这个评测关注什么
面向多步骤、长周期知识工作的专有评测。
阅读时注意
评测结果受任务集合、评分方法和模型版本影响。本站保留原始口径，不用它生成新的综合排名。
适用场景
用于回答一个明确的问题：模型在这一类任务上的相对表现如何，而不是“哪个模型绝对最好”。
长期知识工作
AA-Briefcase
AA-Briefcase 公开评测榜单：面向多步骤、长周期知识工作的专有评测。本站保留来源口径，不合成新的综合评分。
主要指标Elo
快照日期2026-09-18T01:47:51.384Z
| 排名 | 机构 | 模型 | Elo | 95% CI | 发布日期 |
|---|---|---|---|---|---|
| 1 | Anthropic | Claude Fable 5.1 (Adaptive Reasoning, Xhigh Effort, Default Fallback) | 1,650 裁判协议 v4.3.1 | -9 / +9 | Sep 2026 |
| 2 | Anthropic | Claude Fable 5.1 (Adaptive Reasoning, Max Effort, Default Fallback) | 1,646 裁判协议 v4.3.1 | -8 / +8 | Sep 2026 |
| 3 | Anthropic | Claude Opus 5 (Adaptive Reasoning, Max Effort) | 1,633 裁判协议 v4.3.1 | -6 / +7 | Jul 2026 |
| 4 | Anthropic | Claude Opus 5 (Adaptive Reasoning, Xhigh Effort) | 1,625 裁判协议 v4.3.1 | -7 / +8 | Jul 2026 |
| 5 | Alibaba | Qwen3.8 Max (0902) | 1,610 裁判协议 v4.3.1 | -10 / +10 | Sep 2026 |
| 6 | Meta | Muse Spark 1.3 (max) | 1,589 裁判协议 v4.3.1 | -11 / +11 | Sep 2026 |
| 7 | Alibaba | Qwen3.8-Flash-Next | 1,587 裁判协议 v4.3.1 | -10 / +11 | Aug 2026 |
| 8 | Anthropic | Claude Fable 5.1 (Adaptive Reasoning, High Effort, Default Fallback) | 1,578 裁判协议 v4.3.1 | -8 / +9 | Sep 2026 |
| 9 | OpenAI | GPT-6 Astra (max) | 1,562 裁判协议 v4.3.1 | -9 / +10 | Sep 2026 |
| 10 | Anthropic | Claude Opus 5 (Adaptive Reasoning, High Effort) | 1,557 裁判协议 v4.3.1 | -7 / +7 | Jul 2026 |
| 11 | SpaceXAI | Grok 4.6 (xhigh) | 1,545 裁判协议 v4.3.1 | -6 / +7 | Aug 2026 |
| 12 | SpaceXAI | Grok 4.6 (high) | 1,534 裁判协议 v4.3.1 | -6 / +7 | Aug 2026 |
| 13 | OpenAI | GPT-6 Astra (xhigh) | 1,534 裁判协议 v4.3.1 | -9 / +10 | Sep 2026 |
| 14 | Anthropic | Claude Fable 5.1 (Adaptive Reasoning, Medium Effort, Default Fallback) | 1,531 裁判协议 v4.3.1 | -8 / +9 | Sep 2026 |
| 15 | Anthropic | Claude Fable 5 | 1,530 裁判协议 v4.3.1 | -6 / +6 | Jun 2026 |
| 16 | Z AI | GLM-5.3 (max) | 1,504 裁判协议 v4.3.1 | -7 / +8 | Aug 2026 |
| 17 | OpenAI | GPT-6 Astra (high) | 1,494 裁判协议 v4.3.1 | -9 / +10 | Sep 2026 |
| 18 | SpaceXAI | Grok 4.6 (medium) | 1,491 裁判协议 v4.3.1 | -6 / +7 | Aug 2026 |
| 19 | Kimi | Kimi K3 (max) | 1,488 裁判协议 v4.3.1 | -5 / +6 | Jul 2026 |
