Evidence
Fidelity index
Every GGUF file Runner has scored against its publisher's reference weights, one row per measurement, pass or fail. A row reads one file against the reference checkpoint at the revision named, over the window named; the verdict is the bar of the fidelity method, mean KLD at most 0.05 and margin-qualified top-1 at least 97%, which was set over 2,000 positions, so a row over fewer positions is a smaller sample read against the same numbers. Failing rows come first. A side whose report does not say which file it scored is listed as not identified rather than left out.
269 measured sides in 181 reports: 140 pass, 129 fail; the files of 121 are not identified. Method: fidelity-method.md.
| verdict | file | reference | positions | mean KLD | margin-qualified top-1 | top-1 | read with | report |
|---|---|---|---|---|---|---|---|---|
| fail | file not identified (the report calls it lfm) | LiquidAI/LFM2.5-350M at 9e6c6ccf47cd (local checkout) | 300 | 0.9243 | 61.1% | 53.0% | an HTTP endpoint | lfm2.5-reference-2026-10-30/runner-vs-llamacpp-350m-q4_k_m.json#b |
| fail | file not identified (the report calls it score file) | LiquidAI/LFM2.5-350M at 9e6c6ccf47cd (local checkout) | 300 | 0.8744 | 61.9% | 54.3% | runner --score | lfm2.5-reference-2026-10-30/gold-logits-350m-q8-q4.json#b |
| fail | file not identified (the report calls it score file) | LiquidAI/LFM2.5-350M at 9e6c6ccf47cd (local checkout) | 300 | 0.8744 | 61.9% | 54.3% | runner --score | lfm2.5-reference-2026-10-30/runner-vs-llamacpp-350m-q4_k_m.json#a |
| fail | ISTA-DASLab/Qwen3.8-27B-GSQ-RCO-GGUF: Qwen3.8-27B-GSQ-RCO-IQ3_S.gguf | Qwen/Qwen3.8-27B@1d4bf0f2ff6012fd82039f2fa52739d0dd7c60c0 | 300 | 0.5616 | 82.5% | 74.3% | runner --score | r4.32-qwen38-iq3s-reference-2026-11-04/r432-gold-house.json#a |
| fail | file not identified (the report calls it score file) | LiquidAI/LFM2.5-350M-GGUF qad/ at 657e078c9408 (local checkout) | 300 | 0.5275 | 74.2% | 63.0% | runner --score | lfm2.5-reference-2026-10-30/qad-q4_0-350m-against-its-own-parent.json#a |
| fail | file not identified (the report calls it lfm) | LiquidAI/LFM2.5-350M at 9e6c6ccf47cd (local checkout) | 300 | 0.5236 | 74.8% | 65.0% | an HTTP endpoint | lfm2.5-reference-2026-10-30/qad-q4_0-350m-runner-vs-llamacpp.json#b |
| fail | file not identified (the report calls it score file) | LiquidAI/LFM2.5-350M at 9e6c6ccf47cd (local checkout) | 300 | 0.4831 | 76.5% | 67.7% | runner --score | lfm2.5-reference-2026-10-30/qad-q4_0-350m-runner-vs-llamacpp.json#a |
| fail | ISTA-DASLab/Qwen3.8-27B-GSQ-RCO-GGUF: Qwen3.8-27B-GSQ-RCO-IQ3_S.gguf | Qwen/Qwen3.8-27B@1d4bf0f2ff6012fd82039f2fa52739d0dd7c60c0 | 2000 | 0.4718 | 86.4% | 80.8% | runner --score | r4.32-qwen38-iq3s-reference-2026-11-04/r432-gold-d1window.json#a |
| fail | file not identified (the report calls it score file) | LiquidAI/LFM2.5-350M-GGUF qad/ at 657e078c9408 (local checkout) | 300 | 0.4574 | 81.7% | 69.7% | runner --score | lfm2.5-reference-2026-10-30/qad-q4_0-350m-against-its-own-parent.json#b |
| fail | file not identified (the report calls it score file) | google/gemma-4-E2B-it at 3e22461f65e8 (local copy) | 2000 | 0.4374 | 81.1% | 74.4% | runner --score | gemma-4-e2b-ple-quant-2026-10-31/gold-logits-ple-q4k-realistic.json#a |
| fail | file not identified (the report calls it score file) | google/gemma-4-E2B-it at 3e22461f65e8 (local copy) | 2000 | 0.4373 | 81.1% | 74.4% | runner --score | gemma-4-e2b-ple-quant-2026-10-31/gold-logits-ple-q4k-realistic-float32.json#a |
| fail | file not identified (the report calls it score file) | google/gemma-4-E2B-it at 3e22461f65e8 (local copy) | 2000 | 0.4337 | 81.2% | 74.7% | runner --score | gemma-4-e2b-ple-quant-2026-10-31/gold-logits-ple-q4k-realistic.json#b |
| fail | file not identified (the report calls it score file) | google/gemma-4-E2B-it at 3e22461f65e8 (local copy) | 2000 | 0.4336 | 81.2% | 74.7% | runner --score | gemma-4-e2b-ple-quant-2026-10-31/gold-logits-ple-q4k-realistic-float32.json#b |
| fail | file not identified (the report calls it score file) | Qwen/Qwen3.8-27B@1d4bf0f2ff6012fd82039f2fa52739d0dd7c60c0 | 300 | 0.4119 | 85.0% | 77.7% | runner --score | r4.32-qwen38-iq3s-reference-2026-11-04/bf16diag-gold-house.json#b |
| fail | file not identified (the report calls it score file) | Qwen/Qwen3.8-27B@1d4bf0f2ff6012fd82039f2fa52739d0dd7c60c0 | 300 | 0.4119 | 85.0% | 77.7% | runner --score | r4.32-qwen38-iq3s-reference-2026-11-04/r432-gold-house.json#b |
| fail | file not identified (the report calls it score file) | Qwen/Qwen3.8-27B@1d4bf0f2ff6012fd82039f2fa52739d0dd7c60c0 | 2000 | 0.4084 | 86.1% | 80.4% | runner --score | r3.6.3.1-cuda-attn-tc-wide-qwen35-2026-11-04/gold-tcw0-vs-tcw1.json#a |
| fail | file not identified (the report calls it score file) | Qwen/Qwen3.8-27B@1d4bf0f2ff6012fd82039f2fa52739d0dd7c60c0 | 2000 | 0.4084 | 86.1% | 80.4% | runner --score | r4.32-qwen38-iq3s-reference-2026-11-04/bf16diag-gold-d1window.json#b |
| fail | file not identified (the report calls it score file) | Qwen/Qwen3.8-27B@1d4bf0f2ff6012fd82039f2fa52739d0dd7c60c0 | 2000 | 0.4084 | 86.1% | 80.4% | runner --score | r4.32-qwen38-iq3s-reference-2026-11-04/r432-gold-d1window.json#b |
| fail | file not identified (the report calls it score file) | Qwen/Qwen3.8-27B@1d4bf0f2ff6012fd82039f2fa52739d0dd7c60c0 | 2000 | 0.4083 | 86.1% | 80.5% | runner --score | r3.6.3.1-cuda-attn-tc-wide-qwen35-2026-11-04/gold-tcw0-vs-tcw1.json#b |
| fail | file not identified (the report calls it score file) | Accio-Lab/occamy-1.0@72ca9603 | 300 | 0.2578 | 85.7% | 77.0% | runner --score | occamy-1.0-2026-11-06/gold-q4_k_m.json#a |
| fail | file not identified (the report calls it score file) | nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16@a9904d24bcc1d289a1950fa9d2b978c47cf903b9 | 300 | 0.2524 | 84.3% | 72.3% | runner --score | nemotron-3.5-lightning-30b-a3b-2026-11-03/gold-logits-bf16-q8_0-runner-configured-floor.json#b |
| fail | mradermacher/Hemmingway-1-GGUF: Hemmingway-1.Q4_K_M.gguf | HemmingwayAI/Hemmingway-1@faba57469705 | 2000 | 0.2482 | 91.6% | 84.5% | runner --score | hemmingway-reference-2026-11-01/h1-gold-logits-cpu-q4_k_m-cuda-q4_k_m.json#b |
| fail | mradermacher/Hemmingway-1-GGUF: Hemmingway-1.Q4_K_M.gguf | HemmingwayAI/Hemmingway-1@faba57469705 | 2000 | 0.2482 | 91.6% | 84.5% | runner --score | hemmingway-reference-2026-11-01/h1-gold-logits-cuda-q4_k_m-cpu-q8_0.json#a |
| fail | mradermacher/Hemmingway-1-GGUF: Hemmingway-1.Q4_K_M.gguf | HemmingwayAI/Hemmingway-1@faba57469705 | 2000 | 0.2480 | 91.7% | 84.5% | runner --score | hemmingway-reference-2026-11-01/h1-gold-logits-cpu-q4_k_m-cuda-q4_k_m.json#a |
| fail | file not identified (the report calls it score file) | nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16@a9904d24bcc1d289a1950fa9d2b978c47cf903b9 | 300 | 0.2462 | 84.8% | 74.0% | runner --score | nemotron-3.5-lightning-30b-a3b-2026-11-03/gold-logits-bf16-q8_0-runner-configured-floor.json#a |
| fail | file not identified (the report calls it score file) | Qwen/Qwen3.5-4B at 851bf6e806ef (local copy) | 2000 | 0.2399 | 88.5% | 78.8% | runner --score | cuda-nvfp4-tc-2026-10-31/gold-logits-default-vs-tc.json#a |
| fail | file not identified (the report calls it score file) | Qwen/Qwen3.5-4B at 851bf6e806ef (local copy) | 2000 | 0.2398 | 88.5% | 78.8% | runner --score | cuda-nvfp4-tc-2026-10-31/gold-logits-default-vs-tc.json#b |
| fail | LiquidAI/LFM2.5-8B-A1B-GGUF: LFM2.5-8B-A1B-Q4_K_M.gguf (sha256 4923ec14f06b) | LiquidAI/LFM2.5-8B-A1B at 5dd22602c2e9 (local copy) | 300 | 0.2372 | 85.4% | 75.3% | runner --score | lfm2.5-8b-a1b-reference-2026-10-30/gold-logits-8b-a1b-q8-q4.json#b |
| fail | bartowski/Phi-3.5-mini-instruct-GGUF: Phi-3.5-mini-instruct-Q8_0.gguf (sha256 76fbf02f6fe9) | microsoft/Phi-3.5-mini-instruct@2fe19245 | 272 | 0.2071 | 89.1% | 79.8% | an HTTP endpoint | phi-3.5-mini-instruct-2026-11-13/gold-q8_0-w1-llama-c8192.json#b |
| fail | bartowski/Phi-3.5-mini-instruct-GGUF: Phi-3.5-mini-instruct-Q8_0.gguf (sha256 76fbf02f6fe9) | microsoft/Phi-3.5-mini-instruct@2fe19245 | 272 | 0.2052 | 88.5% | 79.0% | runner --score | phi-3.5-mini-instruct-2026-11-13/gold-q8_0-w1-c8192.json#a |
| fail | mradermacher/DeepSeek-V2-Lite-Chat-GGUF: DeepSeek-V2-Lite-Chat.Q4_K_M.gguf | deepseek-ai/DeepSeek-V2-Lite-Chat@85864749cd611b4353ce1decdb286193298f64c7 | 300 | 0.2023 | 89.4% | 79.3% | an HTTP endpoint | deepseek-v2-lite-chat-2026-11-04/gold-q4_k_m-cuda-vs-llamacpp.json#b |
