Calibration
AA-LCR → LongBench v2
LongBench v2 is estimated from AA-LCR with a offset logistic curve fitted on 10 models measured on both: y = 0.0000 + (0.6146 − 0.0000) / (1 + exp(−200.00·(x − 0.5855))), R² = 0.93, cross-validated error 5.9 pp. It is used for 176 estimates.
| Estimated model | AA-LCR | LongBench v2 | Source |
|---|---|---|---|
| A.X K2 | 66.0% | 61.5% | estimated ± 5.9 pp, medium confidence |
| Apodex 1.1 | 79.3% | 61.5% | estimated ± 5.9 pp, low confidence |
| Apodex 1.1 Mini | 79.3% | 61.5% | estimated ± 5.9 pp, low confidence |
| Celeris-1 | 38.0% | 0.0% | estimated ± 5.9 pp, low confidence |
| Claude 3 Haiku | 27.7% | 0.0% | estimated ± 5.9 pp, low confidence |
| Claude 4.1 Opus Thinking | 76.0% | 61.5% | estimated ± 5.9 pp, medium confidence |
| Claude 4 Sonnet | 44.0% | 0.0% | estimated ± 5.9 pp, low confidence |
| Claude Fable 5 | 82.3% | 61.5% | estimated ± 5.9 pp, low confidence |
| Claude Fable 5.1 | 85.3% | 61.5% | estimated ± 5.9 pp, low confidence |
| Claude Haiku 5.5 | 82.7% | 61.5% | estimated ± 5.9 pp, low confidence |
| Claude Opus 4.5 Thinking | 77.3% | 61.5% | estimated ± 5.9 pp, medium confidence |
| Claude Opus 4.6 | 67.0% | 61.5% | estimated ± 5.9 pp, medium confidence |
| Claude Opus 4.6 (Adaptive) | 78.0% | 61.5% | estimated ± 5.9 pp, medium confidence |
| Claude Opus 4.7 | 75.7% | 61.5% | estimated ± 5.9 pp, medium confidence |
| Claude Opus 4.7 (Adaptive) | 78.7% | 61.5% | estimated ± 5.9 pp, low confidence |
| Claude Opus 4.8 | 77.7% | 61.5% | estimated ± 5.9 pp, medium confidence |
| Claude Opus 5 | 79.3% | 61.5% | estimated ± 5.9 pp, low confidence |
| Claude Opus 5.5 | 84.7% | 61.5% | estimated ± 5.9 pp, low confidence |
| Claude Sonnet 4.6 | 68.3% | 61.5% | estimated ± 5.9 pp, medium confidence |
| Claude Sonnet 5 | 82.0% | 61.5% | estimated ± 5.9 pp, low confidence |
| Claude Sonnet 5.5 | 82.7% | 61.5% | estimated ± 5.9 pp, low confidence |
| Command A+ | 52.7% | 0.0% | estimated ± 5.9 pp, low confidence |
| DeepSeek-R1 | 55.7% | 0.2% | estimated ± 5.9 pp, low confidence |
| DeepSeek R1 Distill Qwen 32B | 8.7% | 0.0% | estimated ± 5.9 pp, low confidence |
| DeepSeek V3 | 29.3% | 0.0% | estimated ± 5.9 pp, low confidence |
| DeepSeek V3 0324 | 40.7% | 0.0% | estimated ± 5.9 pp, low confidence |
| DeepSeek V3.1 | 47.0% | 0.0% | estimated ± 5.9 pp, low confidence |
| DeepSeek V3.1 (Reasoning) | 56.7% | 1.5% | estimated ± 5.9 pp, low confidence |
| DeepSeek V3.2 | 45.7% | 0.0% | estimated ± 5.9 pp, low confidence |
| DeepSeek V4.1 Flash | 84.0% | 61.5% | estimated ± 5.9 pp, low confidence |
| DeepSeek V4 Flash 0731 | 79.7% | 61.5% | estimated ± 5.9 pp, low confidence |