| 20 | Anthropic | Claude Fable 5.1 (Adaptive Reasoning, Low Effort, Default Fallback) | 1,484 裁判协议 v4.3.1 | -8 / +8 | Sep 2026 |
| 21 | Meta | Muse Spark 1.3 (xhigh) | 1,483 裁判协议 v4.3.1 | -10 / +11 | Sep 2026 |
| 22 | OpenAI | GPT-5.6 Sol (max) | 1,463 裁判协议 v4.3.1 | -6 / +7 | Jul 2026 |
| 23 | OpenAI | GPT-6 Astra (medium) | 1,451 裁判协议 v4.3.1 | -9 / +9 | Sep 2026 |
| 24 | Z AI | GLM 5.3 Flash | 1,449 裁判协议 v4.3.1 | -9 / +9 | Aug 2026 |
| 25 | Anthropic | Claude Opus 5 (Adaptive Reasoning, Medium Effort) | 1,439 裁判协议 v4.3.1 | -6 / +7 | Jul 2026 |
| 26 | Alibaba | Qwen3.8 2.4T A95B | 1,439 裁判协议 v4.3.1 | -7 / +8 | Aug 2026 |
| 27 | OpenAI | GPT-5.6 Sol (xhigh) | 1,435 裁判协议 v4.3.1 | -8 / +9 | Jul 2026 |
| 28 | DeepSeek | DeepSeek V4 Flash Vision (Reasoning, Max Effort) | 1,432 裁判协议 v4.3.1 | -9 / +10 | Aug 2026 |
| 29 | DeepSeek | DeepSeek V4.1 Flash (Reasoning, Max Effort) | 1,424 裁判协议 v4.3.1 | -10 / +10 | Sep 2026 |
| 30 | Alibaba | Qwen3.8 27B (xhigh) | 1,398 裁判协议 v4.3.1 | -7 / +7 | Aug 2026 |
| 31 | Alibaba | Qwen3.8 Max | 1,388 裁判协议 v4.3.1 | -6 / +7 | Aug 2026 |
| 32 | Alibaba | Qwen3.8 27B (medium) | 1,376 裁判协议 v4.3.1 | -9 / +9 | Aug 2026 |
| 33 | OpenAI | GPT-5.6 Sol (high) | 1,361 裁判协议 v4.3.1 | -8 / +9 | Jul 2026 |
| 34 | Anthropic | Claude Sonnet 5 (Adaptive Reasoning, Max Effort) | 1,355 裁判协议 v4.3.1 | -6 / +6 | Jun 2026 |
| 35 | Meta | Muse Spark 1.2 (xhigh) | 1,344 裁判协议 v4.3.1 | -6 / +7 | Aug 2026 |
| 36 | OpenAI | GPT-5.6 Terra (xhigh) | 1,337 裁判协议 v4.3.1 | -9 / +9 | Jul 2026 |
| 37 | OpenAI | GPT-5.6 Luna (max) | 1,333 裁判协议 v4.3.1 | -6 / +6 | Jul 2026 |
| 38 | OpenAI | GPT-5.6 Terra (max) | 1,330 裁判协议 v4.3.1 | -7 / +7 | Jul 2026 |
| 39 | Anthropic | Claude Opus 4.8 (Adaptive Reasoning, Max Effort) | 1,316 裁判协议 v4.3.1 | -6 / +6 | May 2026 |
| 40 | Institute of Foundation Models | K2 Horizon 375B A23B | 1,298 裁判协议 v4.3.1 | -8 / +8 | Sep 2026 |
| 41 | SpaceXAI | Grok 4.6 (low) | 1,293 裁判协议 v4.3.1 | -7 / +7 | Aug 2026 |
| 42 | SpaceXAI | Grok 4.5 (high) | 1,283 裁判协议 v4.3.1 | -6 / +7 | Jul 2026 |
| 43 | Alibaba | Qwen3.8 27B (low) | 1,277 裁判协议 v4.3.1 | -10 / +10 | Aug 2026 |
| 44 | Anthropic | Claude Sonnet 5 (Adaptive Reasoning, Xhigh Effort) | 1,273 裁判协议 v4.3.1 | -6 / +6 | Jun 2026 |