| fail | IFM/K2-Horizon-0.9B-GGUF: K2-Horizon-1B-Q4_K_M.gguf (sha256 78bff592aeba) | IFM/K2-Horizon-0.9B@d50881f4f8dc | 300 | 0.1951 | 88.8% | 74.0% | an HTTP endpoint | k2-horizon-0.9b-reference-2026-10-30/gold-logits-llamacpp-q4_k_m-text.json#a |
| fail | mradermacher/DeepSeek-V2-Lite-Chat-GGUF: DeepSeek-V2-Lite-Chat.Q4_K_M.gguf | deepseek-ai/DeepSeek-V2-Lite-Chat@85864749cd611b4353ce1decdb286193298f64c7 | 300 | 0.1884 | 90.3% | 81.3% | runner --score | deepseek-v2-lite-chat-2026-11-04/gold-q4_k_m-cpu-vs-cuda.json#a |
| fail | mradermacher/DeepSeek-V2-Lite-Chat-GGUF: DeepSeek-V2-Lite-Chat.Q4_K_M.gguf | deepseek-ai/DeepSeek-V2-Lite-Chat@85864749cd611b4353ce1decdb286193298f64c7 | 300 | 0.1876 | 90.3% | 81.7% | runner --score | deepseek-v2-lite-chat-2026-11-04/gold-q4_k_m-cpu-vs-cuda.json#b |
| fail | mradermacher/DeepSeek-V2-Lite-Chat-GGUF: DeepSeek-V2-Lite-Chat.Q4_K_M.gguf | deepseek-ai/DeepSeek-V2-Lite-Chat@85864749cd611b4353ce1decdb286193298f64c7 | 300 | 0.1876 | 90.3% | 81.7% | runner --score | deepseek-v2-lite-chat-2026-11-04/gold-q4_k_m-cuda-vs-llamacpp.json#a |
| fail | file not identified (the report calls it score file) | google/gemma-4-E2B-it at 3e22461f65e8 (local copy) | 2000 | 0.1728 | 90.6% | 83.8% | runner --score | gemma-4-e2b-ple-quant-2026-10-31/gold-logits-unsloth-q4km-q5km.json#a |
| fail | file not identified (the report calls it score file) | google/gemma-4-E2B-it at 3e22461f65e8 (local copy) | 2000 | 0.1728 | 90.6% | 83.7% | runner --score | gemma-4-e2b-ple-quant-2026-10-31/gold-logits-unsloth-q4km-q5km-float32.json#a |
| fail | file not identified (the report calls it score file) | ornith-ai/Ornith-1.5-35B-A3B@10fbf86f | 300 | 0.1709 | 92.1% | 83.7% | runner --score | ornith-1.5-35b-a3b-2026-11-05/gold-q4_k_m.json#a |
| fail | Colby/apertus-v1.5-8b-text-Q4_K_M-GGUF: apertus-v1.5-8b-text-q4_k_m.gguf (sha256 a037df8d87ff) | swiss-ai/Apertus-v1.5-8B@1519de0a | 300 | 0.1686 | 91.7% | 79.0% | runner --score | apertus-v1.5-8b-text-2026-11-13/gold-q4_k_m-w2.json#a |
| fail | Colby/apertus-v1.5-8b-text-Q4_K_M-GGUF: apertus-v1.5-8b-text-q4_k_m.gguf (sha256 a037df8d87ff) | swiss-ai/Apertus-v1.5-8B@1519de0a | 300 | 0.1629 | 87.7% | 75.3% | runner --score | apertus-v1.5-8b-text-2026-11-13/gold-q4_k_m-w1.json#a |
| fail | ggml-org/Qwen3.6-27B-GGUF: Qwen3.6-27B-Q4_K_M.gguf | Qwen/Qwen3.6-27B@6a9e13bd6fc8f0983b9b99948120bc37f49c13e9 | 300 | 0.1539 | 93.8% | 87.0% | an HTTP endpoint | qwen3.6-27b-2026-11-04/gold-logits-q4_k_m-runner-vs-llamacpp.json#b |
| fail | ibm-granite/granite-4.2-30b-GGUF: granite-4.2-30b-Q4_K_M.gguf | ibm-granite/granite-4.2-30b@9e668ce1c538387ef24d3644e9b0606647762636 | 300 | 0.1512 | 92.5% | 80.7% | an HTTP endpoint | granite-4.2-30b-2026-11-04/gold-logits-q4_k_m-runner-vs-llamacpp.json#b |
| fail | file not identified (the report calls it score file) | InternScience/Sage-A1.5@0e3205cf | 300 | 0.1445 | 92.5% | 84.3% | runner --score | sage-a1.5-2026-11-06/gold-q4_k_m.json#a |
| fail | LiquidAI/LFM2.5-24B-A2B-GGUF: LFM2.5-24B-A2B-Q4_K_M.gguf (sha256 d3334405b962) | LiquidAI/LFM2.5-24B-A2B at 3fcfdbfa1a1b (local copy) | 300 | 0.1441 | 90.5% | 84.0% | runner --score | lfm2.5-24b-a2b-reference-2026-10-30/gold-logits-24b-a2b-q8-q4.json#b |
| fail | file not identified (the report calls it score file) | nvidia/NVIDIA-Nemotron-3.5-Bolt-E4B-BF16@49843eed8db1f1c616c47af75cfd36e07aec0667 | 2000 | 0.1427 | 93.5% | 87.6% | runner --score | r3.6.3.1-cuda-attn-tc-wide-2026-11-03/gold-main-vs-r3631.json#a |
| fail | file not identified (the report calls it score file) | nvidia/NVIDIA-Nemotron-3.5-Bolt-E4B-BF16@49843eed8db1f1c616c47af75cfd36e07aec0667 | 2000 | 0.1422 | 93.4% | 87.6% | runner --score | r3.6.3.1-cuda-attn-tc-wide-2026-11-03/gold-main-vs-r3631.json#b |
| fail | ibm-granite/granite-4.2-30b-GGUF: granite-4.2-30b-Q4_K_M.gguf | ibm-granite/granite-4.2-30b@9e668ce1c538387ef24d3644e9b0606647762636 | 300 | 0.1393 | 93.9% | 83.0% | runner --score | granite-4.2-30b-2026-11-04/gold-logits-q4_k_m-runner-vs-llamacpp.json#a |
| fail | ggml-org/NVIDIA-Nemotron-3-Nano-30B-A3B-GGUF: NVIDIA-Nemotron-3-Nano-30B-A3B-Q4_K_M.gguf (sha256 0f111a0d4977) | nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-BF16@bf77c317 | 300 | 0.1366 | 94.5% | 82.7% | runner --score | nemotron-3-nano-30b-a3b-2026-11-05/gold-q4_k_m-configured.json#a |
| fail | file not identified (the report calls it lfm) | LiquidAI/LFM2.5-1.2B-Instruct at 0f604ada3f76 (local checkout) | 300 | 0.1322 | 93.0% | 82.3% | an HTTP endpoint | lfm2.5-reference-2026-10-30/qad-q4_0-1.2b-runner-vs-llamacpp.json#b |
| fail | file not identified (the report calls it score file) | Qwen/Qwen3-4B at 1cfa9a720891 (local copy) | 2000 | 0.1312 | 92.0% | 83.8% | runner --score | cpu-kquant-one-accumulator-chunked-2026-10-31/gold-logits-pre-vs-commit.json#b |
| fail | file not identified (the report calls it score file) | Qwen/Qwen3-4B at 1cfa9a720891 (local copy) | 2000 | 0.1312 | 92.0% | 83.8% | runner --score | cpu-kquant-one-accumulator-chunked-2026-10-31/gold-logits-pre-vs-main.json#b |
| fail | file not identified (the report calls it score file) | Qwen/Qwen3-4B at 1cfa9a720891 (local copy) | 2000 | 0.1312 | 92.0% | 83.8% | runner --score | cpu-kquant-one-accumulator-chunked-2026-10-31/gold-logits-pre-vs-commit.json#a |
| fail | file not identified (the report calls it score file) | Qwen/Qwen3-4B at 1cfa9a720891 (local copy) | 2000 | 0.1312 | 92.0% | 83.8% | runner --score | cpu-kquant-one-accumulator-chunked-2026-10-31/gold-logits-pre-vs-main.json#a |
| fail | ggml-org/Qwen3.6-27B-GGUF: Qwen3.6-27B-Q4_K_M.gguf | Qwen/Qwen3.6-27B@6a9e13bd6fc8f0983b9b99948120bc37f49c13e9 | 300 | 0.1288 | 95.9% | 89.3% | runner --score | qwen3.6-27b-2026-11-04/gold-logits-q4_k_m-runner-vs-llamacpp.json#a |
| fail | IFM/K2-Horizon-7B-GGUF: K2-Horizon-7B-Q4_K_M.gguf (sha256 eb89c15a0ae9) | IFM/K2-Horizon-7B@4622f94a6632 | 300 | 0.1284 | 95.3% | 82.7% | an HTTP endpoint | k2-horizon-7b-reference-2026-10-30/gold-logits-llamacpp-q4_k_m-text.json#a |
| fail | IFM/K2-Horizon-7B-GGUF: K2-Horizon-7B-Q4_K_M.gguf (sha256 eb89c15a0ae9) | IFM/K2-Horizon-7B@4622f94a6632 | 300 | 0.1284 | 95.3% | 82.7% | an HTTP endpoint | k2-horizon-7b-reference-2026-10-30/gold-logits-llamacpp-q4_k_m.json#a |
| fail | file not identified (the report calls it score file) | LiquidAI/LFM2.5-1.2B-Instruct at 0f604ada3f76 (local checkout) | 300 | 0.1282 | 93.4% | 84.0% | runner --score | lfm2.5-reference-2026-10-30/qad-q4_0-1.2b-runner-vs-llamacpp.json#a |
| fail | file not identified (the report calls it score file) | empero-ai/Qwen3.8-35B-A3B-Distill@bcc2dbe2 | 300 | 0.1278 | 94.6% | 85.0% | runner --score | empero-qwen3.8-35b-a3b-distill-2026-11-06/gold-q4_k_m.json#a |
| fail | HuggingFaceTB/SmolLM2-1.7B-Instruct-GGUF: smollm2-1.7b-instruct-q4_k_m.gguf | HuggingFaceTB/SmolLM2-1.7B-Instruct@31b70e2e | 300 | 0.1248 | 92.3% | 85.0% | an HTTP endpoint | smollm2-2026-11-13/gold-1.7b-publisher-q4_k_m-w1-runner-vs-llamacpp.json#b |
| fail | XHToken/Spark-X2.6-4B-GGUF: Spark-X2.6-4B-Q4_K_M.gguf | XHToken/Spark-X2.6-4B@d1b5f260 | 300 | 0.1232 | 91.5% | 85.0% | an HTTP endpoint | spark-2026-10-30/gold-logits-llamacpp-x2.6-q4_k_m-text.json#a |
| fail | XHToken/Spark-X2.6-4B-GGUF: Spark-X2.6-4B-Q4_K_M.gguf | XHToken/Spark-X2.6-4B@d1b5f260 | 300 | 0.1232 | 91.5% | 85.0% | an HTTP endpoint | spark-2026-10-30/gold-logits-llamacpp-x2.6-q4_k_m.json#a |
| fail | IFM/K2-Horizon-7B-GGUF: K2-Horizon-7B-Q4_K_M.gguf (sha256 eb89c15a0ae9) | IFM/K2-Horizon-7B at 4622f94a (local checkout) | 300 | 0.1225 | 95.8% | 84.3% | runner --score | k2-horizon-7b-reference-2026-10-30/gold-logits-cpu-q8-q4.json#b |
| fail | openbmb/MiniCPM5-2B-GGUF: MiniCPM5-2B-Q4_K_M.gguf (sha256 ec2d58016400) | openbmb/MiniCPM5-2B@f97400052a43 | 300 | 0.1199 | 94.1% | 86.0% | an HTTP endpoint | minicpm5-2b-reference-2026-10-29/gold-logits-llamacpp-q4_k_m-text.json#a |
| fail | file not identified (the report calls it Ornith-1.5-35B-Q6_K) | ornith-ai/Ornith-1.5-35B-A3B@10fbf86f | 300 | 0.1191 | 93.8% | 86.0% | an HTTP endpoint | ornith-1.5-35b-a3b-2026-11-05/gold-q6_k-runner-vs-llamacpp.json#b |
| fail | file not identified (the report calls it lfm) | LiquidAI/LFM2.5-1.2B-Instruct at 0f604ada3f76 (local checkout) | 300 | 0.1189 | 93.0% | 84.3% | an HTTP endpoint | lfm2.5-reference-2026-10-30/runner-vs-llamacpp-1.2b-q4_k_m.json#b |
| fail | mradermacher/stablelm-2-1_6b-chat-GGUF: stablelm-2-1_6b-chat.Q4_K_M.gguf | stabilityai/stablelm-2-1_6b-chat@f3fe6705 | 272 | 0.1185 | 95.7% | 82.4% | runner --score | stablelm-2-1.6b-chat-2026-11-13/gold-q4_k_m-w1.json#a |
