这个评测关注什么
同时观察回答正确性与模型是否知道自己不知道。
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”。
知识边界
同时观察回答正确性与模型是否知道自己不知道。
| 排名 | 机构 | 模型 | 知识可靠指数 | 准确率 | 幻觉率 | 作答率 | 正确 | 错误 | 未作答 |
|---|---|---|---|---|---|---|---|---|---|
| 1 | Anthropic | Claude Fable 5 | 40.15 | 61.4% | 54.9% | 88.6% | 3681 | 1272 | 684 |
| 2 | Gemini 3.1 Pro Preview | 32.93 | 55.3% | 49.9% | 79.4% | 3315 | 1339 | 1238 | |
| 3 | Anthropic | Claude Opus 4.8 (max) | 27.43 | 46.6% | 35.9% | 69.6% | 2796 | 1150 | 1824 |
| 4 | SpaceXAI | Grok 4.5 (high) | 26.38 | 52% | 53.5% | 78.8% | 3123 | 1540 | 1269 |
| 5 | Anthropic | Claude Opus 4.7 (max) | 26.17 | 45.8% | 36.2% | 69.5% | 2747 | 1177 | 1831 |
| 6 | Gemini 3.5 Flash | 22.68 | 51.9% | 60.7% | 83.2% | 3113 | 1752 | 1010 | |
| 7 | OpenAI | GPT-5.6 Sol (max) | 21.7 | 58.5% | 88.8% | 98.3% | 3512 | 2210 | 104 |
| 8 | Gemini 3.5 Flash (medium) | 21.65 | 51.1% | 60.3% | 82.9% | 3067 | 1768 | 1028 | |
| 9 | OpenAI | GPT-5.6 Sol (xhigh) | 20.55 | 58% | 89% | 98.1% | 3478 | 2245 | 114 |
| 10 | OpenAI | GPT-5.5 (xhigh) | 20.07 | 56.9% | 85.5% | 96.9% | 3415 | 2211 | 184 |
| 11 | OpenAI | GPT-5.6 Sol (high) | 19.75 | 57.3% | 88% | 97.8% | 3439 | 2254 | 135 |
| 12 | OpenAI | GPT-5.6 Sol (medium) | 18.95 | 56.7% | 87.1% | 97.5% | 3401 | 2264 | 151 |
| 13 | Kimi | Kimi K3 | 18.42 | 46% | 50.9% | 76.8% | 2757 | 1652 | 1390 |
| 14 | OpenAI | GPT-5.6 Sol (low) | 18.4 | 56.4% | 87.3% | 97% | 3386 | 2282 | 177 |
| 15 | OpenAI | GPT-5.5 (high) | 18.37 | 56% | 85.7% | 97% | 3363 | 2261 | 183 |
| 16 | SpaceXAI | Grok 4.3 (high) | 18.32 | 34.6% | 25% | 52.3% | 2078 | 979 | 2863 |
| 17 | Meta | Muse Spark 1.1 (xhigh) | 18 | 40.6% | 38.1% | 82.3% | 2437 | 1357 | 1062 |
| 18 | OpenAI | GPT-5.5 (medium) | 17.28 | 55.6% | 86.3% | 97% | 3336 | 2299 | 183 |
| 19 | SpaceXAI | Grok 4.3 (medium) | 16.98 | 28.4% | 16% | 41.4% | 1707 | 688 | 3513 |
| 20 | Gemini 3 Pro Preview (high) | 15.8 | 55.9% | 90.9% | 98.6% | 3354 | 2406 | 87 | |
| 21 | SpaceXAI | Grok 4.20 0309 v2 | 15.35 | 27.5% | 16.7% | 40.6% | 1649 | 728 | 3565 |
| 22 | Anthropic | Claude Sonnet 5 (max) | 15.32 | 38.3% | 37.3% | 64.9% | 2298 | 1379 | 2108 |
| 23 | OpenAI | GPT-5.5 (low) | 14.62 | 54.1% | 86% | 96.3% | 3245 | 2368 | 220 |
| 24 | Anthropic | Claude Opus 4.7 (Non-reasoning, high) | 14.22 | 43.5% | 51.9% | 75.6% | 2612 | 1759 | 1461 |
| 25 | Alibaba | Qwen3.7 Max | 14.08 | 30.1% | 22.9% | 48% | 1807 | 962 | 3120 |
| 26 | SpaceXAI | Grok 4.3 (low) | 13.67 | 25.9% | 16.5% | 40% | 1553 | 733 | 3599 |
| 27 | Anthropic | Claude Opus 4.6 (max) | 13.5 | 46.4% | 61.3% | 81.5% | 2783 | 1973 | 1107 |
| 28 | SpaceXAI | Grok 4.20 0309 | 13.37 | 28.9% | 21.9% | 45.1% | 1735 | 933 | 3291 |
| 29 | Anthropic | Claude Opus 4.5 | 13.27 | 45.7% | 59.8% | 81.3% | 2743 | 1947 | 1120 |
| 30 | Anthropic | Claude Sonnet 4.6 (max) | 12.37 | 40% | 46.1% | 70.8% | 2402 | 1660 | 1751 |
| 31 | Gemini 3 Flash | 11.57 | 54% | 92.2% | 98.6% | 3240 | 2546 | 82 | |
| 32 | OpenAI | GPT-5.5 Instant (May 2026) | 10.33 | 45.5% | 64.4% | 83.1% | 2728 | 2108 | 1012 |
| 33 | Alibaba | Qwen3.6 Max Preview | 10.2 | 37.7% | 44.2% | 67.3% | 2263 | 1651 | 1963 |
| 34 | OpenAI | GPT-5.3 Codex (xhigh) | 9.88 | 51.8% | 86.9% | 96.4% | 3107 | 2514 | 218 |
| 35 | SpaceXAI | Grok Build 0.1 0616 | 7.43 | 51.4% | 90.5% | 97.8% | 3084 | 2638 | 133 |
| 36 | Kimi | Kimi K2.6 | 6.42 | 32.8% | 39.3% | 60.8% | 1968 | 1583 | 2352 |
| 37 | OpenAI | GPT-5.4 (xhigh) | 5.65 | 50% | 88.6% | 97% | 2999 | 2660 | 179 |
| 38 | OpenAI | GPT-5.1 (high) | 5.55 | 37.6% | 51.3% | 71% | 2255 | 1922 | 1739 |
| 39 | Xiaomi | MiMo-V2-Pro | 4.92 | 26.8% | 29.9% | 49.4% | 1609 | 1314 | 3039 |
| 40 | OpenAI | GPT-5.4 (low) | 4.72 | 47.4% | 81.1% | 92.4% | 2843 | 2560 | 455 |
| 41 | Meta | Muse Spark | 4.08 | 44.6% | 73.2% | 95.3% | 2677 | 2432 | 284 |
| 42 | Z AI | GLM-5.2 (max) | 3.97 | 25.1% | 28.1% | 47.2% | 1503 | 1265 | 3167 |
| 43 | SpaceXAI | Grok 4 | 3.75 | 41.4% | 64.2% | 79.8% | 2483 | 2258 | 1211 |
| 44 | Xiaomi | MiMo-V2.5-Pro | 3.6 | 22.6% | 24.5% | 42.3% | 1356 | 1140 | 3465 |
| 45 | OpenAI | GPT-5.5 Instant (June 2026) | 3.48 | 44.1% | 72.7% | 86.7% | 2647 | 2438 | 798 |
| 46 | Anthropic | Claude Opus 4.6 (high) | 3.47 | 45.1% | 76% | 90.4% | 2709 | 2501 | 574 |