| 45 | OpenAI | GPT-5.6 Luna (xhigh) | 1,268 裁判协议 v4.3.1 | -9 / +9 | Jul 2026 |
| 46 | DeepSeek | DeepSeek V4 Pro 0813 (Reasoning, Max Effort) | 1,265 裁判协议 v4.3.1 | -6 / +7 | Aug 2026 |
| 47 | OpenAI | GPT-6 Astra (low) | 1,253 裁判协议 v4.3.1 | -9 / +9 | Sep 2026 |
| 48 | DeepSeek | DeepSeek V4 Flash 0731 (Reasoning, Max Effort) | 1,253 裁判协议 v4.3.1 | -6 / +7 | Jul 2026 |
| 49 | Anthropic | Claude Opus 4.7 (Adaptive Reasoning, Max Effort) | 1,252 裁判协议 v4.3.1 | -6 / +6 | Apr 2026 |
| 50 | OpenAI | GPT-5.6 Sol (medium) | 1,237 裁判协议 v4.3.1 | -7 / +7 | Jul 2026 |
| 51 | Z AI | GLM-5.2 (max) | 1,231 裁判协议 v4.3.1 | -6 / +7 | Jun 2026 |
| 52 | Anthropic | Claude Opus 5 (Adaptive Reasoning, Low Effort) | 1,216 裁判协议 v4.3.1 | -6 / +7 | Jul 2026 |
| 53 | Gemini 3.8 Flash (high) | 1,202 裁判协议 v4.3.1 | -8 / +9 | Sep 2026 | |
| 54 | OpenAI | GPT-5.6 Terra (high) | 1,201 裁判协议 v4.3.1 | -9 / +9 | Jul 2026 |
| 55 | Sapiens AI | Agnes 2.5 Pro Alpha | 1,195 裁判协议 v4.3.1 | -9 / +9 | Jul 2026 |
| 56 | Anthropic | Claude Sonnet 5 (Adaptive Reasoning, High Effort) | 1,178 裁判协议 v4.3.1 | -6 / +6 | Jun 2026 |
| 57 | Gemini 3.8 Flash (medium) | 1,177 裁判协议 v4.3.1 | -10 / +10 | Sep 2026 | |
| 58 | OpenAI | GPT-5.6 Luna (high) | 1,173 裁判协议 v4.3.1 | -8 / +8 | Jul 2026 |
| 59 | Institute of Foundation Models | K2 Horizon MoVA 36B A4B | 1,170 裁判协议 v4.3.1 | -10 / +11 | Sep 2026 |
| 60 | OpenAI | GPT-5.5 (xhigh) | 1,136 裁判协议 v4.3.1 | -6 / +7 | Apr 2026 |
| 61 | Multiverse Computing | Quasar 438B (max, based on GLM-5.2) | 1,123 裁判协议 v4.3.1 | -10 / +10 | Aug 2026 |
| 62 | Gemini 3.7 Flash (high) | 1,116 裁判协议 v4.3.1 | -6 / +7 | Aug 2026 | |
| 63 | MiniMax | MiniMax-M3 | 1,096 裁判协议 v4.3.1 | -6 / +7 | Jun 2026 |
| 64 | OpenAI | GPT-5.5 (high) | 1,089 裁判协议 v4.3.1 | -6 / +6 | Apr 2026 |
| 65 | OpenAI | GPT-5.5 Instant (June 2026) | 1,082 裁判协议 v4.3.1 | -9 / +10 | Jun 2026 |
| 66 | Anthropic | Claude Opus 4.7 (Non-reasoning, High Effort) | 1,071 裁判协议 v4.3.1 | -6 / +6 | Apr 2026 |
| 67 | Anthropic | Claude Sonnet 4.6 (Adaptive Reasoning, Max Effort) | 1,065 裁判协议 v4.3.1 | -6 / +6 | Feb 2026 |