| fail | XHToken/Spark-X2.5-4B-GGUF: Spark-X2.5-4B-Q4_K_M.gguf | XHToken/Spark-X2.5-4B@0bcb3567 | 300 | 0.1177 | 95.2% | 87.0% | an HTTP endpoint | spark-2026-10-30/gold-logits-llamacpp-x2.5-q4_k_m-text.json#a |
| fail | XHToken/Spark-X2.5-4B-GGUF: Spark-X2.5-4B-Q4_K_M.gguf | XHToken/Spark-X2.5-4B@0bcb3567 | 300 | 0.1177 | 95.2% | 87.0% | an HTTP endpoint | spark-2026-10-30/gold-logits-llamacpp-x2.5-q4_k_m.json#a |
| fail | IFM/K2-Horizon-0.9B-GGUF: K2-Horizon-1B-Q4_K_M.gguf (sha256 78bff592aeba) | IFM/K2-Horizon-0.9B@d50881f4f8dc | 300 | 0.1174 | 96.1% | 82.7% | an HTTP endpoint | k2-horizon-0.9b-reference-2026-10-30/gold-logits-llamacpp-q4_k_m.json#a |
| fail | file not identified (the report calls it score file) | LiquidAI/LFM2.5-1.2B-Instruct at 0f604ada3f76 (local checkout) | 300 | 0.1153 | 94.7% | 85.7% | runner --score | lfm2.5-reference-2026-10-30/gold-logits-1.2b-q8-q4.json#b |
| fail | file not identified (the report calls it score file) | LiquidAI/LFM2.5-1.2B-Instruct at 0f604ada3f76 (local checkout) | 300 | 0.1153 | 94.7% | 85.7% | runner --score | lfm2.5-reference-2026-10-30/runner-vs-llamacpp-1.2b-q4_k_m.json#a |
| fail | XHToken/Spark-X2.5-4B-GGUF: Spark-X2.5-4B-Q4_K_M.gguf | XHToken/Spark-X2.5-4B at 0bcb3567 (local checkout) | 300 | 0.1136 | 95.6% | 87.0% | runner --score | spark-2026-10-30/gold-logits-cpu-x2.5-q8-q4.json#b |
| fail | XHToken/Spark-X2.6-4B-GGUF: Spark-X2.6-4B-Q4_K_M.gguf | XHToken/Spark-X2.6-4B at d1b5f260 (local checkout) | 300 | 0.1130 | 93.6% | 87.0% | runner --score | spark-2026-10-30/gold-logits-cpu-x2.6-q8-q4.json#b |
| fail | openbmb/MiniCPM5-2B-GGUF: MiniCPM5-2B-Q4_K_M.gguf (sha256 ec2d58016400) | openbmb/MiniCPM5-2B@f97400052a43 | 300 | 0.1079 | 93.7% | 84.7% | an HTTP endpoint | minicpm5-2b-reference-2026-10-29/gold-logits-llamacpp-q4_k_m.json#a |
| fail | IFM/K2-Horizon-0.9B-GGUF: K2-Horizon-1B-Q4_K_M.gguf (sha256 78bff592aeba) | IFM/K2-Horizon-0.9B at d50881f4f8dc6923fb8f301e1c2313fe3a2b6c1d (local checkout) | 300 | 0.1078 | 97.1% | 81.7% | runner --score | k2-horizon-0.9b-reference-2026-10-30/gold-logits-metal-q8-q4.json#b |
| fail | file not identified (the report calls it score file) | IFM/K2-Horizon-0.9B at d50881f4f8dc6923fb8f301e1c2313fe3a2b6c1d (local checkout) | 2000 | 0.1050 | 96.0% | 84.9% | runner --score | r3.24b-metal-microbatch-tiles-2026-11-02/gold-logits-q4_k_m-solo-vs-tiles.json#a |
| fail | file not identified (the report calls it score file) | IFM/K2-Horizon-0.9B at d50881f4f8dc6923fb8f301e1c2313fe3a2b6c1d (local checkout) | 2000 | 0.1049 | 95.8% | 84.8% | runner --score | r3.24b-metal-microbatch-tiles-2026-11-02/gold-logits-q4_k_m-solo-vs-tiles.json#b |
| fail | mradermacher/stablelm-2-1_6b-chat-GGUF: stablelm-2-1_6b-chat.Q4_K_M.gguf | stabilityai/stablelm-2-1_6b-chat@f3fe6705 | 272 | 0.1043 | 92.3% | 81.2% | runner --score | stablelm-2-1.6b-chat-2026-11-13/gold-q4_k_m-w2.json#a |
| fail | HuggingFaceTB/SmolLM2-1.7B-Instruct-GGUF: smollm2-1.7b-instruct-q4_k_m.gguf | HuggingFaceTB/SmolLM2-1.7B-Instruct@31b70e2e | 299 | 0.1033 | 92.7% | 84.6% | runner --score | smollm2-2026-11-13/gold-1.7b-publisher-q4_k_m-w1.json#a |
| fail | file not identified (the report calls it score file) | ornith-ai/Ornith-1.5-35B-A3B@10fbf86f | 300 | 0.1033 | 95.8% | 89.0% | runner --score | ornith-1.5-35b-a3b-2026-11-05/gold-q6_k-runner-vs-llamacpp.json#a |
| fail | file not identified (the report calls it score file) | ornith-ai/Ornith-1.5-35B-A3B@10fbf86f | 300 | 0.1033 | 95.8% | 89.0% | runner --score | ornith-1.5-35b-a3b-2026-11-05/gold-q6_k.json#a |
| fail | HuggingFaceTB/SmolLM2-1.7B-Instruct-GGUF: smollm2-1.7b-instruct-q4_k_m.gguf | HuggingFaceTB/SmolLM2-1.7B-Instruct@31b70e2e | 300 | 0.1032 | 92.8% | 84.7% | runner --score | smollm2-2026-11-13/gold-1.7b-publisher-q4_k_m-w1-runner-vs-llamacpp.json#a |
| fail | openbmb/MiniCPM5-2B-GGUF: MiniCPM5-2B-Q4_K_M.gguf (sha256 ec2d58016400) | openbmb/MiniCPM5-2B at f97400052a43d642bbc6e9975e2397e3ae6a6b52 (local checkout) | 300 | 0.1020 | 94.6% | 86.0% | runner --score | minicpm5-2b-reference-2026-10-29/gold-logits-metal-q8-q4.json#b |
| fail | HuggingFaceTB/SmolLM2-1.7B-Instruct-GGUF: smollm2-1.7b-instruct-q4_k_m.gguf | HuggingFaceTB/SmolLM2-1.7B-Instruct@31b70e2e | 296 | 0.1008 | 96.1% | 84.8% | runner --score | smollm2-2026-11-13/gold-1.7b-publisher-q4_k_m-w2.json#a |
| fail | file not identified (the report calls it score file) | Accio-Lab/occamy-1.0@72ca9603 | 300 | 0.0970 | 95.8% | 87.3% | runner --score | occamy-1.0-2026-11-06/gold-q5_k_m.json#a |
| fail | Qwen/Qwen2.5-1.5B-Instruct-GGUF: qwen2.5-1.5b-instruct-q4_k_m.gguf | Qwen/Qwen2.5-1.5B-Instruct@989aa798 | 300 | 0.0931 | 97.5% | 86.0% | an HTTP endpoint | qwen2.5-1.5b-instruct-2026-11-13/gold-q4_k_m-publisher-w1-runner-vs-llamacpp.json#b |
| fail | file not identified (the report calls it score file) | ibm-granite/granite-4.0-h-micro at d5f01a3ea75f (local copy) | 2000 | 0.0918 | 94.3% | 84.3% | runner --score | r3.24-metal-small-prefill-2026-10-31/gold-w16-granite.json#b |
| fail | file not identified (the report calls it score file) | ibm-granite/granite-4.0-h-micro at d5f01a3ea75f (local copy) | 2000 | 0.0918 | 94.3% | 84.3% | runner --score | r3.24-metal-small-prefill-2026-10-31/gold-w8-granite.json#b |
| fail | file not identified (the report calls it score file) | ibm-granite/granite-4.0-h-micro at d5f01a3ea75f (local copy) | 2000 | 0.0918 | 94.3% | 84.3% | runner --score | granitehybrid-metal-2026-10-31/gold-logits-cpu-vs-metal.json#a |
| fail | file not identified (the report calls it score file) | ibm-granite/granite-4.0-h-micro at d5f01a3ea75f (local copy) | 2000 | 0.0918 | 94.3% | 84.3% | runner --score | granitehybrid-metal-2026-10-31/gold-logits-cpu-vs-metal.json#b |
| fail | file not identified (the report calls it score file) | ibm-granite/granite-4.0-h-micro at d5f01a3ea75f (local copy) | 2000 | 0.0918 | 94.3% | 84.3% | runner --score | r3.24-metal-small-prefill-2026-10-31/gold-w16-granite.json#a |
| fail | file not identified (the report calls it score file) | ibm-granite/granite-4.0-h-micro at d5f01a3ea75f (local copy) | 2000 | 0.0918 | 94.3% | 84.3% | runner --score | r3.24-metal-small-prefill-2026-10-31/gold-w8-granite.json#a |
| fail | IFM/K2-Horizon-0.9B-GGUF: K2-Horizon-1B-Q8_0.gguf (sha256 741fb9ad2639) | IFM/K2-Horizon-0.9B@d50881f4f8dc | 300 | 0.0911 | 96.1% | 84.0% | an HTTP endpoint | k2-horizon-0.9b-reference-2026-10-30/gold-logits-llamacpp-q8_0-text.json#a |
| fail | file not identified (the report calls it score file) | Qwen/Qwen3-30B-A3B@ad44e777bcd18fa416d9da3bd8f70d33ebb85d39 | 2000 | 0.0909 | 94.8% | 85.7% | runner --score | qwen3-30b-a3b-derive-2026-11-02/gold-logits-official-q4_k_m-selective.json#a |
| fail | ibm-granite/granite-4.0-h-micro-GGUF: granite-4.0-h-micro-Q4_K_M.gguf | ibm-granite/granite-4.0-h-micro@d5f01a3e | 272 | 0.0904 | 94.3% | 86.4% | runner --score | granite-4.0-h-micro-2026-11-13/gold-q4_k_m-w2.json#a |
| fail | ibm-granite/granite-4.0-h-micro-GGUF: granite-4.0-h-micro-Q4_K_M.gguf | ibm-granite/granite-4.0-h-micro@d5f01a3e | 272 | 0.0888 | 95.0% | 86.4% | an HTTP endpoint | granite-4.0-h-micro-2026-11-13/gold-q4_k_m-w1-llama.json#b |
| fail | mradermacher/DeepSeek-V2-Lite-Chat-GGUF: DeepSeek-V2-Lite-Chat.Q5_K_M.gguf | deepseek-ai/DeepSeek-V2-Lite-Chat@85864749cd611b4353ce1decdb286193298f64c7 | 300 | 0.0861 | 96.3% | 86.7% | an HTTP endpoint | deepseek-v2-lite-chat-2026-11-04/gold-q5_k_m-cuda-vs-llamacpp.json#b |
| fail | Qwen/Qwen2.5-1.5B-Instruct-GGUF: qwen2.5-1.5b-instruct-q4_k_m.gguf | Qwen/Qwen2.5-1.5B-Instruct@989aa798 | 300 | 0.0846 | 97.5% | 86.7% | runner --score | qwen2.5-1.5b-instruct-2026-11-13/gold-q4_k_m-publisher-w1-runner-vs-llamacpp.json#a |
| fail | Qwen/Qwen2.5-1.5B-Instruct-GGUF: qwen2.5-1.5b-instruct-q4_k_m.gguf | Qwen/Qwen2.5-1.5B-Instruct@989aa798 | 300 | 0.0846 | 97.5% | 86.7% | runner --score | qwen2.5-1.5b-instruct-2026-11-13/gold-q4_k_m-publisher-w1.json#a |
| fail | file not identified (the report calls it score file) | Accio-Lab/occamy-1.0@72ca9603 | 300 | 0.0836 | 95.4% | 88.7% | runner --score | occamy-1.0-2026-11-06/gold-q6_k.json#a |
| fail | ibm-granite/granite-4.0-h-micro-GGUF: granite-4.0-h-micro-Q4_K_M.gguf | ibm-granite/granite-4.0-h-micro@d5f01a3e | 272 | 0.0805 | 95.5% | 87.5% | runner --score | granite-4.0-h-micro-2026-11-13/gold-q4_k_m-w1.json#a |