| DeepSeek V4 Pro 0813 | 80.3% | 61.5% | estimated ± 5.9 pp, low confidence |
| Exaone 4.0 1.2B | 0.0% | 0.0% | estimated ± 5.9 pp, low confidence |
| Gemini 2.5 Flash | 49.9% | 0.0% | estimated ± 5.9 pp, low confidence |
| Gemini 2.5 Pro | 69.0% | 61.5% | estimated ± 5.9 pp, medium confidence |
| Gemini 3.1 Pro | 82.0% | 61.5% | estimated ± 5.9 pp, low confidence |
| Gemini 3.5 Flash | 69.3% | 61.5% | estimated ± 5.9 pp, medium confidence |
| Gemini 3.5 Flash-Lite | 76.0% | 61.5% | estimated ± 5.9 pp, medium confidence |
| Gemini 3.6 Flash | 80.0% | 61.5% | estimated ± 5.9 pp, low confidence |
| Gemini 3.7 Flash | 81.7% | 61.5% | estimated ± 5.9 pp, low confidence |
| Gemini 3.8 Flash | 81.3% | 61.5% | estimated ± 5.9 pp, low confidence |
| Gemini 3 Flash | 55.3% | 0.1% | estimated ± 5.9 pp, low confidence |
| Gemini 3 Pro | 76.0% | 61.5% | estimated ± 5.9 pp, medium confidence |
| Gemini 4 Argon | 79.7% | 61.5% | estimated ± 5.9 pp, low confidence |
| Gemma 3 27B | 7.3% | 0.0% | estimated ± 5.9 pp, low confidence |
| Gemma 4 12B | 63.7% | 61.5% | estimated ± 5.9 pp, medium confidence |
| Gemma 4 26B A4B | 65.7% | 61.5% | estimated ± 5.9 pp, medium confidence |
| Gemma 4 31B | 69.7% | 61.5% | estimated ± 5.9 pp, medium confidence |
| Gemma 4 E2B | 16.3% | 0.0% | estimated ± 5.9 pp, low confidence |
| Gemma 4 E4B | 32.0% | 0.0% | estimated ± 5.9 pp, low confidence |
| GLM-4.5-Air | 46.7% | 0.0% | estimated ± 5.9 pp, low confidence |
| GLM-4.6 | 26.3% | 0.0% | estimated ± 5.9 pp, low confidence |
| GLM-4.7 | 71.0% | 61.5% | estimated ± 5.9 pp, medium confidence |
| GLM-5.1 | 73.7% | 61.5% | estimated ± 5.9 pp, medium confidence |
| GLM-5.2 | 78.3% | 61.5% | estimated ± 5.9 pp, medium confidence |
| GLM-5.3 | 79.7% | 61.5% | estimated ± 5.9 pp, low confidence |
| GLM-5.3-Flash | 80.0% | 61.5% | estimated ± 5.9 pp, low confidence |
| GLM-5-Turbo | 71.7% | 61.5% | estimated ± 5.9 pp, medium confidence |
| GLM-5V-Turbo | 70.3% | 61.5% | estimated ± 5.9 pp, medium confidence |
| GPT-4.1 | 68.3% | 61.5% | estimated ± 5.9 pp, medium confidence |
| GPT-4.1 mini | 44.0% | 0.0% | estimated ± 5.9 pp, low confidence |
| GPT-4.1 nano | 20.3% | 0.0% | estimated ± 5.9 pp, low confidence |
| GPT-4o | 49.3% | 0.0% | estimated ± 5.9 pp, low confidence |
| GPT-5.1 | 80.0% | 61.5% | estimated ± 5.9 pp, low confidence |
| GPT-5.1-Codex | 69.3% | 61.5% | estimated ± 5.9 pp, medium confidence |
| GPT-5.1-Codex-Max | 69.3% | 61.5% | estimated ± 5.9 pp, medium confidence |
| GPT-5.2 | 82.7% | 61.5% | estimated ± 5.9 pp, low confidence |