| 68 | Gemini 3.8 Flash (low) | 1,056 裁判协议 v4.3.1 | -9 / +9 | Sep 2026 | |
| 69 | Anthropic | Claude Sonnet 5 (Adaptive Reasoning, Medium Effort) | 1,054 裁判协议 v4.3.1 | -6 / +6 | Jun 2026 |
| 70 | DeepSeek | DeepSeek V4 Flash 0420 (Reasoning, High Effort) | 1,052 裁判协议 v4.3.1 | -11 / +10 | Apr 2026 |
| 71 | OpenAI | GPT-5.6 Sol (low) | 1,042 裁判协议 v4.3.1 | -8 / +9 | Jul 2026 |
| 72 | OpenAI | GPT-5.6 Terra (medium) | 1,041 裁判协议 v4.3.1 | -8 / +8 | Jul 2026 |
| 73 | Institute of Foundation Models | K2 Horizon 7B | 1,021 裁判协议 v4.3.1 | -10 / +10 | Sep 2026 |
| 74 | OpenAI | GPT-5.6 Sol (Non-reasoning) | 1,015 裁判协议 v4.3.1 | -7 / +8 | Jul 2026 |
| 75 | OpenAI | GPT-5.6 Terra (low) | 1,014 裁判协议 v4.3.1 | -9 / +8 | Jul 2026 |
| 76 | OpenAI | GPT-5.5 (medium) | 1,000 裁判协议 v4.3.1 | -0 / +0 | Apr 2026 |
| 77 | InclusionAI | Ling-3.0-flash-VL | 986 裁判协议 v4.3.1 | -10 / +10 | Sep 2026 |
| 78 | Kimi | Kimi K3 (low) | 985 裁判协议 v4.3.1 | -8 / +8 | Jul 2026 |
| 79 | InclusionAI | Ling-3.0-flash-Fin | 967 裁判协议 v4.3.1 | -10 / +10 | Sep 2026 |
| 80 | Z AI | GLM-5.1 (Reasoning) | 964 裁判协议 v4.3.1 | -6 / +6 | Apr 2026 |
| 81 | Gemini 3.6 Flash (high) | 953 裁判协议 v4.3.1 | -7 / +7 | Jul 2026 | |
| 82 | Tencent | Hy3 | 950 裁判协议 v4.3.1 | -10 / +10 | Jul 2026 |
| 83 | OpenAI | GPT-5.6 Terra (Non-reasoning) | 948 裁判协议 v4.3.1 | -8 / +8 | Jul 2026 |
| 84 | OpenAI | GPT-5.6 Luna (medium) | 935 裁判协议 v4.3.1 | -8 / +8 | Jul 2026 |
| 85 | Anthropic | Claude Sonnet 5 (Adaptive Reasoning, Low Effort) | 927 裁判协议 v4.3.1 | -7 / +7 | Jun 2026 |
| 86 | DeepSeek | DeepSeek V4 Pro 0424 (Reasoning, Max Effort) | 925 裁判协议 v4.3.1 | -6 / +6 | Apr 2026 |
| 87 | Alibaba | Qwen3.7 Max | 918 裁判协议 v4.3.1 | -6 / +6 | May 2026 |
| 88 | Alibaba | Qwen3.7 Plus | 913 裁判协议 v4.3.1 | -9 / +9 | Jun 2026 |
| 89 | Thinking Machines | Inkling Small | 912 裁判协议 v4.3.1 | -7 / +8 | Jul 2026 |
| 90 | Xiaomi | MiMo-V2.5-Pro | 883 裁判协议 v4.3.1 | -7 / +7 | Apr 2026 |
| 91 | NVIDIA | Nemotron 3 Ultra 550B A55B (Reasoning) | 878 裁判协议 v4.3.1 | -7 / +7 | Jun 2026 |
| 92 | Gemini 3.5 Flash (medium) | 871 裁判协议 v4.3.1 | -7 / +7 | May 2026 | |
| 93 | Gemini 3.5 Flash (high) | 870 裁判协议 v4.3.1 | -7 / +6 | May 2026 | |