| 47 | Alibaba | Qwen3.6 Plus | 2.65 | 26.2% | 32% | 51.2% | 1574 | 1415 | 2927 |
| 48 | Alibaba | Qwen3.7 Plus | 2.37 | 22.2% | 25.5% | 43.6% | 1332 | 1190 | 3386 |
| 49 | Gemini 3 Pro Preview (low) | 2.15 | 47.1% | 84.9% | 98.2% | 2824 | 2695 | 111 | |
| 50 | Thinking Machines | Inkling | 2.05 | 40% | 63.1% | 81% | 2397 | 2274 | 1141 |
| 51 | Z AI | GLM-5 | 2 | 26.9% | 34% | 53.1% | 1612 | 1492 | 2811 |
| 52 | Z AI | GLM-5.1 | 1.93 | 24.2% | 29.4% | 47.5% | 1452 | 1336 | 3152 |
| 53 | MiniMax | MiniMax-M3 | 1.37 | 15% | 16.1% | 30.9% | 902 | 820 | 4146 |
| 54 | Gemini 3.5 Flash (minimal) | 1.3 | 43.1% | 73.4% | 87% | 2585 | 2507 | 782 | |
| 55 | OpenAI | GPT-5.6 Sol (Non-reasoning) | 0.85 | 48.2% | 91.3% | 97.9% | 2891 | 2840 | 129 |
| 56 | MiniMax | MiniMax-M2.7 | 0.68 | 26.1% | 34.4% | 53.2% | 1567 | 1526 | 2808 |
| 57 | Anthropic | Claude 4.5 Sonnet | 0.33 | 32.4% | 47.4% | 67.2% | 1943 | 1923 | 1967 |
| 58 | Anthropic | Claude 3.7 Sonnet | 0.32 | 28% | 38.5% | 58.1% | 1682 | 1663 | 2516 |
| 59 | Anthropic | Claude 4 Sonnet | 0.18 | 22.3% | 28.5% | 46.3% | 1340 | 1329 | 3223 |
| 60 | OpenAI | GPT-5.2 (medium) | 0.02 | 37.7% | 60.6% | 77.4% | 2264 | 2263 | 1358 |
| 61 | OpenAI | GPT-5.6 Terra (max) | -0.22 | 45.9% | 85.2% | 94.6% | 2753 | 2766 | 322 |
| 62 | OpenBMB | MiniCPM5-1B | -0.78 | 0.1% | 0.9% | 0.9% | 4 | 51 | 5944 |
| 63 | NVIDIA | Nemotron 3 Ultra | -0.83 | 21.6% | 28.5% | 47.2% | 1293 | 1343 | 3166 |
| 64 | OpenAI | GPT-5.2 (xhigh) | -1 | 43.8% | 79.7% | 90.7% | 2628 | 2688 | 560 |
| 65 | Anthropic | Claude Sonnet 5 (Non-reasoning) | -1.18 | 32.6% | 50.1% | 69% | 1955 | 2026 | 1861 |
| 66 | Anthropic | Claude Sonnet 4.6 (Non-reasoning, Low Effort) | -1.9 | 36.1% | 59.6% | 76.3% | 2168 | 2282 | 1424 |
| 67 | OpenAI | GPT-5.2 Codex (xhigh) | -2.48 | 40.7% | 72.8% | 85.8% | 2442 | 2591 | 850 |
| 68 | Anthropic | Claude Sonnet 4.6 (Non-reasoning) | -2.93 | 38% | 65.9% | 81.5% | 2278 | 2454 | 1108 |
| 69 | OpenAI | GPT-5.6 Terra (xhigh) | -3.3 | 44.7% | 86.7% | 95.3% | 2681 | 2879 | 281 |
| 70 | Gemini 3 Flash | -3.62 | 45.5% | 90.2% | 98.2% | 2731 | 2948 | 107 | |
| 71 | Anthropic | Claude Opus 4.5 | -3.95 | 40.7% | 75.4% | 88.3% | 2444 | 2681 | 702 |
| 72 | Cohere | Command A+ | -3.98 | 8.8% | 14.1% | 22.3% | 531 | 770 | 4665 |
| 73 | OpenAI | GPT-5.6 Terra (high) | -4.07 | 44.5% | 87.5% | 95.5% | 2669 | 2913 | 269 |
| 74 | Anthropic | Claude 4.5 Haiku | -4.22 | 17.4% | 26.2% | 41.5% | 1046 | 1299 | 3512 |
| 75 | OpenAI | GPT-5.5 (Non-reasoning) | -5.05 | 45.1% | 91.2% | 97.7% | 2704 | 3007 | 137 |
| 76 | OpenAI | GPT-5.6 Terra (medium) | -5.17 | 44% | 87.9% | 95.7% | 2641 | 2951 | 256 |
| 77 | OpenAI | GPT-5.1 Codex (high) | -6.02 | 39.2% | 74.4% | 87.7% | 2352 | 2713 | 739 |
| 78 | SpaceXAI | Grok 3 mini Reasoning (high) | -6.05 | 15.4% | 25.4% | 37.5% | 924 | 1287 | 3751 |
| 79 | Z AI | GLM-5.2 | -6.4 | 20.3% | 33.5% | 48% | 1219 | 1603 | 3120 |
| 80 | OpenAI | GPT-5 Codex (high) | -6.77 | 38.7% | 74.1% | 86.1% | 2320 | 2726 | 837 |
| 81 | OpenAI | GPT-5.6 Terra (low) | -7.12 | 43% | 87.8% | 95.6% | 2577 | 3004 | 266 |
| 82 | Anthropic | Claude 4.5 Haiku | -7.67 | 13.7% | 24.7% | 37% | 820 | 1280 | 3777 |
| 83 | OpenAI | GPT-5 (high) | -8.08 | 40.6% | 82.1% | 90.6% | 2439 | 2924 | 562 |
| 84 | Kimi | Kimi K2.5 | -8.12 | 34.3% | 64.6% | 79.8% | 2059 | 2546 | 1211 |
| 85 | Anthropic | Claude 4.5 Sonnet | -9.2 | 27.5% | 50.7% | 68.7% | 1651 | 2203 | 1876 |
| 86 | Anthropic | Claude 3.7 Sonnet | -9.23 | 28.3% | 52.3% | 67.9% | 1698 | 2252 | 1924 |
| 87 | Anthropic | Claude 4 Sonnet | -9.23 | 22.4% | 40.8% | 56.4% | 1346 | 1900 | 2617 |
| 88 | Xiaomi | MiMo-V2.5 | -9.33 | 17.3% | 32.2% | 44.4% | 1036 | 1596 | 3336 |
| 89 | DeepSeek | DeepSeek V4 Pro (high) | -9.7 | 41.8% | 88.6% | 94.8% | 2510 | 3092 | 313 |
| 90 | Kimi | Kimi K2.6 | -9.97 | 23.1% | 42.9% | 59.4% | 1384 | 1982 | 2435 |
| 91 | DeepSeek | DeepSeek V4 Pro (max) | -10.02 | 43.3% | 94% | 98.2% | 2597 | 3198 | 109 |
| 92 | OpenAI | GPT-5 (low) | -10.02 | 37.4% | 75.8% | 88.2% | 2246 | 2847 | 706 |
| 93 | OpenAI | GPT-5 (medium) | -10.08 | 38.9% | 80.1% | 91.1% | 2332 | 2937 | 536 |
| 94 | China Mobile | JT-4.1 Flash 236B A21B | -10.12 | 23.1% | 43.3% | 59.3% | 1388 | 1995 | 2442 |
| 95 | OpenAI | o1 | -10.55 | 34.7% | 69.3% | 81.2% | 2081 | 2714 | 1130 |