| fail | ibm-granite/granite-4.0-h-micro-GGUF: granite-4.0-h-micro-Q4_K_M.gguf | ibm-granite/granite-4.0-h-micro@d5f01a3e | 272 | 0.0805 | 95.5% | 87.5% | runner --score | granite-4.0-h-micro-2026-11-13/gold-q4_k_m-w1-llama.json#a |
| fail | ggml-org/gpt-oss-20b-GGUF: gpt-oss-20b-MXFP4.gguf (sha256 27cd6c432c76) | openai/gpt-oss-20b@6cee5e81 | 272 | 0.0783 | 98.4% | 96.0% | runner --score | gpt-oss-20b-2026-11-13/gold-mxfp4-w2.json#a |
| fail | Qwen/Qwen2.5-1.5B-Instruct-GGUF: qwen2.5-1.5b-instruct-q4_k_m.gguf | Qwen/Qwen2.5-1.5B-Instruct@989aa798 | 300 | 0.0750 | 96.4% | 86.7% | runner --score | qwen2.5-1.5b-instruct-2026-11-13/gold-q4_k_m-publisher-w2.json#a |
| fail | bartowski/MiMo-V2.6-Distill-Qwen-9B-GGUF: MiMo-V2.6-Distill-Qwen-9B-Q4_K_M.gguf | XiaomiMiMo/MiMo-V2.6-Distill-Qwen-9B (local checkout) | 300 | 0.0717 | 98.2% | 91.7% | runner --score | mimo-v2.6-distill-qwen-9b-2026-10-29/gold-logits-cpu-q8-q4.json#b |
| fail | mradermacher/stablelm-2-1_6b-chat-i1-GGUF: stablelm-2-1_6b-chat.i1-Q4_K_M.gguf | stabilityai/stablelm-2-1_6b-chat@f3fe6705 | 272 | 0.0705 | 98.4% | 87.1% | runner --score | stablelm-2-1.6b-chat-2026-11-13/gold-i1-q4_k_m-w1.json#a |
| fail | mradermacher/DeepSeek-V2-Lite-Chat-GGUF: DeepSeek-V2-Lite-Chat.Q5_K_M.gguf | deepseek-ai/DeepSeek-V2-Lite-Chat@85864749cd611b4353ce1decdb286193298f64c7 | 300 | 0.0705 | 96.3% | 88.0% | runner --score | deepseek-v2-lite-chat-2026-11-04/gold-q5_k_m-cpu-vs-cuda.json#b |
| fail | mradermacher/DeepSeek-V2-Lite-Chat-GGUF: DeepSeek-V2-Lite-Chat.Q5_K_M.gguf | deepseek-ai/DeepSeek-V2-Lite-Chat@85864749cd611b4353ce1decdb286193298f64c7 | 300 | 0.0705 | 96.3% | 88.0% | runner --score | deepseek-v2-lite-chat-2026-11-04/gold-q5_k_m-cuda-vs-llamacpp.json#a |
| fail | mradermacher/DeepSeek-V2-Lite-Chat-GGUF: DeepSeek-V2-Lite-Chat.Q5_K_M.gguf | deepseek-ai/DeepSeek-V2-Lite-Chat@85864749cd611b4353ce1decdb286193298f64c7 | 300 | 0.0704 | 96.8% | 88.3% | runner --score | deepseek-v2-lite-chat-2026-11-04/gold-q5_k_m-cpu-vs-cuda.json#a |
| fail | ggml-org/gpt-oss-20b-GGUF: gpt-oss-20b-MXFP4.gguf (sha256 27cd6c432c76) | openai/gpt-oss-20b@6cee5e81 | 272 | 0.0671 | 98.0% | 94.9% | an HTTP endpoint | gpt-oss-20b-2026-11-13/gold-mxfp4-w1-llama.json#b |
| fail | ggml-org/gpt-oss-20b-GGUF: gpt-oss-20b-MXFP4.gguf (sha256 27cd6c432c76) | openai/gpt-oss-20b@6cee5e81 | 272 | 0.0634 | 97.6% | 93.8% | runner --score | gpt-oss-20b-2026-11-13/gold-mxfp4-w1-llama.json#a |
| fail | ggml-org/gpt-oss-20b-GGUF: gpt-oss-20b-MXFP4.gguf (sha256 27cd6c432c76) | openai/gpt-oss-20b@6cee5e81 | 272 | 0.0634 | 97.6% | 93.8% | runner --score | gpt-oss-20b-2026-11-13/gold-mxfp4-w1.json#a |
| fail | ggml-org/Qwen3.6-27B-GGUF: Qwen3.6-27B-Q8_0.gguf | Qwen/Qwen3.6-27B@6a9e13bd6fc8f0983b9b99948120bc37f49c13e9 | 300 | 0.0626 | 98.3% | 96.3% | an HTTP endpoint | qwen3.6-27b-2026-11-04/gold-logits-q8_0-runner-vs-llamacpp.json#b |
| fail | bartowski/SmolLM2-135M-Instruct-GGUF: SmolLM2-135M-Instruct-Q4_K_M.gguf | HuggingFaceTB/SmolLM2-135M-Instruct@12fd25f7 | 296 | 0.0625 | 98.0% | 87.8% | runner --score | smollm2-2026-11-13/gold-135m-q4_k_m-w2.json#a |
| fail | mistralai/Ministral-3-3B-Instruct-2512-GGUF: Ministral-3-3B-Instruct-2512-Q4_K_M.gguf | mistralai/Ministral-3-3B-Instruct-2512@b35d4dfe56c1 | 300 | 0.0623 | 99.5% | 92.0% | runner --score | ministral-3-3b-reference-2026-11-01/gold-logits-cpu-q4_k_m-cpu-q8_0.json#a |
| fail | bartowski/SmolLM2-135M-Instruct-GGUF: SmolLM2-135M-Instruct-Q4_K_M.gguf | HuggingFaceTB/SmolLM2-135M-Instruct@12fd25f7 | 299 | 0.0621 | 97.9% | 89.6% | runner --score | smollm2-2026-11-13/gold-135m-q4_k_m-w1.json#a |
| fail | mradermacher/stablelm-2-1_6b-chat-i1-GGUF: stablelm-2-1_6b-chat.i1-Q4_K_M.gguf | stabilityai/stablelm-2-1_6b-chat@f3fe6705 | 272 | 0.0619 | 96.4% | 89.0% | runner --score | stablelm-2-1.6b-chat-2026-11-13/gold-i1-q4_k_m-w2.json#a |
| fail | file not identified (the report calls it Qwen_Qwen3.6-35B-A3B-Q8_0.gguf) | Qwen/Qwen3.6-35B-A3B at 995ad96eacd9 (local copy) | 300 | 0.0618 | 98.1% | 94.0% | an HTTP endpoint | qwen3.6-35b-a3b-reference-2026-10-31/gold-logits-q8-runner-vs-other-engine.json#b |
| fail | file not identified (the report calls it score file) | openbmb/MiniCPM-V-4.7-35B-A3B@98e79ba9 (its text_config) | 300 | 0.0594 | 95.4% | 88.3% | runner --score | minicpm-v-4.7-35b-a3b-2026-11-06/gold-q4_k_m.json#a |
| fail | file not identified (the report calls it score file) | Accio-Lab/occamy-1.0@72ca9603 | 300 | 0.0590 | 98.7% | 93.0% | runner --score | occamy-1.0-2026-11-06/gold-q8_0.json#a |
| fail | file not identified (the report calls it LFM2.5-8B-A1B-Q8_0.gguf) | LiquidAI/LFM2.5-8B-A1B at 5dd22602c2e9 (local copy) | 300 | 0.0572 | 96.7% | 90.0% | an HTTP endpoint | lfm2.5-8b-a1b-reference-2026-10-30/gold-logits-8b-a1b-bf16-vs-other-engine-q8.json#b |
| fail | bartowski/Qwen2.5-1.5B-Instruct-GGUF: Qwen2.5-1.5B-Instruct-Q4_K_M.gguf | Qwen/Qwen2.5-1.5B-Instruct@989aa798 | 300 | 0.0556 | 97.0% | 86.0% | runner --score | qwen2.5-1.5b-instruct-2026-11-13/gold-q4_k_m-bartowski-w1.json#a |
| fail | unsloth/granite-4.0-h-micro-GGUF: granite-4.0-h-micro-Q4_K_M.gguf (sha256 1f21ba20ca24) | ibm-granite/granite-4.0-h-micro@d5f01a3e | 272 | 0.0553 | 97.6% | 90.8% | runner --score | granite-4.0-h-micro-2026-11-13/gold-unsloth-q4_k_m-w2.json#a |
| fail | file not identified (the report calls it score file) | google/gemma-4-E2B-it at 3e22461f65e8 (local copy) | 2000 | 0.0539 | 96.8% | 90.8% | runner --score | gemma-4-e2b-ple-quant-2026-10-31/gold-logits-unsloth-q4km-q5km-float32.json#b |
| fail | file not identified (the report calls it score file) | google/gemma-4-E2B-it at 3e22461f65e8 (local copy) | 2000 | 0.0539 | 96.8% | 90.8% | runner --score | gemma-4-e2b-ple-quant-2026-10-31/gold-logits-unsloth-q4km-q5km.json#b |
| fail | bartowski/swiss-ai_Apertus-8B-Instruct-2509-GGUF: swiss-ai_Apertus-8B-Instruct-2509-Q4_K_M.gguf (sha256 5901007f15ae) | swiss-ai/Apertus-8B-Instruct-2509@b946d404 | 300 | 0.0526 | 98.1% | 88.3% | runner --score | apertus-8b-instruct-2509-2026-11-12/gold-q4_k_m-w2.json#a |
| fail | file not identified (the report calls it score file) | empero-ai/Qwen3.8-35B-A3B-Distill@bcc2dbe2 | 300 | 0.0508 | 95.5% | 89.3% | runner --score | empero-qwen3.8-35b-a3b-distill-2026-11-06/gold-q5_k_m.json#a |
| fail | mradermacher/DeepSeek-V2-Lite-Chat-GGUF: DeepSeek-V2-Lite-Chat.Q6_K.gguf | deepseek-ai/DeepSeek-V2-Lite-Chat@85864749cd611b4353ce1decdb286193298f64c7 | 300 | 0.0466 | 96.8% | 88.7% | an HTTP endpoint | deepseek-v2-lite-chat-2026-11-04/dcause-gold-naivecpu-vs-llamacpp.json#b |
| fail | mradermacher/DeepSeek-V2-Lite-Chat-GGUF: DeepSeek-V2-Lite-Chat.Q6_K.gguf | deepseek-ai/DeepSeek-V2-Lite-Chat@85864749cd611b4353ce1decdb286193298f64c7 | 300 | 0.0466 | 96.8% | 88.7% | an HTTP endpoint | deepseek-v2-lite-chat-2026-11-04/gold-q6_k-cuda-vs-llamacpp.json#b |
| pass | ggml-org/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-GGUF: NVIDIA-Nemotron-3.5-Lightning-30B-A3B-Q4_0.gguf | nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16@a9904d24bcc1d289a1950fa9d2b978c47cf903b9 | 300 | 0.0499 | 99.0% | 90.3% | an HTTP endpoint | nemotron-3.5-lightning-30b-a3b-2026-11-03/gold-logits-q4_0-runner-vs-llamacpp.json#b |
| pass | bartowski/SmolLM2-1.7B-Instruct-GGUF: SmolLM2-1.7B-Instruct-Q4_K_M.gguf (sha256 77665ea48159) | HuggingFaceTB/SmolLM2-1.7B-Instruct@31b70e2e | 296 | 0.0495 | 99.5% | 88.2% | runner --score | smollm2-2026-11-13/gold-1.7b-q4_k_m-w2.json#a |
| pass | bartowski/SmolLM2-1.7B-Instruct-GGUF: SmolLM2-1.7B-Instruct-Q4_K_M.gguf (sha256 77665ea48159) | HuggingFaceTB/SmolLM2-1.7B-Instruct@31b70e2e | 299 | 0.0493 | 97.7% | 91.0% | runner --score | smollm2-2026-11-13/gold-1.7b-q4_k_m-w1.json#a |
| pass | file not identified (the report calls it score file) | ornith-ai/Ornith-1.5-35B-A3B@10fbf86f | 300 | 0.0484 | 97.5% | 91.3% | runner --score | ornith-1.5-35b-a3b-2026-11-05/gold-bf16-vs-q8_0.json#b |
| pass | file not identified (the report calls it score file) | ornith-ai/Ornith-1.5-35B-A3B@10fbf86f | 300 | 0.0484 | 97.5% | 91.3% | runner --score | ornith-1.5-35b-a3b-2026-11-05/gold-q8_0.json#a |
| pass | ggml-org/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-GGUF: NVIDIA-Nemotron-3.5-Lightning-30B-A3B-Q4_0.gguf | nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16@a9904d24bcc1d289a1950fa9d2b978c47cf903b9 | 300 | 0.0476 | 97.6% | 89.3% | runner --score | nemotron-3.5-lightning-30b-a3b-2026-11-03/gold-logits-q4_0-runner-vs-llamacpp.json#a |