| GPT-5.2-Codex | 82.3% | 61.5% | estimated ± 5.9 pp, low confidence |
| GPT-5.3 Codex | 83.3% | 61.5% | estimated ± 5.9 pp, low confidence |
| GPT-5.4 | 82.0% | 61.5% | estimated ± 5.9 pp, low confidence |
| GPT-5.4 mini | 77.0% | 61.5% | estimated ± 5.9 pp, medium confidence |
| GPT-5.4 nano | 76.7% | 61.5% | estimated ± 5.9 pp, medium confidence |
| GPT-5.5 | 84.3% | 61.5% | estimated ± 5.9 pp, low confidence |
| GPT-5.6 Luna | 83.7% | 61.5% | estimated ± 5.9 pp, low confidence |
| GPT-5.6 Sol | 84.0% | 61.5% | estimated ± 5.9 pp, low confidence |
| GPT-5.6 Terra | 83.0% | 61.5% | estimated ± 5.9 pp, low confidence |
| GPT-5 (high) | 78.2% | 61.5% | estimated ± 5.9 pp, medium confidence |
| GPT-5 (medium) | 76.0% | 61.5% | estimated ± 5.9 pp, medium confidence |
| GPT-6.1 Sol | 83.0% | 61.5% | estimated ± 5.9 pp, low confidence |
| GPT-6 Astra | 80.7% | 61.5% | estimated ± 5.9 pp, low confidence |
| GPT-6 Luna | 83.3% | 61.5% | estimated ± 5.9 pp, low confidence |
| GPT-6 Sol | 83.7% | 61.5% | estimated ± 5.9 pp, low confidence |
| GPT-OSS 120B | 52.0% | 0.0% | estimated ± 5.9 pp, low confidence |
| GPT-OSS 20B | 34.7% | 0.0% | estimated ± 5.9 pp, low confidence |
| Granite-4.0-350M | 0.0% | 0.0% | estimated ± 5.9 pp, low confidence |
| Granite-4.0-H-1B | 7.0% | 0.0% | estimated ± 5.9 pp, low confidence |
| Granite-4.0-H-350M | 0.0% | 0.0% | estimated ± 5.9 pp, low confidence |
| Granite 4.2 30B | 49.0% | 0.0% | estimated ± 5.9 pp, low confidence |
| Granite 4.2 3B | 24.3% | 0.0% | estimated ± 5.9 pp, low confidence |
| Granite 4.2 8B | 45.0% | 0.0% | estimated ± 5.9 pp, low confidence |
| Grok 4 | 68.0% | 61.5% | estimated ± 5.9 pp, medium confidence |
| Grok 4.1 Fast | 31.3% | 0.0% | estimated ± 5.9 pp, low confidence |
| Grok 4.1 Fast (Reasoning) | 74.0% | 61.5% | estimated ± 5.9 pp, medium confidence |
| Grok 4.3 | 64.3% | 61.5% | estimated ± 5.9 pp, medium confidence |
| Grok 4.5 | 79.3% | 61.5% | estimated ± 5.9 pp, low confidence |
| Grok 4.6 | 80.3% | 61.5% | estimated ± 5.9 pp, low confidence |
| Grok 4.7 | 76.7% | 61.5% | estimated ± 5.9 pp, medium confidence |
| Grok 4 Fast (Reasoning) | 73.7% | 61.5% | estimated ± 5.9 pp, medium confidence |
| Grok Code Fast 1 | 53.0% | 0.0% | estimated ± 5.9 pp, low confidence |
| Hy3 | 79.0% | 61.5% | estimated ± 5.9 pp, low confidence |
| Hy3 Preview | 66.7% | 61.5% | estimated ± 5.9 pp, medium confidence |
| Inkling | 77.3% | 61.5% | estimated ± 5.9 pp, medium confidence |
| Inkling-Small | 75.7% | 61.5% | estimated ± 5.9 pp, medium confidence |
| K-Exaone | 61.3% | 61.2% | estimated ± 5.9 pp, medium confidence |
| K-EXAONE 2.0 | 56.2% | 0.6% | estimated ± 5.9 pp, low confidence |