| 94 | OpenAI | GPT-5.3 Codex (xhigh) | 866 裁判协议 v4.3.1 | -7 / +6 | Feb 2026 |
| 95 | Meta | Muse Spark 1.1 (xhigh) | 860 裁判协议 v4.3.1 | -8 / +8 | Jul 2026 |
| 96 | Kimi | Kimi K2.7 Code | 856 裁判协议 v4.3.1 | -10 / +9 | Jun 2026 |
| 97 | Thinking Machines | Inkling (xhigh) | 836 裁判协议 v4.3.1 | -8 / +7 | Jul 2026 |
| 98 | DeepSeek | DeepSeek V4 Flash 0420 (Reasoning, Max Effort) | 833 裁判协议 v4.3.1 | -6 / +6 | Apr 2026 |
| 99 | Kimi | Kimi K2.6 | 824 裁判协议 v4.3.1 | -6 / +6 | Apr 2026 |
| 100 | Alibaba | Qwen3.6 27B (Reasoning) | 812 裁判协议 v4.3.1 | -7 / +7 | Apr 2026 |
| 101 | InclusionAI | Ling 3.0 Flash | 802 裁判协议 v4.3.1 | -8 / +7 | Aug 2026 |
| 102 | Institute of Foundation Models | K2 Horizon 3.7B | 793 裁判协议 v4.3.1 | -11 / +10 | Sep 2026 |
| 103 | SpaceXAI | Grok 4.3 (high) | 762 裁判协议 v4.3.1 | -7 / +7 | Apr 2026 |
| 104 | Xiaomi | MiMo-V2.5 | 751 裁判协议 v4.3.1 | -9 / +9 | Apr 2026 |
| 105 | OpenAI | GPT-5.6 Luna (low) | 744 裁判协议 v4.3.1 | -9 / +9 | Jul 2026 |
| 106 | LongCat | LongCat 2.0 | 739 裁判协议 v4.3.1 | -10 / +10 | Jun 2026 |
| 107 | SpaceXAI | Grok 4.3 (Non-reasoning) | 724 裁判协议 v4.3.1 | -10 / +10 | Apr 2026 |
| 108 | OpenAI | GPT-5.4 mini (xhigh) | 719 裁判协议 v4.3.1 | -7 / +7 | Mar 2026 |
| 109 | MiniMax | MiniMax-M2.7 | 715 裁判协议 v4.3.1 | -9 / +10 | Mar 2026 |
| 110 | Anthropic | Claude 4.5 Sonnet (Reasoning) | 712 裁判协议 v4.3.1 | -10 / +9 | Sep 2025 |
| 111 | Alibaba | Qwen3.6 35B A3B (Reasoning) | 711 裁判协议 v4.3.1 | -9 / +8 | Apr 2026 |
| 112 | OpenAI | GPT-5.6 Luna (Non-reasoning) | 710 裁判协议 v4.3.1 | -8 / +8 | Jul 2026 |
| 113 | OpenAI | GPT-5.4 nano (xhigh) | 670 裁判协议 v4.3.1 | -10 / +10 | Mar 2026 |
| 114 | InclusionAI | Ring-2.6-1T | 654 裁判协议 v4.3.1 | -10 / +9 | May 2026 |
| 115 | Gemini 3.5 Flash-Lite | 645 裁判协议 v4.3.1 | -8 / +8 | Jul 2026 | |
| 116 | Meta | Muse Spark | 644 裁判协议 v4.3.1 | -7 / +7 | Apr 2026 |
| 117 | Anthropic | Claude 4.5 Haiku (Reasoning) | 614 裁判协议 v4.3.1 | -7 / +7 | Oct 2025 |
| 118 | KwaiKAT | KAT-Coder-Pro V1 | 602 裁判协议 v4.3.1 | -8 / +8 | Nov 2025 |
| 119 | Alibaba | Qwen3.5 397B A17B (Reasoning) | 563 裁判协议 v4.3.1 | -8 / +7 | Feb 2026 |