| 96 | Kimi | Kimi K2.7 Code | -10.7 | 38.6% | 80.3% | 90.6% | 2316 | 2958 | 561 |
| 97 | OpenAI | GPT-4o (Nov) | -10.75 | 19.7% | 37.9% | 51.7% | 1181 | 1826 | 2897 |
| 98 | OpenAI | GPT-5 mini (medium) | -10.78 | 22.2% | 42.4% | 58.8% | 1332 | 1979 | 2474 |
| 99 | OpenAI | GPT-5.6 Luna (max) | -11.23 | 41.5% | 90.1% | 96.9% | 2490 | 3164 | 184 |
| 100 | Z AI | GLM-5 | -11.23 | 22.7% | 43.9% | 59.5% | 1361 | 2035 | 2429 |
| 101 | OpenAI | GPT-5.6 Luna (xhigh) | -11.6 | 41.3% | 90.1% | 96.8% | 2478 | 3174 | 191 |
| 102 | Alibaba | Qwen3.5 Omni Plus | -12.3 | 17.4% | 36% | 48.9% | 1045 | 1783 | 3069 |
| 103 | OpenAI | GPT-5.6 Luna (high) | -12.45 | 40.7% | 89.5% | 96.6% | 2440 | 3187 | 206 |
| 104 | OpenAI | GPT-5.2 | -13.72 | 29.3% | 60.8% | 75.4% | 1757 | 2580 | 1477 |
| 105 | OpenAI | GPT-5.6 Luna (medium) | -13.82 | 39.6% | 88.5% | 95.6% | 2378 | 3207 | 263 |
| 106 | Kimi | Kimi K2.5 | -14 | 22.9% | 47.9% | 64.3% | 1374 | 2214 | 2142 |
| 107 | Xiaomi | MiMo-V2-Omni-0327 | -14.02 | 18.5% | 39.9% | 51.7% | 1109 | 1950 | 2900 |
| 108 | Gemini 2.5 Pro | -14.3 | 39% | 87.4% | 95.8% | 2340 | 3198 | 251 | |
| 109 | Z AI | GLM-5-Turbo | -15.08 | 29% | 62.2% | 74.6% | 1742 | 2647 | 1525 |
| 110 | OpenAI | GPT-5.6 Luna (low) | -15.2 | 38.6% | 87.7% | 95.1% | 2318 | 3230 | 295 |
| 111 | OpenAI | o3 | -15.27 | 38.4% | 87.1% | 93.9% | 2304 | 3220 | 365 |
| 112 | Gemini 3.1 Flash-Lite | -15.52 | 36.4% | 81.6% | 91.1% | 2184 | 3115 | 533 | |
| 113 | OpenAI | GPT-5.4 (Non-reasoning) | -15.88 | 36.8% | 83.2% | 91.9% | 2205 | 3158 | 484 |
| 114 | DeepSeek | DeepSeek V3.2 Speciale | -16.48 | 38.4% | 88.9% | 94.6% | 2301 | 3290 | 325 |
| 115 | OpenAI | GPT-5.1 Codex mini (high) | -16.52 | 22.7% | 50.8% | 63.9% | 1363 | 2354 | 2166 |
| 116 | OpenBMB | MiniCPM5-1B | -16.62 | 1.7% | 18.7% | 20.2% | 104 | 1101 | 4790 |
| 117 | OpenAI | GPT-5 mini (high) | -17.18 | 23.9% | 54.1% | 68.9% | 1437 | 2468 | 1866 |
| 118 | Motif Technologies | Motif 3 (Beta) | -17.28 | 21% | 48.5% | 60.6% | 1262 | 2299 | 2367 |
| 119 | Meta | Llama 3.1 405B | -17.3 | 22.3% | 51% | 64.6% | 1340 | 2378 | 2126 |
| 120 | Xiaomi | MiMo-V2-Omni | -17.43 | 18.7% | 44.4% | 61.7% | 1121 | 2167 | 2299 |
| 121 | Xiaomi | MiMo-V2-Flash (Feb 2026) | -18.47 | 20.2% | 48.4% | 59.6% | 1210 | 2318 | 2424 |
| 122 | Tencent | Hy3 | -18.5 | 31.5% | 73% | 83.7% | 1890 | 3000 | 978 |
| 123 | OpenAI | GPT-5.4 mini (xhigh) | -18.68 | 37.5% | 89.8% | 95.3% | 2248 | 3369 | 283 |
| 124 | OpenAI | GPT-5.4 nano | -18.82 | 21% | 50.4% | 62.5% | 1260 | 2389 | 2247 |
| 125 | Z AI | GLM 5V Turbo | -18.98 | 29.1% | 67.9% | 79.7% | 1748 | 2887 | 1218 |
| 126 | Alibaba | Qwen3.6 27B | -19.78 | 19.2% | 48.3% | 59.4% | 1153 | 2340 | 2435 |
| 127 | Gemma 4 E4B | -20.05 | 8.6% | 31.3% | 37.9% | 513 | 1716 | 3726 | |
| 128 | OpenAI | GPT-5.4 mini (medium) | -20.12 | 36.5% | 89.2% | 94.8% | 2190 | 3397 | 313 |
| 129 | Kimi | Kimi K2 Thinking | -20.48 | 30% | 72.2% | 84.6% | 1801 | 3030 | 922 |
| 130 | DeepSeek | DeepSeek V3.2 | -20.88 | 33.5% | 81.7% | 89.8% | 2008 | 3261 | 613 |
| 131 | Z AI | GLM-5.1 | -21.4 | 25.7% | 63.4% | 73.7% | 1543 | 2827 | 1581 |
| 132 | Alibaba | Qwen3.6 35B A3B | -21.43 | 18.9% | 49.7% | 60.3% | 1133 | 2419 | 2381 |
| 133 | KwaiKAT | KAT-Coder-Pro V2 | -21.6 | 22% | 55.8% | 66.8% | 1318 | 2614 | 1995 |
| 134 | OpenAI | GPT-4o (Aug) | -21.72 | 23.7% | 59.5% | 70.3% | 1420 | 2723 | 1785 |
| 135 | Alibaba | Qwen3 Coder 480B | -22.2 | 14.7% | 43.3% | 55.9% | 883 | 2215 | 2649 |
| 136 | DeepSeek | DeepSeek V4 Flash (high) | -22.28 | 35.5% | 89.7% | 94.5% | 2132 | 3469 | 327 |
| 137 | LongCat | LongCat 2.0 | -22.63 | 29.8% | 74.7% | 83.8% | 1788 | 3146 | 974 |
| 138 | DeepSeek | DeepSeek V4 Flash (max) | -22.9 | 37.2% | 95.8% | 99.1% | 2233 | 3607 | 57 |
| 139 | Anthropic | Claude 3.5 Haiku | -23.18 | 13% | 41.5% | 50.4% | 778 | 2169 | 2976 |
| 140 | China Mobile | JT-35B-Flash | -23.25 | 24.5% | 63.3% | 74% | 1471 | 2866 | 1559 |
| 141 | OpenAI | GPT-5.6 Terra (Non-reasoning) | -23.47 | 36.2% | 93.5% | 98.4% | 2172 | 3580 | 97 |
| 142 | Gemma 4 E2B | -23.98 | 6.7% | 32.9% | 38.4% | 401 | 1840 | 3694 | |
| 143 | DeepSeek | DeepSeek V3.1 Terminus | -24.32 | 27.9% | 72.5% | 83% | 1676 | 3135 | 1023 |
| 144 | OpenAI | GPT-5 nano (medium) | -24.88 | 16.1% | 48.8% | 63% | 966 | 2459 | 2217 |