| Kimi K2.6 | 81.0% | 61.5% | estimated ± 5.9 pp, low confidence |
| Kimi K2 | 53.0% | 0.0% | estimated ± 5.9 pp, low confidence |
| Kimi K2.5 (Reasoning) | 78.0% | 61.5% | estimated ± 5.9 pp, medium confidence |
| Kimi K2.7 Code | 79.3% | 61.5% | estimated ± 5.9 pp, low confidence |
| Kimi K3 | 88.7% | 61.5% | estimated ± 5.9 pp, low confidence |
| LFM2.5-2.6B | 5.7% | 0.0% | estimated ± 5.9 pp, low confidence |
| LFM2.5-8B-A1B | 0.0% | 0.0% | estimated ± 5.9 pp, low confidence |
| LFM2.5-VL-1.6B-Extract | 0.0% | 0.0% | estimated ± 5.9 pp, low confidence |
| Ling 2.6 Flash | 31.3% | 0.0% | estimated ± 5.9 pp, low confidence |
| Ling 3.0 Flash | 73.0% | 61.5% | estimated ± 5.9 pp, medium confidence |
| Ling 3.0 Flash FP8 | 73.0% | 61.5% | estimated ± 5.9 pp, medium confidence |
| Ling 3.0 Flash VL | 78.3% | 61.5% | estimated ± 5.9 pp, medium confidence |
| Ling 3.0 Tiny | 60.3% | 59.7% | estimated ± 5.9 pp, medium confidence |
| Ling 3.1 Flash | 83.0% | 61.5% | estimated ± 5.9 pp, low confidence |
| Llama 3.1 405B | 25.3% | 0.0% | estimated ± 5.9 pp, low confidence |
| Llama 4 Maverick | 50.0% | 0.0% | estimated ± 5.9 pp, low confidence |
| Llama 4 Scout | 27.7% | 0.0% | estimated ± 5.9 pp, low confidence |
| Mercury 2.5 | 68.0% | 61.5% | estimated ± 5.9 pp, medium confidence |
| MiMo-V2.5-Pro | 79.7% | 61.5% | estimated ± 5.9 pp, low confidence |
| MiMo-V2.6-Flash | 74.3% | 61.5% | estimated ± 5.9 pp, medium confidence |
| MiMo-V2.6-Pro | 86.3% | 61.5% | estimated ± 5.9 pp, low confidence |
| MiMo-V2-Flash | 36.3% | 0.0% | estimated ± 5.9 pp, low confidence |
| MiMo-V2-Omni | 75.0% | 61.5% | estimated ± 5.9 pp, medium confidence |
| MiMo-V2-Pro | 68.3% | 61.5% | estimated ± 5.9 pp, medium confidence |
| MiniMax M2.7 | 78.3% | 61.5% | estimated ± 5.9 pp, medium confidence |
| MiniMax M3 | 83.0% | 61.5% | estimated ± 5.9 pp, low confidence |
| Mistral Large 3 | 36.0% | 0.0% | estimated ± 5.9 pp, low confidence |
| Mistral Large 4 | 81.3% | 61.5% | estimated ± 5.9 pp, low confidence |
| Mistral Medium 3 | 31.3% | 0.0% | estimated ± 5.9 pp, low confidence |
| Mistral Medium 3.5 128B | 69.3% | 61.5% | estimated ± 5.9 pp, medium confidence |
| Mistral Small 4 | 49.7% | 0.0% | estimated ± 5.9 pp, low confidence |
| Mistral Small 4 (Reasoning) | 49.7% | 0.0% | estimated ± 5.9 pp, low confidence |
| Muse Glimmer 30B | 83.3% | 61.5% | estimated ± 5.9 pp, low confidence |
| Muse Spark | 78.0% | 61.5% | estimated ± 5.9 pp, medium confidence |
| Muse Spark 1.1 | 77.7% | 61.5% | estimated ± 5.9 pp, medium confidence |
| Muse Spark 1.2 | 79.0% | 61.5% | estimated ± 5.9 pp, low confidence |