| pass | bartowski/Qwen2.5-1.5B-Instruct-GGUF: Qwen2.5-1.5B-Instruct-Q4_K_M.gguf | Qwen/Qwen2.5-1.5B-Instruct@989aa798 | 300 | 0.0476 | 99.0% | 90.0% | runner --score | qwen2.5-1.5b-instruct-2026-11-13/gold-q4_k_m-bartowski-w2.json#a |
| pass | bartowski/Qwen_Qwen3.6-35B-A3B-GGUF: Qwen_Qwen3.6-35B-A3B-Q4_K_M.gguf (sha256 b46fedd33e0b) | Qwen/Qwen3.6-35B-A3B at 995ad96eacd9 (local copy) | 300 | 0.0475 | 97.2% | 89.0% | runner --score | qwen3.6-35b-a3b-reference-2026-10-31/gold-logits-q8-q4.json#b |
| pass | bartowski/swiss-ai_Apertus-8B-Instruct-2509-GGUF: swiss-ai_Apertus-8B-Instruct-2509-Q4_K_M.gguf (sha256 5901007f15ae) | swiss-ai/Apertus-8B-Instruct-2509@b946d404 | 300 | 0.0459 | 98.0% | 89.0% | runner --score | apertus-8b-instruct-2509-2026-11-12/gold-q4_k_m-w1.json#a |
| pass | file not identified (the report calls it LFM2.5-8B-A1B-Q8_0.gguf) | LiquidAI/LFM2.5-8B-A1B at 5dd22602c2e9 (local copy) | 300 | 0.0442 | 97.6% | 93.7% | runner --score | lfm2.5-8b-a1b-reference-2026-10-30/gold-logits-8b-a1b-q8-q4.json#a |
| pass | unsloth/granite-4.0-h-micro-GGUF: granite-4.0-h-micro-Q4_K_M.gguf (sha256 1f21ba20ca24) | ibm-granite/granite-4.0-h-micro@d5f01a3e | 272 | 0.0426 | 99.0% | 93.0% | runner --score | granite-4.0-h-micro-2026-11-13/gold-unsloth-q4_k_m-w1.json#a |
| pass | bartowski/Phi-3.5-mini-instruct-GGUF: Phi-3.5-mini-instruct-Q4_K_M.gguf (sha256 e4165e3a71af) | microsoft/Phi-3.5-mini-instruct@2fe19245 | 272 | 0.0424 | 99.0% | 91.9% | runner --score | phi-3.5-mini-instruct-2026-11-13/gold-q4_k_m-w1.json#a |
| pass | file not identified (the report calls it score file) | Qwen/Qwen3-30B-A3B at ad44e777bcd1 (local copy) | 2000 | 0.0424 | 98.3% | 90.9% | runner --score | qwen3-30b-a3b-imatrix-2026-10-31/gold-logits-n11-vs-imx.json#a |
| pass | bartowski/Phi-3.5-mini-instruct-GGUF: Phi-3.5-mini-instruct-Q4_K_M.gguf (sha256 e4165e3a71af) | microsoft/Phi-3.5-mini-instruct@2fe19245 | 272 | 0.0421 | 99.0% | 90.8% | runner --score | phi-3.5-mini-instruct-2026-11-13/gold-q4_k_m-w2.json#a |
| pass | openbmb/MiniCPM5-2B-GGUF: MiniCPM5-2B-Q8_0.gguf (sha256 c5415f8989bf) | openbmb/MiniCPM5-2B@f97400052a43 | 300 | 0.0401 | 98.0% | 92.7% | an HTTP endpoint | minicpm5-2b-reference-2026-10-29/gold-logits-llamacpp-q8_0-text.json#a |
| pass | file not identified (the report calls it score file) | InternScience/Sage-A1.5@0e3205cf | 300 | 0.0395 | 97.9% | 95.3% | runner --score | sage-a1.5-2026-11-06/gold-q6_k.json#a |
| pass | file not identified (the report calls it score file) | Qwen/Qwen3-30B-A3B at ad44e777bcd1 (local copy) | 2000 | 0.0377 | 98.9% | 91.6% | runner --score | r4.7.3-sensitivity-sweep-2026-10-31/gold-k16.json#a |
| pass | file not identified (the report calls it score file) | Qwen/Qwen3-30B-A3B at ad44e777bcd1 (local copy) | 2000 | 0.0355 | 98.6% | 91.5% | runner --score | qwen3-30b-a3b-imatrix-2026-10-31/gold-logits-n11-vs-imx.json#b |
| pass | file not identified (the report calls it score file) | Qwen/Qwen3-30B-A3B at ad44e777bcd1 (local copy) | 2000 | 0.0355 | 98.6% | 91.5% | runner --score | qwen3-30b-a3b-sensitivity-plan-2026-10-31/gold-logits-n11-vs-sens.json#a |
| pass | file not identified (the report calls it Qwen_Qwen3.6-35B-A3B-Q8_0.gguf) | Qwen/Qwen3.6-35B-A3B at 995ad96eacd9 (local copy) | 300 | 0.0349 | 99.1% | 94.7% | runner --score | qwen3.6-35b-a3b-reference-2026-10-31/gold-logits-q8-q4.json#a |
| pass | file not identified (the report calls it Qwen_Qwen3.6-35B-A3B-Q8_0.gguf) | Qwen/Qwen3.6-35B-A3B at 995ad96eacd9 (local copy) | 300 | 0.0349 | 99.1% | 94.7% | runner --score | qwen3.6-35b-a3b-reference-2026-10-31/gold-logits-q8-runner-vs-other-engine.json#a |
| pass | file not identified (the report calls it score file) | empero-ai/Qwen3.8-35B-A3B-Distill@bcc2dbe2 | 300 | 0.0336 | 98.6% | 91.7% | runner --score | empero-qwen3.8-35b-a3b-distill-2026-11-06/gold-q6_k.json#a |
| pass | file not identified (the report calls it score file) | Qwen/Qwen3-30B-A3B at ad44e777bcd1 (local copy) | 2000 | 0.0331 | 98.8% | 92.2% | runner --score | r4.7.3-sensitivity-sweep-2026-10-31/gold-k6.json#a |
| pass | Qwen/Qwen2.5-1.5B-Instruct-GGUF: qwen2.5-1.5b-instruct-q8_0.gguf (sha256 d7efb072e772) | Qwen/Qwen2.5-1.5B-Instruct@989aa798 | 300 | 0.0324 | 100.0% | 93.0% | an HTTP endpoint | qwen2.5-1.5b-instruct-2026-11-13/gold-q8_0-publisher-w1-runner-vs-llamacpp.json#b |
| pass | file not identified (the report calls it score file) | Qwen/Qwen3-30B-A3B at ad44e777bcd1 (local copy) | 2000 | 0.0310 | 98.8% | 93.0% | runner --score | r4.7.3-sensitivity-sweep-2026-10-31/gold-k16.json#b |
| pass | Qwen/Qwen2.5-1.5B-Instruct-GGUF: qwen2.5-1.5b-instruct-q8_0.gguf (sha256 d7efb072e772) | Qwen/Qwen2.5-1.5B-Instruct@989aa798 | 300 | 0.0307 | 100.0% | 91.3% | runner --score | qwen2.5-1.5b-instruct-2026-11-13/gold-q8_0-publisher-w1-runner-vs-llamacpp.json#a |
| pass | Qwen/Qwen2.5-1.5B-Instruct-GGUF: qwen2.5-1.5b-instruct-q8_0.gguf (sha256 d7efb072e772) | Qwen/Qwen2.5-1.5B-Instruct@989aa798 | 300 | 0.0307 | 100.0% | 91.3% | runner --score | qwen2.5-1.5b-instruct-2026-11-13/gold-q8_0-publisher-w1.json#a |
| pass | file not identified (the report calls it score file) | deepseek-ai/DeepSeek-V2-Lite-Chat@85864749cd611b4353ce1decdb286193298f64c7 | 300 | 0.0297 | 98.6% | 93.0% | runner --score | deepseek-v2-lite-chat-2026-11-04/dcause-gold-naivecpu-vs-naivecuda.json#b |
| pass | file not identified (the report calls it LFM2.5-24B-A2B-Q8_0.gguf) | LiquidAI/LFM2.5-24B-A2B at 3fcfdbfa1a1b (local copy) | 300 | 0.0297 | 99.1% | 91.3% | runner --score | lfm2.5-24b-a2b-reference-2026-10-30/gold-logits-24b-a2b-q8-q4.json#a |
| pass | mradermacher/DeepSeek-V2-Lite-Chat-GGUF: DeepSeek-V2-Lite-Chat.Q6_K.gguf | deepseek-ai/DeepSeek-V2-Lite-Chat@85864749cd611b4353ce1decdb286193298f64c7 | 300 | 0.0296 | 98.6% | 93.0% | runner --score | deepseek-v2-lite-chat-2026-11-04/gold-q6_k-cpu-vs-cuda.json#b |
| pass | mradermacher/DeepSeek-V2-Lite-Chat-GGUF: DeepSeek-V2-Lite-Chat.Q6_K.gguf | deepseek-ai/DeepSeek-V2-Lite-Chat@85864749cd611b4353ce1decdb286193298f64c7 | 300 | 0.0296 | 98.6% | 93.0% | runner --score | deepseek-v2-lite-chat-2026-11-04/gold-q6_k-cuda-vs-llamacpp.json#a |
| pass | file not identified (the report calls it score file) | deepseek-ai/DeepSeek-V2-Lite-Chat@85864749cd611b4353ce1decdb286193298f64c7 | 300 | 0.0294 | 98.6% | 93.0% | runner --score | deepseek-v2-lite-chat-2026-11-04/dcause-gold-abscpu-vs-naivecpu.json#b |
| pass | file not identified (the report calls it score file) | deepseek-ai/DeepSeek-V2-Lite-Chat@85864749cd611b4353ce1decdb286193298f64c7 | 300 | 0.0294 | 98.6% | 93.0% | runner --score | deepseek-v2-lite-chat-2026-11-04/dcause-gold-naivecpu-vs-llamacpp.json#a |
| pass | file not identified (the report calls it score file) | deepseek-ai/DeepSeek-V2-Lite-Chat@85864749cd611b4353ce1decdb286193298f64c7 | 300 | 0.0294 | 98.6% | 93.0% | runner --score | deepseek-v2-lite-chat-2026-11-04/dcause-gold-naivecpu-vs-naivecuda.json#a |
| pass | file not identified (the report calls it score file) | deepseek-ai/DeepSeek-V2-Lite-Chat@85864749cd611b4353ce1decdb286193298f64c7 | 300 | 0.0293 | 98.6% | 93.0% | runner --score | deepseek-v2-lite-chat-2026-11-04/dcause-gold-abscpu-vs-naivecpu.json#a |
| pass | mradermacher/DeepSeek-V2-Lite-Chat-GGUF: DeepSeek-V2-Lite-Chat.Q6_K.gguf | deepseek-ai/DeepSeek-V2-Lite-Chat@85864749cd611b4353ce1decdb286193298f64c7 | 300 | 0.0293 | 98.6% | 93.0% | runner --score | deepseek-v2-lite-chat-2026-11-04/gold-q6_k-cpu-vs-cuda.json#a |
| pass | file not identified (the report calls it score file) | Qwen/Qwen3-30B-A3B at ad44e777bcd1 (local copy) | 2000 | 0.0292 | 99.0% | 93.0% | runner --score | qwen3-30b-a3b-sensitivity-plan-2026-10-31/gold-logits-n11-vs-sens.json#b |
| pass | file not identified (the report calls it score file) | Qwen/Qwen3-30B-A3B@ad44e777bcd18fa416d9da3bd8f70d33ebb85d39 | 2000 | 0.0290 | 99.0% | 92.9% | runner --score | qwen3-30b-a3b-derive-2026-11-02/gold-logits-derive.json#a |
| pass | Qwen/Qwen2.5-1.5B-Instruct-GGUF: qwen2.5-1.5b-instruct-q8_0.gguf (sha256 d7efb072e772) | Qwen/Qwen2.5-1.5B-Instruct@989aa798 | 300 | 0.0287 | 99.5% | 92.0% | runner --score | qwen2.5-1.5b-instruct-2026-11-13/gold-q8_0-publisher-w2.json#a |
| pass | file not identified (the report calls it score file) | Qwen/Qwen3-30B-A3B@ad44e777bcd18fa416d9da3bd8f70d33ebb85d39 | 2000 | 0.0279 | 98.8% | 93.0% | runner --score | qwen3-30b-a3b-derive-2026-11-02/gold-logits-official-q4_k_m-selective.json#b |