| 145 | Deep Cogito | Cogito v2.1 | -24.98 | 30.2% | 79.1% | 87.6% | 1812 | 3311 | 744 |
| 146 | OpenAI | GPT-5.6 Luna (Non-reasoning) | -25.12 | 27.9% | 73.5% | 83.5% | 1673 | 3180 | 990 |
| 147 | Z AI | GLM-4.5 | -25.45 | 24.9% | 67.2% | 79.4% | 1497 | 3024 | 1235 |
| 148 | Mistral | Magistral Medium 1.2 | -26.23 | 20.8% | 59.5% | 69.9% | 1251 | 2825 | 1808 |
| 149 | Mistral | Magistral Medium 1 | -26.42 | 20.7% | 59.4% | 68.8% | 1243 | 2828 | 1871 |
| 150 | Kimi | Kimi K2 0905 | -26.47 | 25.4% | 69.6% | 79% | 1526 | 3114 | 1259 |
| 151 | Z AI | GLM-4.6V | -26.48 | 15.6% | 49.9% | 60.8% | 937 | 2526 | 2351 |
| 152 | DeepSeek | DeepSeek R1 0528 | -27.08 | 30.9% | 84% | 90.3% | 1857 | 3482 | 584 |
| 153 | Kimi | Kimi K2 | -27.52 | 26.8% | 74.2% | 84.6% | 1609 | 3260 | 922 |
| 154 | Nex AGI | Nex-N2-Pro | -27.62 | 34.4% | 94.5% | 98.4% | 2064 | 3721 | 98 |
| 155 | OpenAI | GPT-5 nano (high) | -27.75 | 18.3% | 56.3% | 68.1% | 1097 | 2762 | 1912 |
| 156 | Gemini 2.5 Flash | -27.85 | 25.1% | 70.7% | 83.7% | 1505 | 3176 | 980 | |
| 157 | DeepSeek | DeepSeek V3.1 | -28.37 | 28.8% | 80.3% | 89.1% | 1729 | 3431 | 653 |
| 158 | SpaceXAI | Grok 4 Fast | -28.4 | 22.6% | 66% | 76.4% | 1358 | 3062 | 1413 |
| 159 | SpaceXAI | Grok 4.1 Fast | -28.73 | 25.3% | 72.4% | 80.8% | 1520 | 3244 | 1155 |
| 160 | OpenAI | GPT-5.4 nano (xhigh) | -29.55 | 25.4% | 73.6% | 81.8% | 1522 | 3295 | 1093 |
| 161 | Alibaba | Qwen3.5 397B A17B | -29.78 | 31.4% | 89.1% | 93.8% | 1881 | 3668 | 374 |
| 162 | Mistral | Mistral Small 4 | -29.9 | 22.1% | 66.8% | 75.3% | 1327 | 3121 | 1484 |
| 163 | DeepSeek | DeepSeek V3.2 Exp | -29.93 | 28.2% | 81.1% | 87.3% | 1695 | 3491 | 765 |
| 164 | DeepSeek | DeepSeek V4 Pro | -30.08 | 30.8% | 87.9% | 93.8% | 1846 | 3651 | 371 |
| 165 | Meta | Llama 3.1 8B | -30.33 | 8.1% | 41.8% | 49.3% | 484 | 2304 | 3041 |
| 166 | Gemma 3 270M | -31.32 | 0.9% | 32.5% | 35% | 52 | 1931 | 3898 | |
| 167 | DeepSeek | DeepSeek R1 (Jan) | -31.32 | 30.7% | 89.5% | 94.6% | 1843 | 3722 | 326 |
| 168 | Mistral | Mistral Medium 3 | -31.47 | 18.3% | 60.9% | 69.7% | 1098 | 2986 | 1817 |
| 169 | Z AI | GLM-4.6 | -31.6 | 20.8% | 66.1% | 76.5% | 1247 | 3143 | 1412 |
| 170 | Mistral | Devstral Medium | -31.62 | 19.1% | 62.6% | 71% | 1143 | 3040 | 1738 |
| 171 | SpaceXAI | Grok 4.3 (Non-reasoning) | -32.3 | 24.2% | 74.5% | 81.7% | 1450 | 3388 | 1098 |
| 172 | SpaceXAI | Grok 3 | -32.33 | 28.3% | 84.6% | 93.5% | 1699 | 3639 | 392 |
| 173 | MiniMax | MiniMax-M2.1 | -32.83 | 20.5% | 67% | 77.2% | 1228 | 3198 | 1368 |
| 174 | Nous Research | Hermes 4 405B | -33 | 26.3% | 80.4% | 87.4% | 1576 | 3556 | 756 |
| 175 | Liquid AI | LFM2.5-8B-A1B | -33.27 | 9.3% | 47% | 52.6% | 561 | 2557 | 2845 |
| 176 | MBZUAI Institute of Foundation Models | K2 Think V2 | -33.92 | 15.7% | 58.9% | 66.8% | 943 | 2978 | 1991 |
| 177 | OpenAI | GPT-5 (minimal) | -33.95 | 28.1% | 86.2% | 94.6% | 1684 | 3721 | 325 |
| 178 | Mistral | Mistral Large 2 (Nov) | -34.05 | 20.1% | 67.8% | 75.6% | 1208 | 3251 | 1462 |
| 179 | Alibaba | Qwen3 Max Thinking | -34.42 | 29.9% | 91.8% | 97.4% | 1795 | 3860 | 155 |
| 180 | ByteDance Seed | Doubao Seed Code | -34.43 | 25.1% | 79.4% | 86% | 1504 | 3570 | 841 |
| 181 | Z AI | GLM-4.7 | -34.6 | 29.3% | 90.3% | 96.8% | 1756 | 3832 | 193 |
| 182 | Tencent | Hy3-preview | -34.63 | 28% | 86.9% | 92.4% | 1679 | 3757 | 455 |
| 183 | Gemini 2.5 Flash (Sep) | -34.72 | 28.9% | 89.5% | 93.9% | 1734 | 3817 | 368 | |
| 184 | OpenAI | GPT-5.1 | -34.95 | 29.1% | 90.2% | 95% | 1744 | 3841 | 299 |
| 185 | Alibaba | Qwen3.5 0.8B | -35.18 | 1.1% | 36.7% | 38.5% | 65 | 2176 | 3693 |
| 186 | OpenAI | o4-mini (high) | -35.75 | 24.6% | 80.1% | 86.2% | 1478 | 3623 | 830 |
| 187 | Baidu | ERNIE 4.5 300B A47B | -35.88 | 18.7% | 67.1% | 75.8% | 1121 | 3274 | 1449 |
| 188 | SpaceXAI | Grok Code Fast 1 | -36 | 23.8% | 78.5% | 85.2% | 1429 | 3589 | 891 |
| 189 | Nous Research | Hermes 4 405B | -36.05 | 30.2% | 94.8% | 98.1% | 1810 | 3973 | 114 |
| 190 | Alibaba | Qwen3.5 397B A17B | -36.08 | 24.3% | 79.8% | 88.6% | 1460 | 3625 | 683 |
| 191 | OpenAI | GPT-4.1 | -36.18 | 24.2% | 79.6% | 97.9% | 1451 | 3622 | 128 |
| 192 | Amazon | Nova Premier | -36.27 | 19.1% | 68.5% | 80.7% | 1147 | 3323 | 1158 |