| 120 | Mistral | Mistral Medium 3.1 | 538 裁判协议 v4.3.1 | -9 / +8 | Aug 2025 |
| 121 | Mistral | Mistral Medium 3.5 | 524 裁判协议 v4.3.1 | -8 / +7 | Apr 2026 |
| 122 | NVIDIA | Nemotron 3.5 Lightning | 509 裁判协议 v4.3.1 | -10 / +10 | Aug 2026 |
| 123 | Alibaba | Qwen3.5 122B A10B (Reasoning) | 499 裁判协议 v4.3.1 | -9 / +9 | Feb 2026 |
| 124 | Mistral | Devstral Small 2 | 496 裁判协议 v4.3.1 | -10 / +9 | Dec 2025 |
| 125 | Mistral | Devstral 2 | 483 裁判协议 v4.3.1 | -11 / +11 | Dec 2025 |
| 126 | InclusionAI | Ling 3.0 Tiny | 481 裁判协议 v4.3.1 | -10 / +9 | Aug 2026 |
| 127 | Meta | Muse Glimmer (high) | 481 裁判协议 v4.3.1 | -9 / +8 | Aug 2026 |
| 128 | DeepSeek | DeepSeek V3.1 Terminus (Reasoning) | 464 裁判协议 v4.3.1 | -10 / +9 | Sep 2025 |
| 129 | Gemini 3.1 Pro Preview | 459 裁判协议 v4.3.1 | -8 / +8 | Feb 2026 | |
| 130 | OpenAI | GPT-5 mini (high) | 438 裁判协议 v4.3.1 | -10 / +9 | Aug 2025 |
| 131 | OpenBMB | MiniCPM5-2B | 434 裁判协议 v4.3.1 | -11 / +11 | Sep 2026 |
| 132 | Alibaba | Qwen3 Coder Next | 414 裁判协议 v4.3.1 | -10 / +10 | Feb 2026 |
| 133 | Gemma 4 31B (Reasoning) | 377 裁判协议 v4.3.1 | -8 / +8 | Apr 2026 | |
| 134 | Cohere | Command A+ | 366 裁判协议 v4.3.1 | -10 / +9 | May 2026 |
| 135 | Mistral | Mistral Small 3.1 | 325 裁判协议 v4.3.1 | -10 / +9 | Mar 2025 |
| 136 | IBM | Granite 4.2 8B | 320 裁判协议 v4.3.1 | -10 / +9 | Aug 2026 |
| 137 | Gemini 2.5 Pro | 315 裁判协议 v4.3.1 | -10 / +9 | Jun 2025 | |
| 138 | Inception | Mercury 2 | 280 裁判协议 v4.3.1 | -9 / +9 | Feb 2026 |
| 139 | Arcee AI | Trinity Large Thinking | 271 裁判协议 v4.3.1 | -10 / +9 | Apr 2026 |
| 140 | Mistral | Mistral Small 4 (Reasoning) | 256 裁判协议 v4.3.1 | -10 / +10 | Mar 2026 |
| 141 | Multiverse Computing | HyperNova 60B 2605 (high, based on gpt-oss-120b) | 249 裁判协议 v4.3.1 | -10 / +10 | May 2026 |
| 142 | Mistral | Mistral Large 3 | 246 裁判协议 v4.3.1 | -9 / +9 | Dec 2025 |
| 143 | Cohere | North Mini Code | 234 裁判协议 v4.3.1 | -10 / +8 | Jun 2026 |
| 144 | Gemini 3.1 Flash-Lite | 226 裁判协议 v4.3.1 | -9 / +8 | Mar 2026 | |
| 145 | Alibaba | Qwen3 235B A22B 2507 (Reasoning) | 190 裁判协议 v4.3.1 | -12 / +10 | Jul 2025 |
| 146 | Celeris | Celeris-1 | 172 裁判协议 v4.3.1 | -10 / +9 | Jul 2026 |