| pass | ggml-org/Qwen3.6-27B-GGUF: Qwen3.6-27B-Q8_0.gguf | Qwen/Qwen3.6-27B@6a9e13bd6fc8f0983b9b99948120bc37f49c13e9 | 300 | 0.0273 | 99.6% | 98.0% | runner --score | qwen3.6-27b-2026-11-04/gold-logits-q8_0-runner-vs-llamacpp.json#a |
| pass | file not identified (the report calls it score file) | Qwen/Qwen3-30B-A3B at ad44e777bcd1 (local copy) | 2000 | 0.0267 | 99.1% | 93.0% | runner --score | r4.7.3-sensitivity-sweep-2026-10-31/gold-k6.json#b |
| pass | file not identified (the report calls it score file) | ornith-ai/Ornith-1.5-35B-A3B@10fbf86f | 300 | 0.0262 | 98.3% | 92.7% | runner --score | ornith-1.5-35b-a3b-2026-11-05/gold-bf16-vs-q8_0.json#a |
| pass | unsloth/Devstral-Small-2-24B-Instruct-2512-GGUF: Devstral-Small-2-24B-Instruct-2512-Q4_K_M.gguf (sha256 d14ba9edee1b) | mistralai/Devstral-Small-2-24B-Instruct-2512 at 55c5b41e98c2dbd21b0c8afffc540dcfc9eb5128 (local checkout) | 300 | 0.0259 | 99.5% | 92.3% | an HTTP endpoint | devstral-small-2-reference-2026-11-02/gold-logits-runner-ggml-org-q8_0-llamacpp-q4_k_m.json#b |
| pass | bartowski/Phi-3.5-mini-instruct-GGUF: Phi-3.5-mini-instruct-Q8_0.gguf (sha256 76fbf02f6fe9) | microsoft/Phi-3.5-mini-instruct@2fe19245 | 114 | 0.0252 | 98.8% | 94.7% | an HTTP endpoint | phi-3.5-mini-instruct-longrope-2026-11-13/gold-q8_0-past4096.json#b |
| pass | ggml-org/NVIDIA-Nemotron-3-Nano-30B-A3B-GGUF: NVIDIA-Nemotron-3-Nano-30B-A3B-Q4_K_M.gguf (sha256 0f111a0d4977) | nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-BF16@bf77c317 | 300 | 0.0243 | 99.5% | 94.0% | an HTTP endpoint | nemotron-3-nano-30b-a3b-2026-11-05/gold-q4_k_m-runner-vs-llamacpp.json#b |
| pass | ibm-granite/granite-4.2-30b-GGUF: granite-4.2-30b-Q8_0.gguf | ibm-granite/granite-4.2-30b@9e668ce1c538387ef24d3644e9b0606647762636 | 300 | 0.0241 | 99.1% | 95.0% | an HTTP endpoint | granite-4.2-30b-2026-11-04/gold-logits-q8_0-runner-vs-llamacpp.json#b |
| pass | file not identified (the report calls it score file) | mistralai/Devstral-Small-2-24B-Instruct-2512 at 55c5b41e98c2dbd21b0c8afffc540dcfc9eb5128 (local checkout) | 300 | 0.0223 | 99.5% | 92.3% | runner --score | devstral-small-2-reference-2026-11-02/gold-logits-runner-q4_k_m-q8_0.json#a |
| pass | ggml-org/NVIDIA-Nemotron-3-Nano-30B-A3B-GGUF: NVIDIA-Nemotron-3-Nano-30B-A3B-Q4_K_M.gguf (sha256 0f111a0d4977) | nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-BF16@bf77c317 | 300 | 0.0222 | 99.5% | 93.7% | runner --score | nemotron-3-nano-30b-a3b-2026-11-05/gold-q4_k_m-floor0.json#a |
| pass | ggml-org/NVIDIA-Nemotron-3-Nano-30B-A3B-GGUF: NVIDIA-Nemotron-3-Nano-30B-A3B-Q4_K_M.gguf (sha256 0f111a0d4977) | nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-BF16@bf77c317 | 300 | 0.0222 | 99.5% | 93.7% | runner --score | nemotron-3-nano-30b-a3b-2026-11-05/gold-q4_k_m-runner-vs-llamacpp.json#a |
| pass | bartowski/Mistral-7B-Instruct-v0.3-GGUF: Mistral-7B-Instruct-v0.3-Q4_K_M.gguf (sha256 1270d22c0fbb) | mistralai/Mistral-7B-Instruct-v0.3@c170c708 | 272 | 0.0222 | 99.5% | 93.0% | runner --score | mistral-7b-instruct-v0.3-2026-11-13/gold-q4_k_m-w1.json#a |
| pass | bartowski/Mistral-7B-Instruct-v0.3-GGUF: Mistral-7B-Instruct-v0.3-Q4_K_M.gguf (sha256 1270d22c0fbb) | mistralai/Mistral-7B-Instruct-v0.3@c170c708 | 271 | 0.0216 | 99.0% | 94.5% | runner --score | mistral-7b-instruct-v0.3-2026-11-13/gold-q4_k_m-w2.json#a |
| pass | bartowski/HemmingwayAI_Hemmingway-1.5-GGUF: HemmingwayAI_Hemmingway-1.5-Q4_K_M.gguf | HemmingwayAI/Hemmingway-1.5@2f66c2569b99 | 2000 | 0.0203 | 99.7% | 95.5% | runner --score | hemmingway-reference-2026-11-01/h15-gold-logits-cpu-q4_k_m-cuda-q4_k_m.json#b |
| pass | bartowski/HemmingwayAI_Hemmingway-1.5-GGUF: HemmingwayAI_Hemmingway-1.5-Q4_K_M.gguf | HemmingwayAI/Hemmingway-1.5@2f66c2569b99 | 2000 | 0.0203 | 99.7% | 95.5% | runner --score | hemmingway-reference-2026-11-01/h15-gold-logits-cuda-q4_k_m-cpu-q8_0.json#a |
| pass | bartowski/HemmingwayAI_Hemmingway-1.5-GGUF: HemmingwayAI_Hemmingway-1.5-Q4_K_M.gguf | HemmingwayAI/Hemmingway-1.5@2f66c2569b99 | 2000 | 0.0203 | 99.7% | 95.6% | runner --score | hemmingway-reference-2026-11-01/h15-gold-logits-cpu-q4_k_m-cuda-q4_k_m.json#a |
| pass | IFM/K2-Horizon-7B-GGUF: K2-Horizon-7B-Q8_0.gguf (sha256 5a98a289aba5) | IFM/K2-Horizon-7B@4622f94a6632 | 300 | 0.0176 | 100.0% | 93.7% | an HTTP endpoint | k2-horizon-7b-reference-2026-10-30/gold-logits-llamacpp-q8_0-text.json#a |
| pass | IFM/K2-Horizon-7B-GGUF: K2-Horizon-7B-Q8_0.gguf (sha256 5a98a289aba5) | IFM/K2-Horizon-7B@4622f94a6632 | 300 | 0.0176 | 100.0% | 93.7% | an HTTP endpoint | k2-horizon-7b-reference-2026-10-30/gold-logits-llamacpp-q8_0.json#a |
| pass | mistralai/Ministral-3-3B-Instruct-2512-GGUF: Ministral-3-3B-Instruct-2512-Q8_0.gguf | mistralai/Ministral-3-3B-Instruct-2512@b35d4dfe56c1 | 300 | 0.0174 | 100.0% | 94.3% | runner --score | ministral-3-3b-reference-2026-11-01/gold-logits-cpu-q4_k_m-cpu-q8_0.json#b |
| pass | file not identified (the report calls it NVIDIA-Nemotron-3.5-Lightning-30B-A3B-Q8_0.gguf) | nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16@a9904d24bcc1d289a1950fa9d2b978c47cf903b9 | 300 | 0.0173 | 100.0% | 93.3% | an HTTP endpoint | nemotron-3.5-lightning-30b-a3b-2026-11-03/gold-logits-q8_0-runner-vs-llamacpp.json#b |
| pass | IFM/K2-Horizon-7B-GGUF: K2-Horizon-7B-Q8_0.gguf (sha256 5a98a289aba5) | IFM/K2-Horizon-7B at 4622f94a (local checkout) | 300 | 0.0159 | 100.0% | 95.0% | runner --score | k2-horizon-7b-reference-2026-10-30/gold-logits-cpu-q8-q4.json#a |
| pass | mradermacher/DeepSeek-V2-Lite-Chat-GGUF: DeepSeek-V2-Lite-Chat.Q8_0.gguf | deepseek-ai/DeepSeek-V2-Lite-Chat@85864749cd611b4353ce1decdb286193298f64c7 | 300 | 0.0155 | 99.5% | 94.7% | an HTTP endpoint | deepseek-v2-lite-chat-2026-11-04/gold-q8_0-cuda-vs-llamacpp.json#b |
| pass | file not identified (the report calls it score file) | nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16@a9904d24bcc1d289a1950fa9d2b978c47cf903b9 | 300 | 0.0146 | 99.0% | 92.3% | runner --score | nemotron-3.5-lightning-30b-a3b-2026-11-03/gold-logits-bf16-q8_0-runner.json#b |
| pass | file not identified (the report calls it score file) | nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16@a9904d24bcc1d289a1950fa9d2b978c47cf903b9 | 300 | 0.0146 | 99.0% | 92.3% | runner --score | nemotron-3.5-lightning-30b-a3b-2026-11-03/gold-logits-q8_0-runner-vs-llamacpp.json#a |
| pass | XHToken/Spark-X2.6-4B-GGUF: Spark-X2.6-4B-Q8_0.gguf | XHToken/Spark-X2.6-4B at d1b5f260 (local checkout) | 300 | 0.0132 | 100.0% | 98.7% | runner --score | spark-2026-10-30/gold-logits-cpu-x2.6-q8-q4.json#a |
| pass | mradermacher/Hemmingway-1-GGUF: Hemmingway-1.Q8_0.gguf | HemmingwayAI/Hemmingway-1@faba57469705 | 2000 | 0.0106 | 99.5% | 98.2% | runner --score | hemmingway-reference-2026-11-01/h1-gold-logits-cuda-q4_k_m-cpu-q8_0.json#b |
| pass | mradermacher/DeepSeek-V2-Lite-Chat-GGUF: DeepSeek-V2-Lite-Chat.Q8_0.gguf | deepseek-ai/DeepSeek-V2-Lite-Chat@85864749cd611b4353ce1decdb286193298f64c7 | 300 | 0.0096 | 100.0% | 96.0% | runner --score | deepseek-v2-lite-chat-2026-11-04/gold-q8_0-cpu-vs-cuda.json#b |
| pass | mradermacher/DeepSeek-V2-Lite-Chat-GGUF: DeepSeek-V2-Lite-Chat.Q8_0.gguf | deepseek-ai/DeepSeek-V2-Lite-Chat@85864749cd611b4353ce1decdb286193298f64c7 | 300 | 0.0096 | 100.0% | 96.0% | runner --score | deepseek-v2-lite-chat-2026-11-04/gold-q8_0-cuda-vs-llamacpp.json#a |
| pass | mradermacher/DeepSeek-V2-Lite-Chat-GGUF: DeepSeek-V2-Lite-Chat.Q8_0.gguf | deepseek-ai/DeepSeek-V2-Lite-Chat@85864749cd611b4353ce1decdb286193298f64c7 | 300 | 0.0094 | 100.0% | 96.7% | runner --score | deepseek-v2-lite-chat-2026-11-04/gold-q8_0-cpu-vs-cuda.json#a |
| pass | file not identified (the report calls it score file) | google/gemma-4-E2B-it at 3e22461f65e8 (local copy) | 2000 | 0.0094 | 99.9% | 96.0% | runner --score | gemma-4-e2b-ple-quant-2026-10-31/gold-logits-ple-q40-isolated-float32.json#b |
| pass | file not identified (the report calls it score file) | google/gemma-4-E2B-it at 3e22461f65e8 (local copy) | 2000 | 0.0094 | 99.9% | 96.0% | runner --score | gemma-4-e2b-ple-quant-2026-10-31/gold-logits-ple-q40-isolated.json#b |
| pass | file not identified (the report calls it score file) | google/gemma-4-E2B-it at 3e22461f65e8 (local copy) | 2000 | 0.0089 | 99.6% | 96.0% | runner --score | gemma-4-e2b-ple-quant-2026-10-31/gold-logits-ple-q4k-isolated-float32.json#b |
| pass | file not identified (the report calls it score file) | google/gemma-4-E2B-it at 3e22461f65e8 (local copy) | 2000 | 0.0088 | 99.6% | 96.0% | runner --score | gemma-4-e2b-ple-quant-2026-10-31/gold-logits-ple-q4k-isolated.json#b |