| 193 | Mistral | Mistral Medium 3.5 | -36.32 | 25.1% | 82% | 87.4% | 1506 | 3685 | 757 |
| 194 | Tencent | Hy3-preview | -36.4 | 22.4% | 75.7% | 83.8% | 1341 | 3525 | 971 |
| 195 | KwaiKAT | KAT-Coder-Pro V1 | -37.43 | 17.5% | 66.6% | 74.4% | 1049 | 3295 | 1538 |
| 196 | StepFun | Step 3.7 Flash | -37.55 | 25.4% | 84.4% | 90.2% | 1524 | 3777 | 588 |
| 197 | DeepSeek | DeepSeek V3 0324 | -37.57 | 22.2% | 76.7% | 91.1% | 1330 | 3584 | 532 |
| 198 | InclusionAI | Ring-2.6-1T | -37.75 | 25.1% | 83.8% | 89.7% | 1504 | 3769 | 618 |
| 199 | Alibaba | Qwen3 Max Thinking (Preview) | -37.83 | 28% | 91.3% | 94.8% | 1677 | 3947 | 312 |
| 200 | Z AI | GLM-4.6V | -37.87 | 17.5% | 67.1% | 74.5% | 1050 | 3322 | 1533 |
| 201 | Gemini 2.5 Flash (Sep) | -38.12 | 26.7% | 88.3% | 94.9% | 1599 | 3886 | 307 | |
| 202 | Xiaomi | MiMo-V2.5-Pro | -38.3 | 27% | 89.3% | 94.3% | 1617 | 3915 | 345 |
| 203 | OpenAI | GPT-5.4 nano | -38.58 | 13.9% | 60.9% | 68.6% | 832 | 3147 | 1885 |
| 204 | Mistral | Mistral Large 3 | -39.43 | 24.1% | 83.7% | 91.8% | 1447 | 3813 | 493 |
| 205 | Alibaba | Qwen3.5 122B A10B | -39.58 | 24.7% | 85.5% | 91.4% | 1484 | 3859 | 516 |
| 206 | MiniMax | MiniMax-M2.5 | -39.7 | 26.2% | 89.3% | 92.8% | 1572 | 3954 | 429 |
| 207 | Amazon | Nova Lite | -40.42 | 9.7% | 55.6% | 62.6% | 584 | 3009 | 2242 |
| 208 | Alibaba | Qwen3 Max (Preview) | -40.58 | 24.3% | 85.7% | 92.6% | 1457 | 3892 | 445 |
| 209 | Meta | Llama 3.1 70B | -40.85 | 19.4% | 74.7% | 84.4% | 1163 | 3614 | 934 |
| 210 | StepFun | Step 3.5 Flash | -40.9 | 23.9% | 85.2% | 90.6% | 1434 | 3888 | 564 |
| 211 | NVIDIA | Llama 3.1 Nemotron 70B | -41.05 | 16.4% | 68.8% | 78.5% | 987 | 3450 | 1289 |
| 212 | DeepSeek | DeepSeek V3.1 | -41.07 | 23.1% | 83.5% | 90.8% | 1388 | 3852 | 553 |
| 213 | OpenAI | o3-mini (high) | -41.33 | 21% | 79% | 86.6% | 1262 | 3742 | 806 |
| 214 | DeepSeek | DeepSeek V3 (Dec) | -41.35 | 25.4% | 89.4% | 94% | 1522 | 4003 | 357 |
| 215 | Gemma 4 E4B | -41.52 | 8.3% | 54.4% | 58.9% | 499 | 2990 | 2467 | |
| 216 | Z AI | GLM-4.6 | -41.73 | 27.3% | 95% | 97.5% | 1639 | 4143 | 151 |
| 217 | Meta | Llama 4 Maverick | -41.77 | 24.3% | 87.3% | 93.7% | 1459 | 3965 | 380 |
| 218 | DeepSeek | DeepSeek V3.1 Terminus | -41.88 | 23.4% | 85.3% | 92.2% | 1407 | 3920 | 469 |
| 219 | Nanbeige | Nanbeige4.1-3B | -41.93 | 9.6% | 57% | 62.5% | 575 | 3091 | 2248 |
| 220 | Gemini 2.5 Flash | -42.02 | 26.5% | 93.3% | 96.6% | 1592 | 4113 | 204 | |
| 221 | Alibaba | Qwen3.5 27B | -42.02 | 21% | 79.7% | 86.3% | 1258 | 3779 | 824 |
| 222 | NVIDIA | NVIDIA Nemotron 3 Super | -42.07 | 24% | 87% | 91.4% | 1442 | 3966 | 516 |
| 223 | Alibaba | Qwen3 235B 2507 | -42.1 | 16.8% | 70.8% | 83.4% | 1007 | 3533 | 995 |
| 224 | Gemini 2.0 Flash | -42.55 | 23.6% | 86.5% | 91.3% | 1415 | 3968 | 519 | |
| 225 | Gemini 2.5 Flash-Lite (Sep) | -42.67 | 14.3% | 66.4% | 72% | 857 | 3417 | 1678 | |
| 226 | SpaceXAI | Grok 4.20 0309 | -42.68 | 26% | 92.9% | 96.5% | 1561 | 4122 | 210 |
| 227 | NVIDIA | NVIDIA Nemotron Nano 9B V2 | -42.87 | 11.3% | 61% | 66.2% | 676 | 3248 | 2026 |
| 228 | Alibaba | Qwen3 Max | -43.12 | 24.4% | 89.4% | 94% | 1466 | 4053 | 363 |
| 229 | Mistral | Magistral Small 1 | -43.37 | 18.3% | 75.5% | 81.3% | 1098 | 3700 | 1120 |
| 230 | Xiaomi | MiMo-V2-Flash | -43.45 | 24.9% | 91.1% | 96.4% | 1495 | 4102 | 215 |
| 231 | Z AI | GLM-4.5V | -43.57 | 20.6% | 80.9% | 90% | 1237 | 3851 | 599 |
| 232 | StepFun | Step 3.5 Flash 2603 | -43.67 | 25% | 91.6% | 94.8% | 1502 | 4122 | 309 |
| 233 | Allen Institute for AI | Olmo 3.1 32B Think | -44.12 | 13.6% | 66.8% | 72.9% | 816 | 3463 | 1629 |
| 234 | Baidu | ERNIE 5.0 Thinking Preview | -44.13 | 21.9% | 84.6% | 91.3% | 1315 | 3963 | 525 |
| 235 | DeepSeek | DeepSeek V4 Flash | -44.18 | 26.1% | 95.1% | 97.9% | 1566 | 4217 | 127 |
| 236 | Arcee AI | Trinity Large Thinking | -44.18 | 22.8% | 86.6% | 90.6% | 1365 | 4016 | 563 |
| 237 | Alibaba | Qwen3 235B | -44.58 | 18.4% | 77.2% | 83% | 1104 | 3779 | 1019 |
| 238 | SpaceXAI | Grok 4.20 0309 v2 | -44.6 | 26.6% | 96.9% | 99.5% | 1594 | 4270 | 33 |
| 239 | Alibaba | Qwen3 VL 235B A22B | -44.62 | 20.8% | 82.5% | 89.1% | 1246 | 3923 | 651 |
| 240 | Gemini 2.5 Flash-Lite | -44.63 | 17.5% | 75.4% | 82.8% | 1053 | 3731 | 1030 | |