| 147 | OpenAI | o3-mini (high) | 161 裁判协议 v4.3.1 | -10 / +10 | Jan 2025 |
| 148 | Mistral | Ministral 3 14B | 150 裁判协议 v4.3.1 | -10 / +9 | Dec 2025 |
| 149 | DeepSeek | DeepSeek R1 (Jan '25) | 150 裁判协议 v4.3.1 | -10 / +10 | Jan 2025 |
| 150 | Upstage | Solar Pro 3 | 122 裁判协议 v4.3.1 | -9 / +9 | Apr 2026 |
| 151 | IBM | Granite 4.2 3B | 121 裁判协议 v4.3.1 | -11 / +10 | Aug 2026 |
| 152 | NVIDIA | NVIDIA Nemotron 3 Nano 30B A3B (Reasoning) | 100 裁判协议 v4.3.1 | -11 / +9 | Dec 2025 |
| 153 | DeepSeek | DeepSeek V3 (Dec '24) | 100 裁判协议 v4.3.1 | -12 / +11 | Dec 2024 |
| 154 | Mistral | Ministral 3 8B | 99 裁判协议 v4.3.1 | -11 / +10 | Dec 2025 |
| 155 | DeepSeek | DeepSeek V3 0324 | 92 裁判协议 v4.3.1 | -11 / +10 | Mar 2025 |
| 156 | Institute of Foundation Models | K2 Think V2 | 34 裁判协议 v4.3.1 | -9 / +9 | Dec 2025 |
| 157 | Alibaba | Qwen3 30B A3B 2507 (Reasoning) | 32 裁判协议 v4.3.1 | -12 / +11 | Jul 2025 |
| 158 | Mistral | Ministral 3 3B | 16 裁判协议 v4.3.1 | -11 / +9 | Dec 2025 |
| 159 | Alibaba | Qwen3 32B (Reasoning) | 9 裁判协议 v4.3.1 | -9 / +10 | Apr 2025 |
| 160 | NVIDIA | Nemotron 3 Super 120B A12B (Reasoning) | 0 裁判协议 v4.3.1 | -0 / +0 | Mar 2026 |
| 161 | OpenAI | gpt-oss-120b (high) | 0 裁判协议 v4.3.1 | -0 / +0 | Aug 2025 |
| 162 | Meta | Llama 4 Maverick | 0 裁判协议 v4.3.1 | -0 / +0 | Apr 2025 |
| 163 | OpenAI | gpt-oss-20b (high) | 0 裁判协议 v4.3.1 | -0 / +0 | Aug 2025 |
| 164 | Mistral | Mistral Small 3.2 | 0 裁判协议 v4.3.1 | -0 / +0 | Jun 2025 |
| 165 | Meta | Llama 4 Scout | 0 裁判协议 v4.3.1 | -0 / +0 | Apr 2025 |
| 166 | Alibaba | Qwen3 14B (Reasoning) | 0 裁判协议 v4.3.1 | -0 / +0 | Apr 2025 |
| 167 | Alibaba | Qwen3 8B (Reasoning) | 0 裁判协议 v4.3.1 | -0 / +0 | Apr 2025 |
| 168 | Institute of Foundation Models | K2 Horizon 0.9B | 0 裁判协议 v4.3.1 | -0 / +0 | Sep 2026 |
| — | OpenAI | GPT-6 Astra (Non-reasoning) 本轮来源未列出，保留上次观测 | 1,471 历史参考 2026-09-12 | -8 / +9 历史参考 2026-09-12 | Sep 2026 |
实时 Elo 的裁判协议与可比边界
AA-Briefcase 与 GDPval-AA v2 的当前成对比较使用裁判协议 v4.3.1，原评测名称与任务版本不变，Intelligence Index 仍为 v4.3。未区分裁判协议的旧 Elo 保留原始观测，仅作历史参考，不与当前协议混排。
指数中的 Elo 贡献在模型加入时冻结；实时榜单分数不能用于重算指数。来源未给出逐模型重跑日期，不能把裁判更新或跨协议分数差解释成新的模型运行或能力变化。
查看裁判、核验时间与完整方法边界 →查看 AA 原始方法 ↗