| pass | file not identified (the report calls it score file) | openbmb/MiniCPM-V-4.7-35B-A3B@98e79ba9 (its text_config) | 300 | 0.0088 | 100.0% | 96.0% | runner --score | minicpm-v-4.7-35b-a3b-2026-11-06/gold-q8_0.json#a |
| pass | XHToken/Spark-X2.6-4B-GGUF: Spark-X2.6-4B-Q8_0.gguf | XHToken/Spark-X2.6-4B@d1b5f260 | 300 | 0.0086 | 100.0% | 98.0% | an HTTP endpoint | spark-2026-10-30/gold-logits-llamacpp-x2.6-q8_0-text.json#a |
| pass | XHToken/Spark-X2.6-4B-GGUF: Spark-X2.6-4B-Q8_0.gguf | XHToken/Spark-X2.6-4B@d1b5f260 | 300 | 0.0086 | 100.0% | 98.0% | an HTTP endpoint | spark-2026-10-30/gold-logits-llamacpp-x2.6-q8_0.json#a |
| pass | file not identified (the report calls it score file) | LiquidAI/LFM2.5-350M at 9e6c6ccf47cd (local checkout) | 300 | 0.0081 | 100.0% | 95.3% | runner --score | lfm2.5-reference-2026-10-30/gold-logits-350m-q8-q4.json#a |
| pass | LiquidAI/LFM2.5-8B-A1B-GGUF: LFM2.5-8B-A1B-BF16.gguf | LiquidAI/LFM2.5-8B-A1B at 5dd22602c2e9 (local copy) | 300 | 0.0078 | 100.0% | 97.3% | runner --score | lfm2.5-8b-a1b-reference-2026-10-30/gold-logits-8b-a1b-bf16-vs-other-engine-q8.json#a |
| pass | bartowski/SmolLM2-135M-Instruct-GGUF: SmolLM2-135M-Instruct-Q8_0.gguf (sha256 5a1395716f79) | HuggingFaceTB/SmolLM2-135M-Instruct@12fd25f7 | 296 | 0.0077 | 100.0% | 98.0% | runner --score | smollm2-2026-11-13/gold-135m-q8_0-w2.json#a |
| pass | bartowski/MiMo-V2.6-Distill-Qwen-9B-GGUF: MiMo-V2.6-Distill-Qwen-9B-Q8_0.gguf | XiaomiMiMo/MiMo-V2.6-Distill-Qwen-9B (local checkout) | 300 | 0.0075 | 100.0% | 99.0% | runner --score | mimo-v2.6-distill-qwen-9b-2026-10-29/gold-logits-cpu-q8-q4.json#a |
| pass | file not identified (the report calls it score file) | Qwen/Qwen3-0.6B at c1899de289a04d12100db370d81485cdf75e47ca (local checkout) | 2000 | 0.0075 | 100.0% | 97.0% | runner --score | r3.1.5b-cpu-int8-prefill-2026-11-02/gold-logits-q8_0-f32-vs-int8-prefill.json#b |
| pass | bartowski/Phi-3.5-mini-instruct-GGUF: Phi-3.5-mini-instruct-Q8_0.gguf (sha256 76fbf02f6fe9) | microsoft/Phi-3.5-mini-instruct@2fe19245 | 272 | 0.0072 | 100.0% | 96.3% | an HTTP endpoint | phi-3.5-mini-instruct-2026-11-13/gold-q8_0-w1-llama-c4096.json#b |
| pass | bartowski/SmolLM2-135M-Instruct-GGUF: SmolLM2-135M-Instruct-Q8_0.gguf (sha256 5a1395716f79) | HuggingFaceTB/SmolLM2-135M-Instruct@12fd25f7 | 299 | 0.0071 | 100.0% | 99.0% | runner --score | smollm2-2026-11-13/gold-135m-q8_0-w1.json#a |
| pass | XHToken/Spark-X2.5-4B-GGUF: Spark-X2.5-4B-Q8_0.gguf | XHToken/Spark-X2.5-4B@0bcb3567 | 300 | 0.0065 | 100.0% | 98.0% | an HTTP endpoint | spark-2026-10-30/gold-logits-llamacpp-x2.5-q8_0-text.json#a |
| pass | XHToken/Spark-X2.5-4B-GGUF: Spark-X2.5-4B-Q8_0.gguf | XHToken/Spark-X2.5-4B@0bcb3567 | 300 | 0.0065 | 100.0% | 98.0% | an HTTP endpoint | spark-2026-10-30/gold-logits-llamacpp-x2.5-q8_0.json#a |
| pass | IFM/K2-Horizon-0.9B-GGUF: K2-Horizon-1B-Q8_0.gguf (sha256 741fb9ad2639) | IFM/K2-Horizon-0.9B@d50881f4f8dc | 300 | 0.0064 | 100.0% | 97.0% | an HTTP endpoint | k2-horizon-0.9b-reference-2026-10-30/gold-logits-llamacpp-q8_0.json#a |
| pass | mistralai/Ministral-3-3B-Instruct-2512-GGUF: Ministral-3-3B-Instruct-2512-Q8_0.gguf | mistralai/Ministral-3-3B-Instruct-2512@b35d4dfe56c1 | 246 | 0.0061 | 100.0% | 100.0% | runner --score | ministral-3-3b-reference-2026-11-01/gold-logits-long-cpu-q8_0.json#a |
| pass | file not identified (the report calls it score file) | Qwen/Qwen3-0.6B at c1899de289a04d12100db370d81485cdf75e47ca (local checkout) | 2000 | 0.0061 | 100.0% | 97.7% | runner --score | r3.1.5b-cpu-int8-prefill-2026-11-02/gold-logits-q8_0-f32-vs-int8-prefill.json#a |
| pass | bartowski/swiss-ai_Apertus-8B-Instruct-2509-GGUF: swiss-ai_Apertus-8B-Instruct-2509-Q8_0.gguf (sha256 68a9d338961e) | swiss-ai/Apertus-8B-Instruct-2509@b946d404 | 300 | 0.0057 | 100.0% | 96.7% | runner --score | apertus-8b-instruct-2509-2026-11-12/gold-q8_0-w2.json#a |
| pass | file not identified (the report calls it score file) | nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16@a9904d24bcc1d289a1950fa9d2b978c47cf903b9 | 300 | 0.0055 | 100.0% | 96.3% | runner --score | nemotron-3.5-lightning-30b-a3b-2026-11-03/gold-logits-bf16-q8_0-runner.json#a |
| pass | file not identified (the report calls it score file) | Qwen/Qwen3-0.6B@c1899de289a04d12100db370d81485cdf75e47ca | 2000 | 0.0055 | 100.0% | 97.5% | runner --score | r3.29-cuda-decode-attention-2026-11-03/gold-main-vs-r329.json#b |
| pass | file not identified (the report calls it score file) | Qwen/Qwen3-0.6B@c1899de289a04d12100db370d81485cdf75e47ca | 2000 | 0.0055 | 100.0% | 97.5% | runner --score | r3.29-cuda-decode-attention-2026-11-03/gold-main-vs-r329.json#a |
| pass | openbmb/MiniCPM5-2B-GGUF: MiniCPM5-2B-Q8_0.gguf (sha256 c5415f8989bf) | openbmb/MiniCPM5-2B@f97400052a43 | 300 | 0.0050 | 100.0% | 98.0% | an HTTP endpoint | minicpm5-2b-reference-2026-10-29/gold-logits-llamacpp-q8_0.json#a |
| pass | bartowski/swiss-ai_Apertus-8B-Instruct-2509-GGUF: swiss-ai_Apertus-8B-Instruct-2509-Q8_0.gguf (sha256 68a9d338961e) | swiss-ai/Apertus-8B-Instruct-2509@b946d404 | 300 | 0.0050 | 100.0% | 97.3% | runner --score | apertus-8b-instruct-2509-2026-11-12/gold-q8_0-w1.json#a |
| pass | file not identified (the report calls it score file) | Qwen/Qwen3-30B-A3B at ad44e777bcd1 (local copy) | 2000 | 0.0049 | 99.9% | 97.8% | runner --score | qwen3-30b-a3b-imatrix-2026-10-31/gold-logits-q8.json#a |
| pass | file not identified (the report calls it score file) | google/gemma-4-E2B-it at 3e22461f65e8 (local copy) | 2000 | 0.0046 | 99.8% | 97.2% | runner --score | gemma-4-e2b-ple-quant-2026-10-31/gold-logits-ple-q40-isolated-float32.json#a |
| pass | file not identified (the report calls it score file) | google/gemma-4-E2B-it at 3e22461f65e8 (local copy) | 2000 | 0.0046 | 99.8% | 97.2% | runner --score | gemma-4-e2b-ple-quant-2026-10-31/gold-logits-ple-q4k-isolated-float32.json#a |
| pass | file not identified (the report calls it score file) | google/gemma-4-E2B-it at 3e22461f65e8 (local copy) | 2000 | 0.0046 | 99.8% | 97.4% | runner --score | gemma-4-e2b-ple-quant-2026-10-31/gold-logits-ple-q40-isolated.json#a |
| pass | file not identified (the report calls it score file) | google/gemma-4-E2B-it at 3e22461f65e8 (local copy) | 2000 | 0.0046 | 99.8% | 97.4% | runner --score | gemma-4-e2b-ple-quant-2026-10-31/gold-logits-ple-q4k-isolated.json#a |
| pass | bartowski/Phi-3.5-mini-instruct-GGUF: Phi-3.5-mini-instruct-Q8_0.gguf (sha256 76fbf02f6fe9) | microsoft/Phi-3.5-mini-instruct@2fe19245 | 114 | 0.0045 | 100.0% | 99.1% | runner --score | phi-3.5-mini-instruct-longrope-2026-11-13/gold-q8_0-past4096.json#a |
| pass | bartowski/SmolLM2-1.7B-Instruct-GGUF: SmolLM2-1.7B-Instruct-Q8_0.gguf (sha256 0c6e8955788b) | HuggingFaceTB/SmolLM2-1.7B-Instruct@31b70e2e | 296 | 0.0045 | 100.0% | 97.0% | runner --score | smollm2-2026-11-13/gold-1.7b-q8_0-w2.json#a |
| pass | XHToken/Spark-X2.5-4B-GGUF: Spark-X2.5-4B-Q8_0.gguf | XHToken/Spark-X2.5-4B at 0bcb3567 (local checkout) | 300 | 0.0044 | 100.0% | 98.7% | runner --score | spark-2026-10-30/gold-logits-cpu-x2.5-q8-q4.json#a |
| pass | mradermacher/stablelm-2-1_6b-chat-GGUF: stablelm-2-1_6b-chat.Q8_0.gguf (sha256 70fa97127815) | stabilityai/stablelm-2-1_6b-chat@f3fe6705 | 272 | 0.0043 | 100.0% | 97.8% | runner --score | stablelm-2-1.6b-chat-2026-11-13/gold-q8_0-w1.json#a |
| pass | bartowski/Phi-3.5-mini-instruct-GGUF: Phi-3.5-mini-instruct-Q8_0.gguf (sha256 76fbf02f6fe9) | microsoft/Phi-3.5-mini-instruct@2fe19245 | 272 | 0.0042 | 100.0% | 98.9% | runner --score | phi-3.5-mini-instruct-2026-11-13/gold-q8_0-w1-llama-c4096.json#a |
| pass | bartowski/Phi-3.5-mini-instruct-GGUF: Phi-3.5-mini-instruct-Q8_0.gguf (sha256 76fbf02f6fe9) | microsoft/Phi-3.5-mini-instruct@2fe19245 | 272 | 0.0042 | 100.0% | 98.9% | runner --score | phi-3.5-mini-instruct-2026-11-13/gold-q8_0-w1-llama-c8192.json#a |
| pass | bartowski/Phi-3.5-mini-instruct-GGUF: Phi-3.5-mini-instruct-Q8_0.gguf (sha256 76fbf02f6fe9) | microsoft/Phi-3.5-mini-instruct@2fe19245 | 272 | 0.0042 | 100.0% | 98.9% | runner --score | phi-3.5-mini-instruct-2026-11-13/gold-q8_0-w1.json#a |
| pass | file not identified (the report calls it score file) | LiquidAI/LFM2.5-1.2B-Instruct at 0f604ada3f76 (local checkout) | 300 | 0.0042 | 100.0% | 99.0% | runner --score | lfm2.5-reference-2026-10-30/gold-logits-1.2b-q8-q4.json#a |