| 241 | Gemma 4 31B | -45.42 | 19.9% | 81.6% | 89.7% | 1195 | 3920 | 618 | |
| 242 | Mistral | Mistral Medium 3.1 | -45.47 | 19.8% | 81.3% | 90% | 1186 | 3914 | 598 |
| 243 | NVIDIA | Llama Nemotron Ultra | -45.52 | 19.9% | 81.7% | 86.8% | 1194 | 3925 | 792 |
| 244 | Alibaba | Qwen3 235B A22B 2507 | -45.55 | 23.1% | 89.2% | 93.8% | 1385 | 4118 | 371 |
| 245 | NVIDIA | Llama Nemotron Super 49B v1.5 | -45.63 | 12.2% | 65.9% | 72.6% | 732 | 3470 | 1643 |
| 246 | Amazon | Nova 2.0 Pro Preview (low) | -45.77 | 22.1% | 87.2% | 91.5% | 1328 | 4074 | 509 |
| 247 | NVIDIA | Llama Nemotron Super 49B v1.5 | -45.82 | 16.9% | 75.5% | 82.3% | 1016 | 3765 | 1064 |
| 248 | OpenAI | GPT-5.4 mini | -45.85 | 25.3% | 95.3% | 98.1% | 1520 | 4271 | 113 |
| 249 | Mistral | Devstral 2 | -46.17 | 21% | 85.1% | 90% | 1261 | 4031 | 602 |
| 250 | Z AI | GLM-4.7 | -46.28 | 23.9% | 92.3% | 96.1% | 1435 | 4212 | 236 |
| 251 | Alibaba | Qwen3.5 35B A3B | -46.38 | 20.4% | 84% | 89.6% | 1227 | 4010 | 626 |
| 252 | DeepSeek | DeepSeek R1 Distill Llama 70B | -46.45 | 19.3% | 81.5% | 86.9% | 1159 | 3946 | 787 |
| 253 | DeepSeek | DeepSeek V3.2 | -46.73 | 24.2% | 93.5% | 96.1% | 1450 | 4254 | 232 |
| 254 | Alibaba | Qwen3.5 27B | -46.75 | 15.7% | 74.1% | 79.7% | 944 | 3749 | 1218 |
| 255 | MiniMax | MiniMax-M2 | -46.93 | 22% | 88.4% | 94.8% | 1321 | 4137 | 315 |
| 256 | DeepSeek | DeepSeek V3.2 Exp | -47.08 | 22.7% | 90.3% | 95% | 1362 | 4187 | 299 |
| 257 | Nous Research | Hermes 4 70B | -47.37 | 18.2% | 80.1% | 87.9% | 1091 | 3933 | 724 |
| 258 | MiniMax | MiniMax M1 80k | -47.42 | 21.1% | 86.8% | 95% | 1265 | 4110 | 302 |
| 259 | Alibaba | Qwen3.5 2B | -47.48 | 4.7% | 54.7% | 58.6% | 281 | 3130 | 2485 |
| 260 | Amazon | Nova Pro | -47.58 | 17% | 77.9% | 83.3% | 1022 | 3877 | 1005 |
| 261 | Anthropic | Claude 3 Haiku | -47.65 | 17.2% | 78.2% | 86.1% | 1030 | 3889 | 837 |
| 262 | InclusionAI | Ring-1T | -47.87 | 19.8% | 84.4% | 92.3% | 1188 | 4060 | 462 |
| 263 | Amazon | Nova 2.0 Pro Preview (medium) | -48.05 | 22% | 89.8% | 93.9% | 1319 | 4202 | 368 |
| 264 | Gemma 4 26B A4B | -48.07 | 18.1% | 80.9% | 90.5% | 1089 | 3973 | 570 | |
| 265 | MBZUAI Institute of Foundation Models | K2-V2 (low) | -48.07 | 16.1% | 76.4% | 81.4% | 965 | 3849 | 1116 |
| 266 | Amazon | Nova 2.0 Pro Preview | -48.15 | 16.1% | 76.5% | 84.5% | 965 | 3854 | 929 |
| 267 | Mistral | Mistral Small 4 | -48.37 | 16.3% | 77.2% | 83.3% | 975 | 3877 | 1003 |
| 268 | Cohere | Command A | -48.47 | 15.7% | 76.2% | 83.2% | 943 | 3851 | 1006 |
| 269 | Xiaomi | MiMo-V2-Flash | -48.48 | 15.2% | 75.1% | 81.5% | 913 | 3822 | 1111 |
| 270 | Amazon | Nova Micro | -48.67 | 9.9% | 65% | 69% | 593 | 3513 | 1857 |
| 271 | Cohere | North Mini Code | -49.07 | 18.6% | 83.1% | 87.5% | 1114 | 4058 | 750 |
| 272 | Alibaba | Qwen3 14B | -49.23 | 14.9% | 75.5% | 80.5% | 897 | 3851 | 1168 |
| 273 | NVIDIA | Llama 3.3 Nemotron Super 49B | -49.68 | 13.5% | 73% | 78.3% | 808 | 3789 | 1300 |
| 274 | SpaceXAI | Grok 4 Fast | -49.87 | 15.9% | 78.2% | 91.5% | 955 | 3947 | 509 |
| 275 | Nous Research | Hermes 4 70B | -49.92 | 23.3% | 95.4% | 97.9% | 1397 | 4392 | 124 |
| 276 | MBZUAI Institute of Foundation Models | K2-V2 (medium) | -49.97 | 17.4% | 81.5% | 86.3% | 1042 | 4040 | 824 |
| 277 | Amazon | Nova 2.0 Omni (low) | -49.98 | 18.5% | 84% | 88.8% | 1108 | 4107 | 674 |
| 278 | OpenAI | gpt-oss-120b (high) | -50.05 | 21.5% | 91.2% | 94.2% | 1291 | 4294 | 346 |
| 279 | OpenAI | GPT-4.1 mini | -50.13 | 17.5% | 82% | 96.2% | 1050 | 4058 | 228 |
| 280 | OpenAI | gpt-oss-120b (low) | -50.15 | 15.7% | 78.1% | 95% | 943 | 3952 | 300 |
| 281 | Alibaba | Qwen3 VL 32B | -50.2 | 16.8% | 80.4% | 87.5% | 1005 | 4017 | 747 |
| 282 | Mistral | Mistral Small 3.1 | -50.32 | 15% | 76.8% | 83% | 900 | 3919 | 1023 |
| 283 | Mistral | Mistral Small 3.2 | -50.42 | 14.7% | 76.3% | 82.2% | 880 | 3905 | 1067 |
| 284 | Alibaba | Qwen3 32B | -50.53 | 17.3% | 82% | 86.7% | 1036 | 4068 | 797 |
| 285 | Alibaba | Qwen3 Next 80B A3B | -50.57 | 17.4% | 82.2% | 91.6% | 1042 | 4076 | 506 |
| 286 | Alibaba | Qwen3 VL 235B A22B | -50.67 | 20.2% | 88.7% | 94.8% | 1210 | 4250 | 310 |
| 287 | Alibaba | Qwen3 Coder 30B A3B | -50.85 | 15.8% | 79.2% | 83.8% | 950 | 4001 | 974 |
| 288 | ByteDance Seed | Seed-OSS-36B-Instruct | -50.85 | 17.8% | 83.4% | 90.2% | 1065 | 4116 | 587 |