| Muse Spark 1.3 | 83.0% | 61.5% | estimated ± 5.9 pp, low confidence |
| Nemotron 3.5 Lightning 30B A3B NVFP4 | 49.2% | 0.0% | estimated ± 5.9 pp, low confidence |
| Nemotron 3 Nano 30B | 38.0% | 0.0% | estimated ± 5.9 pp, low confidence |
| Nemotron 3 Nano Omni 30B A3B | 39.7% | 0.0% | estimated ± 5.9 pp, low confidence |
| Nemotron 3 Super 100B | 65.7% | 61.5% | estimated ± 5.9 pp, medium confidence |
| North Mini Code | 37.3% | 0.0% | estimated ± 5.9 pp, low confidence |
| Nova Pro | 21.0% | 0.0% | estimated ± 5.9 pp, low confidence |
| o1 | 65.0% | 61.5% | estimated ± 5.9 pp, medium confidence |
| o3 | 74.7% | 61.5% | estimated ± 5.9 pp, medium confidence |
| Phi-4 | 0.0% | 0.0% | estimated ± 5.9 pp, low confidence |
| Quasar 438B | 76.3% | 61.5% | estimated ± 5.9 pp, medium confidence |
| Qwen3.5 397B (Reasoning) | 64.3% | 61.5% | estimated ± 5.9 pp, medium confidence |
| Qwen3.6-27B | 77.3% | 61.5% | estimated ± 5.9 pp, medium confidence |
| Qwen3.6-35B-A3B | 71.7% | 61.5% | estimated ± 5.9 pp, medium confidence |
| Qwen 3.6 Max (preview) | 80.7% | 61.5% | estimated ± 5.9 pp, low confidence |
| Qwen3.7 Max | 79.0% | 61.5% | estimated ± 5.9 pp, low confidence |
| Qwen3.7 Plus | 73.0% | 61.5% | estimated ± 5.9 pp, medium confidence |
| Qwen3.8-27B | 82.0% | 61.5% | estimated ± 5.9 pp, low confidence |
| Qwen3.8-Flash-Next | 79.7% | 61.5% | estimated ± 5.9 pp, low confidence |
| Qwen3.8 Max Preview | 80.3% | 61.5% | estimated ± 5.9 pp, low confidence |
| Qwen3 Max | 50.0% | 0.0% | estimated ± 5.9 pp, low confidence |
| Qwen3-Omni-30B-A3B-Instruct | 0.0% | 0.0% | estimated ± 5.9 pp, low confidence |
| Qwen3-Omni-30B-A3B-Thinking | 0.0% | 0.0% | estimated ± 5.9 pp, low confidence |
| Sarvam 105B | 0.0% | 0.0% | estimated ± 5.9 pp, low confidence |
| Sarvam 30B | 0.0% | 0.0% | estimated ± 5.9 pp, low confidence |
| Solar Open 2 | 62.3% | 61.4% | estimated ± 5.9 pp, medium confidence |
| Solar Pro 2 | 0.0% | 0.0% | estimated ± 5.9 pp, low confidence |
| Solar Pro 3 | 32.3% | 0.0% | estimated ± 5.9 pp, low confidence |
| Solar Pro 4 | 71.0% | 61.5% | estimated ± 5.9 pp, medium confidence |
| Step 3.7 Flash | 73.7% | 61.5% | estimated ± 5.9 pp, medium confidence |
| Step 5 Preview | 88.3% | 61.5% | estimated ± 5.9 pp, low confidence |
| Ternary Bonsai 2 27B | 77.0% | 61.5% | estimated ± 5.9 pp, medium confidence |
| Trinity-Large-Preview | 38.0% | 0.0% | estimated ± 5.9 pp, low confidence |
| Trinity-Large-Thinking | 38.0% | 0.0% | estimated ± 5.9 pp, low confidence |
| Ultravox v0.6 Llama 3.3 70B | 15.7% | 0.0% | estimated ± 5.9 pp, low confidence |