| pass | LiquidAI/LFM2.5-1.2B-Instruct-GGUF: LFM2.5-1.2B-Instruct-Q8_0.gguf (sha256 f6b981dcb869) | LiquidAI/LFM2.5-1.2B-Instruct at 0f604ada3f76 (local copy) | 300 | 0.0042 | 100.0% | 99.0% | runner --score | lfm2.5-8b-a1b-reference-2026-10-30/gold-logits-1.2b-bf16-dense-baseline.json#b |
| pass | ibm-granite/granite-4.2-30b-GGUF: granite-4.2-30b-Q8_0.gguf | ibm-granite/granite-4.2-30b@9e668ce1c538387ef24d3644e9b0606647762636 | 300 | 0.0041 | 99.6% | 97.7% | runner --score | granite-4.2-30b-2026-11-04/gold-logits-q8_0-runner-vs-llamacpp.json#a |
| pass | mradermacher/stablelm-2-1_6b-chat-GGUF: stablelm-2-1_6b-chat.Q8_0.gguf (sha256 70fa97127815) | stabilityai/stablelm-2-1_6b-chat@f3fe6705 | 272 | 0.0039 | 100.0% | 98.5% | runner --score | stablelm-2-1.6b-chat-2026-11-13/gold-q8_0-w2.json#a |
| pass | bartowski/SmolLM2-1.7B-Instruct-GGUF: SmolLM2-1.7B-Instruct-Q8_0.gguf (sha256 0c6e8955788b) | HuggingFaceTB/SmolLM2-1.7B-Instruct@31b70e2e | 299 | 0.0038 | 100.0% | 98.3% | runner --score | smollm2-2026-11-13/gold-1.7b-q8_0-w1.json#a |
| pass | IFM/K2-Horizon-0.9B-GGUF: K2-Horizon-1B-Q8_0.gguf (sha256 741fb9ad2639) | IFM/K2-Horizon-0.9B at d50881f4f8dc6923fb8f301e1c2313fe3a2b6c1d (local checkout) | 300 | 0.0037 | 100.0% | 97.3% | runner --score | k2-horizon-0.9b-reference-2026-10-30/gold-logits-metal-q8-q4.json#a |
| pass | andreasmartin/apertus-v1.5-8b-text-Q8_0-GGUF: apertus-v1.5-8b-text-q8_0.gguf (sha256 dea904ad80cd) | swiss-ai/Apertus-v1.5-8B@1519de0a | 300 | 0.0037 | 100.0% | 98.7% | runner --score | apertus-v1.5-8b-text-2026-11-13/gold-q8_0-w2.json#a |
| pass | file not identified (the report calls it score file) | IFM/K2-Horizon-0.9B at d50881f4f8dc6923fb8f301e1c2313fe3a2b6c1d (local checkout) | 2000 | 0.0036 | 100.0% | 97.7% | runner --score | r3.24b-metal-microbatch-tiles-2026-11-02/gold-logits-q8_0-solo-vs-tiles.json#b |
| pass | file not identified (the report calls it score file) | IFM/K2-Horizon-0.9B at d50881f4f8dc6923fb8f301e1c2313fe3a2b6c1d (local checkout) | 2000 | 0.0036 | 100.0% | 97.7% | runner --score | r3.24b-metal-microbatch-tiles-2026-11-02/gold-logits-q8_0-solo-vs-tiles.json#a |
| pass | andreasmartin/apertus-v1.5-8b-text-Q8_0-GGUF: apertus-v1.5-8b-text-q8_0.gguf (sha256 dea904ad80cd) | swiss-ai/Apertus-v1.5-8B@1519de0a | 300 | 0.0035 | 100.0% | 97.0% | runner --score | apertus-v1.5-8b-text-2026-11-13/gold-q8_0-w1.json#a |
| pass | file not identified (the report calls it score file) | Qwen/Qwen2.5-1.5B-Instruct@989aa798 | 300 | 0.0034 | 100.0% | 98.7% | runner --score | qwen2.5-1.5b-instruct-2026-11-13/gold-lab-q8_0-w2.json#a |
| pass | bartowski/Qwen2.5-1.5B-Instruct-GGUF: Qwen2.5-1.5B-Instruct-Q8_0.gguf (sha256 7185d306cf45) | Qwen/Qwen2.5-1.5B-Instruct@989aa798 | 300 | 0.0034 | 100.0% | 98.7% | runner --score | qwen2.5-1.5b-instruct-2026-11-13/gold-q8_0-bartowski-w2.json#a |
| pass | openbmb/MiniCPM5-2B-GGUF: MiniCPM5-2B-Q8_0.gguf (sha256 c5415f8989bf) | openbmb/MiniCPM5-2B at f97400052a43d642bbc6e9975e2397e3ae6a6b52 (local checkout) | 300 | 0.0033 | 100.0% | 98.0% | runner --score | minicpm5-2b-reference-2026-10-29/gold-logits-metal-q8-q4.json#a |
| pass | bartowski/Phi-3.5-mini-instruct-GGUF: Phi-3.5-mini-instruct-Q8_0.gguf (sha256 76fbf02f6fe9) | microsoft/Phi-3.5-mini-instruct@2fe19245 | 272 | 0.0033 | 100.0% | 98.2% | runner --score | phi-3.5-mini-instruct-2026-11-13/gold-q8_0-w2.json#a |
| pass | file not identified (the report calls it score file) | Qwen/Qwen3.5-0.8B at 2fc06364715b (local copy) | 2000 | 0.0032 | 100.0% | 98.3% | runner --score | qwen35-metal-2026-10-31/gold-logits-cpu-vs-metal.json#b |
| pass | file not identified (the report calls it score file) | Qwen/Qwen3.5-0.8B at 2fc06364715b (local copy) | 2000 | 0.0032 | 100.0% | 98.3% | runner --score | r3.24-metal-small-prefill-2026-10-31/gold-w16-qwen.json#a |
| pass | file not identified (the report calls it score file) | Qwen/Qwen3.5-0.8B at 2fc06364715b (local copy) | 2000 | 0.0032 | 100.0% | 98.3% | runner --score | r3.24-metal-small-prefill-2026-10-31/gold-w8-qwen.json#a |
| pass | file not identified (the report calls it score file) | Qwen/Qwen3.5-0.8B at 2fc06364715b (local copy) | 2000 | 0.0032 | 100.0% | 98.5% | runner --score | r3.24-metal-small-prefill-2026-10-31/gold-w16-qwen.json#b |
| pass | file not identified (the report calls it score file) | Qwen/Qwen3.5-0.8B at 2fc06364715b (local copy) | 2000 | 0.0032 | 100.0% | 98.5% | runner --score | r3.24-metal-small-prefill-2026-10-31/gold-w8-qwen.json#b |
| pass | file not identified (the report calls it score file) | Qwen/Qwen3.5-0.8B at 2fc06364715b (local copy) | 2000 | 0.0032 | 100.0% | 98.3% | runner --score | qwen35-metal-2026-10-31/gold-logits-cpu-vs-metal.json#a |
| pass | file not identified (the report calls it score file) | Qwen/Qwen2.5-1.5B-Instruct@989aa798 | 300 | 0.0032 | 100.0% | 98.0% | runner --score | qwen2.5-1.5b-instruct-2026-11-13/gold-lab-q8_0-w1.json#a |
| pass | bartowski/Qwen2.5-1.5B-Instruct-GGUF: Qwen2.5-1.5B-Instruct-Q8_0.gguf (sha256 7185d306cf45) | Qwen/Qwen2.5-1.5B-Instruct@989aa798 | 300 | 0.0032 | 100.0% | 98.0% | runner --score | qwen2.5-1.5b-instruct-2026-11-13/gold-q8_0-bartowski-w1.json#a |
| pass | ibm-granite/granite-4.0-h-micro-GGUF: granite-4.0-h-micro-Q8_0.gguf (sha256 a009111abf28) | ibm-granite/granite-4.0-h-micro@d5f01a3e | 272 | 0.0026 | 100.0% | 98.9% | runner --score | granite-4.0-h-micro-2026-11-13/gold-q8_0-w2.json#a |
| pass | ibm-granite/granite-4.0-h-micro-GGUF: granite-4.0-h-micro-Q8_0.gguf (sha256 a009111abf28) | ibm-granite/granite-4.0-h-micro@d5f01a3e | 272 | 0.0024 | 100.0% | 99.3% | runner --score | granite-4.0-h-micro-2026-11-13/gold-q8_0-w1.json#a |
| pass | bartowski/HemmingwayAI_Hemmingway-1.5-GGUF: HemmingwayAI_Hemmingway-1.5-Q8_0.gguf | HemmingwayAI/Hemmingway-1.5@2f66c2569b99 | 2000 | 0.0021 | 99.9% | 99.1% | runner --score | hemmingway-reference-2026-11-01/h15-gold-logits-cuda-q4_k_m-cpu-q8_0.json#b |
| pass | unsloth/Devstral-Small-2-24B-Instruct-2512-GGUF: Devstral-Small-2-24B-Instruct-2512-Q8_0.gguf (sha256 0760502e9228) | mistralai/Devstral-Small-2-24B-Instruct-2512 at 55c5b41e98c2dbd21b0c8afffc540dcfc9eb5128 (local checkout) | 300 | 0.0021 | 100.0% | 99.3% | an HTTP endpoint | devstral-small-2-reference-2026-11-02/gold-logits-llamacpp-q8_0.json#a |
| pass | bartowski/Mistral-7B-Instruct-v0.3-GGUF: Mistral-7B-Instruct-v0.3-Q8_0.gguf (sha256 404857e77611) | mistralai/Mistral-7B-Instruct-v0.3@c170c708 | 272 | 0.0015 | 100.0% | 98.2% | runner --score | mistral-7b-instruct-v0.3-2026-11-13/gold-q8_0-w1.json#a |
| pass | file not identified (the report calls it score file) | mistralai/Devstral-Small-2-24B-Instruct-2512 at 55c5b41e98c2dbd21b0c8afffc540dcfc9eb5128 (local checkout) | 300 | 0.0013 | 100.0% | 99.3% | runner --score | devstral-small-2-reference-2026-11-02/gold-logits-runner-ggml-org-q8_0-llamacpp-q4_k_m.json#a |
| pass | file not identified (the report calls it score file) | mistralai/Devstral-Small-2-24B-Instruct-2512 at 55c5b41e98c2dbd21b0c8afffc540dcfc9eb5128 (local checkout) | 300 | 0.0013 | 100.0% | 99.3% | runner --score | devstral-small-2-reference-2026-11-02/gold-logits-runner-q4_k_m-q8_0.json#b |
| pass | bartowski/Mistral-7B-Instruct-v0.3-GGUF: Mistral-7B-Instruct-v0.3-Q8_0.gguf (sha256 404857e77611) | mistralai/Mistral-7B-Instruct-v0.3@c170c708 | 271 | 0.0008 | 100.0% | 98.9% | runner --score | mistral-7b-instruct-v0.3-2026-11-13/gold-q8_0-w2.json#a |
| pass | file not identified (the report calls it score file) | Qwen/Qwen2.5-1.5B-Instruct@989aa798 | 300 | 0.0002 | 100.0% | 99.7% | runner --score | qwen2.5-1.5b-instruct-2026-11-13/gold-lab-bf16-w1.json#a |
| pass | file not identified (the report calls it score file) | Qwen/Qwen2.5-1.5B-Instruct@989aa798 | 300 | 0.0001 | 100.0% | 99.7% | runner --score | qwen2.5-1.5b-instruct-2026-11-13/gold-lab-bf16-w2.json#a |
| pass | file not identified (the report calls it LFM2.5-1.2B-Instruct-BF16.gguf) | LiquidAI/LFM2.5-1.2B-Instruct at 0f604ada3f76 (local copy) | 300 | 0.0001 | 100.0% | 100.0% | runner --score | lfm2.5-8b-a1b-reference-2026-10-30/gold-logits-1.2b-bf16-dense-baseline.json#a |
| pass | file not identified (the report calls it score file) | Qwen/Qwen3.8-27B@1d4bf0f2ff6012fd82039f2fa52739d0dd7c60c0 | 2000 | 0.0000 | 100.0% | 100.0% | runner --score | r4.32-qwen38-iq3s-reference-2026-11-04/bf16diag-gold-d1window.json#a |
| pass | file not identified (the report calls it score file) | Qwen/Qwen3.8-27B@1d4bf0f2ff6012fd82039f2fa52739d0dd7c60c0 | 300 | 0.0000 | 100.0% | 99.7% | runner --score | r4.32-qwen38-iq3s-reference-2026-11-04/bf16diag-gold-house.json#a |