| 289 | SpaceXAI | Grok 4.1 Fast | -50.95 | 17% | 81.8% | 86.8% | 1018 | 4075 | 790 |
| 290 | InclusionAI | Ling-2.6-1T | -51 | 21.4% | 92.2% | 96.4% | 1287 | 4347 | 215 |
| 291 | Prime Intellect | INTELLECT-3 | -51.02 | 19.2% | 87% | 91.3% | 1154 | 4215 | 522 |
| 292 | Amazon | Nova 2.0 Lite (low) | -51.07 | 15.9% | 79.6% | 91% | 952 | 4016 | 542 |
| 293 | Allen Institute for AI | Olmo 3.1 32B Instruct | -51.08 | 11.3% | 70.4% | 77.7% | 680 | 3745 | 1336 |
| 294 | Mistral | Devstral Small | -51.15 | 14.7% | 77.1% | 82.2% | 880 | 3949 | 1071 |
| 295 | Alibaba | Qwen3 30B | -51.2 | 16.1% | 80.3% | 84.9% | 968 | 4040 | 905 |
| 296 | Gemma 4 31B | -51.35 | 16.9% | 82.1% | 86.9% | 1012 | 4093 | 786 | |
| 297 | LG AI Research | EXAONE 4.5 33B | -51.48 | 16.2% | 80.7% | 84.5% | 970 | 4059 | 933 |
| 298 | NVIDIA | NVIDIA Nemotron 3 Nano | -51.63 | 17.1% | 82.9% | 87.3% | 1027 | 4125 | 765 |
| 299 | Gemma 4 12B | -51.88 | 16% | 80.8% | 84.9% | 960 | 4073 | 907 | |
| 300 | Meta | Llama 3.3 70B | -51.92 | 17.9% | 85.1% | 94.2% | 1076 | 4191 | 346 |
| 301 | Alibaba | Qwen3 VL 8B | -52.05 | 20.2% | 90.4% | 94.6% | 1209 | 4332 | 324 |
| 302 | Liquid AI | LFM2 2.6B | -52.1 | 5.2% | 60.5% | 65.9% | 314 | 3440 | 2048 |
| 303 | Alibaba | Qwen2.5 72B | -52.22 | 17.6% | 84.7% | 89.5% | 1055 | 4188 | 629 |
| 304 | Inception | Mercury 2 | -52.28 | 20.5% | 91.5% | 94.6% | 1229 | 4366 | 324 |
| 305 | Meta | Llama 4 Scout | -52.37 | 14.5% | 78.3% | 84.3% | 873 | 4015 | 940 |
| 306 | Alibaba | Qwen3 VL 8B | -52.42 | 20.1% | 90.7% | 94.8% | 1205 | 4350 | 314 |
| 307 | NVIDIA | Nemotron Cascade 2 30B A3B | -52.47 | 17.6% | 85% | 89.2% | 1056 | 4204 | 646 |
| 308 | Alibaba | Qwen3.5 9B | -52.52 | 15.9% | 81.4% | 85.3% | 954 | 4105 | 881 |
| 309 | Gemma 4 12B (Non-reasoning) | -52.73 | 12% | 73.5% | 77.8% | 718 | 3882 | 1333 | |
| 310 | Alibaba | Qwen3.6 27B | -52.78 | 17% | 84% | 88% | 1019 | 4186 | 721 |
| 311 | Naver | HyperCLOVA X SEED Think (32B) | -52.87 | 14.6% | 79.1% | 85.6% | 878 | 4050 | 862 |
| 312 | Alibaba | Qwen3 235B | -52.97 | 18.2% | 87% | 91.4% | 1092 | 4270 | 518 |
| 313 | Gemini 2.5 Flash-Lite (Sep) | -53.08 | 18% | 86.8% | 91.1% | 1082 | 4267 | 533 | |
| 314 | Z AI | GLM-4.5V | -53.52 | 17.9% | 87% | 93.8% | 1074 | 4285 | 374 |
| 315 | Meta | Llama 3 70B | -53.58 | 17.6% | 86.5% | 91.9% | 1059 | 4274 | 487 |
| 316 | Upstage | Solar Pro 3 | -53.78 | 18.1% | 87.8% | 91.7% | 1087 | 4314 | 499 |
| 317 | Alibaba | Qwen3.5 122B A10B | -53.9 | 18.6% | 89.1% | 94.8% | 1116 | 4350 | 311 |
| 318 | Upstage | Solar Open 100B | -54.1 | 18.1% | 88.1% | 92.6% | 1085 | 4331 | 443 |
| 319 | Amazon | Nova 2.0 Lite (high) | -54.12 | 18.8% | 89.8% | 93.7% | 1129 | 4376 | 377 |
| 320 | OpenAI | GPT-5 mini (minimal) | -54.12 | 18.2% | 88.4% | 93.2% | 1091 | 4338 | 411 |
| 321 | Swiss AI Initiative | Apertus 70B Instruct | -55.1 | 13.2% | 78.6% | 85.2% | 790 | 4096 | 891 |
| 322 | Meta | Llama 3.2 1B | -55.22 | 6.9% | 66.8% | 69.8% | 416 | 3729 | 1810 |
| 323 | Trillion Labs | Tri-21B-think Preview | -55.28 | 8.7% | 70.1% | 75.4% | 524 | 3841 | 1475 |
| 324 | InclusionAI | Ling-1T | -55.48 | 19.9% | 94% | 97.5% | 1191 | 4520 | 151 |
| 325 | AI21 Labs | Jamba 1.7 Large | -55.55 | 19.7% | 93.8% | 97.8% | 1183 | 4516 | 130 |
| 326 | Amazon | Nova 2.0 Lite (medium) | -55.62 | 18.3% | 90.4% | 94.3% | 1097 | 4434 | 342 |
| 327 | ServiceNow | Apriel-v1.5-15B-Thinker | -55.78 | 16.1% | 85.7% | 89.4% | 966 | 4313 | 635 |
| 328 | NVIDIA | Nemotron 3 Nano Omni 30B A3B Reasoning | -56.03 | 14.8% | 83.1% | 86.5% | 888 | 4250 | 809 |
| 329 | OpenAI | GPT-4.1 nano | -56.42 | 13.3% | 80.4% | 86.5% | 798 | 4183 | 810 |
| 330 | Multiverse Computing | HyperNova 60B 2605 | -56.43 | 14.1% | 82.2% | 86.4% | 849 | 4235 | 819 |
| 331 | Upstage | Solar Pro 2 | -56.57 | 19.5% | 94.5% | 97.3% | 1169 | 4563 | 162 |
| 332 | MBZUAI Institute of Foundation Models | K2-V2 (high) | -56.7 | 18.3% | 91.8% | 94.5% | 1097 | 4499 | 332 |
| 333 | Microsoft | Phi-4 | -56.72 | 13.2% | 80.5% | 86.1% | 791 | 4194 | 832 |
| 334 | Mistral | Devstral Small 2 | -56.73 | 15.3% | 85.1% | 91.2% | 920 | 4324 | 528 |
| 335 | Alibaba | Qwen3.5 4B | -56.77 | 12.8% | 79.7% | 84.4% | 767 | 4173 | 934 |
| 336 | Alibaba | Qwen3 30B A3B 2507 | -56.87 | 16.2% | 87.3% | 90.4% | 974 | 4386 | 576 |
| 337 | Mistral | Devstral Small (May) | -56.98 | 15.8% | 86.6% | 90.1% | 951 | 4370 | 593 |
| 338 | Gemini 2.5 Flash-Lite | -57 | 15.3% | 85.4% | 91.1% | 919 | 4339 | 536 | |
| 339 | NVIDIA | NVIDIA Nemotron Nano 9B V2 | -57.1 | 9.5% | 73.7% | 78.6% | 572 | 3998 | 1286 |
| 340 | Korea Telecom | Mi:dm K 2.5 Pro | -57.27 | 16.8% | 88.9% | 95.9% | 1005 | 4441 | 249 |
| 341 | Alibaba | Qwen3 VL 30B A3B | -57.32 | 16.9% | 89.2% | 92.8% | 1011 | 4450 | 435 |
| 342 | Amazon | Nova 2.0 Omni (medium) | -57.83 | 17.9% | 92.2% | 95.3% | 1073 | 4543 | 280 |
| 343 | LG AI Research | K-EXAONE | -57.93 | 16.5% | 89.1% | 91.7% | 989 | 4465 | 497 |
| 344 | NVIDIA | Llama 3.3 Nemotron Super 49B | -58.33 | 16.9% | 90.5% | 93.9% | 1014 | 4514 | 367 |
| 345 | Allen Institute for AI | Olmo 3 32B Think | -58.4 | 14.4% | 85.2% | 89.8% | 867 | 4371 | 610 |
| 346 | ServiceNow | Apriel-v1.6-15B-Thinker | -58.53 | 17% | 90.9% | 94.3% | 1017 | 4529 | 345 |
| 347 | InclusionAI | Ring-flash-2.0 | -58.67 | 16.3% | 89.6% | 92.7% | 978 | 4498 | 437 |
| 348 | StepFun | Step3 VL 10B | -59 | 12.7% | 82.2% | 85.9% | 763 | 4303 | 846 |
| 349 | DiffusionGemma 26B A4B | -59.03 | 16.8% | 91.1% | 93.8% | 1008 | 4550 | 373 | |
| 350 | Liquid AI | LFM2 24B A2B | -59.07 | 6.4% | 70% | 72.4% | 385 | 3929 | 1655 |
| 351 | Amazon | Nova 2.0 Lite | -59.1 | 13.5% | 83.9% | 89.6% | 810 | 4356 | 626 |
| 352 | Alibaba | Qwen3 Next 80B A3B | -59.1 | 17.6% | 93.2% | 95.4% | 1058 | 4604 | 275 |
| 353 | Z AI | GLM-4.7-Flash | -59.33 | 15.9% | 89.5% | 94.5% | 954 | 4514 | 328 |
| 354 | Sarvam | Sarvam 105B (high) | -59.53 | 17.6% | 93.5% | 96.1% | 1054 | 4626 | 234 |
| 355 | Alibaba | Qwen3 4B 2507 | -59.63 | 10.3% | 78% | 84.3% | 619 | 4197 | 943 |
| 356 | Alibaba | Qwen3.6 35B A3B | -59.83 | 17.1% | 92.7% | 95.3% | 1024 | 4614 | 285 |
| 357 | OpenAI | gpt-oss-20b (low) | -60.13 | 14.2% | 86.6% | 89.8% | 851 | 4459 | 609 |
| 358 | Alibaba | Qwen3 Coder Next | -60.72 | 15.8% | 90.9% | 95.7% | 949 | 4592 | 259 |
| 359 | Alibaba | Qwen3 Omni 30B A3B | -60.72 | 14.8% | 88.7% | 91.5% | 889 | 4532 | 509 |
| 360 | LG AI Research | EXAONE 4.0 32B | -60.8 | 13.6% | 86.1% | 89.7% | 816 | 4464 | 617 |
| 361 | Alibaba | Qwen3 4B 2507 | -60.82 | 12.7% | 84.1% | 86.8% | 759 | 4408 | 793 |
| 362 | IBM | Granite 4.0 H Small | -60.92 | 14.4% | 87.9% | 90.7% | 862 | 4517 | 558 |
| 363 | Microsoft | Phi-4 Mini | -61.02 | 8.3% | 75.5% | 81.8% | 496 | 4157 | 1093 |
| 364 | LG AI Research | K-EXAONE | -61.17 | 12.2% | 83.5% | 93.9% | 731 | 4401 | 364 |
| 365 | Motif Technologies | Motif-2-12.7B | -61.18 | 15.3% | 90.4% | 93% | 921 | 4592 | 422 |
| 366 | Gemma 4 E2B | -61.37 | 8% | 75.4% | 78.1% | 479 | 4161 | 1311 | |
| 367 | Upstage | Solar Pro 2 | -61.68 | 15.6% | 91.5% | 96.4% | 934 | 4635 | 218 |
| 368 | Alibaba | Qwen3.5 35B A3B | -62 | 15.7% | 92.1% | 96.2% | 940 | 4660 | 231 |
| 369 | Korea Telecom | Mi:dm K 2.5 Pro Preview | -62.02 | 16.1% | 93.1% | 96.3% | 965 | 4686 | 222 |
| 370 | Gemma 4 26B A4B | -62.03 | 15.7% | 92.2% | 94.8% | 943 | 4665 | 310 | |
| 371 | TII UAE | Falcon-H1R-7B | -62.08 | 14.1% | 88.7% | 92% | 845 | 4570 | 481 |
| 372 | Z AI | GLM-4.7-Flash | -62.23 | 12% | 84.3% | 96.9% | 719 | 4453 | 186 |
| 373 | Sarvam | Sarvam M | -62.25 | 14.8% | 90.5% | 93.2% | 890 | 4625 | 407 |
| 374 | LG AI Research | EXAONE 4.0 32B | -62.3 | 10.3% | 81% | 86% | 621 | 4359 | 842 |
| 375 | Z AI | GLM-4.5-Air | -62.52 | 15.5% | 92.3% | 95.5% | 930 | 4681 | 271 |
| 376 | Alibaba | Qwen3 VL 32B | -62.82 | 14.9% | 91.2% | 93.9% | 892 | 4661 | 365 |
| 377 | Alibaba | Qwen3 VL 30B A3B | -63 | 15.5% | 92.9% | 95.7% | 931 | 4711 | 259 |
| 378 | Meta | Llama 3.2 11B (Vision) | -63.08 | 10.2% | 81.6% | 85.6% | 612 | 4397 | 866 |
| 379 | Trillion Labs | Tri-21B-Think | -63.3 | 12% | 85.5% | 87.9% | 718 | 4516 | 724 |
| 380 | InclusionAI | Ling-flash-2.0 | -63.37 | 14% | 90% | 93.2% | 840 | 4642 | 408 |
| 381 | OpenAI | gpt-oss-20b (high) | -63.92 | 15.5% | 94.1% | 96.4% | 932 | 4767 | 215 |
| 382 | OpenAI | GPT-5 nano (minimal) | -63.98 | 11% | 84.3% | 91.7% | 662 | 4501 | 500 |
| 383 | Mistral | Ministral 3 3B | -64.1 | 7.6% | 77.5% | 82.5% | 454 | 4300 | 1051 |
| 384 | DeepSeek | DeepSeek R1 0528 Qwen3 8B | -64.12 | 12% | 86.5% | 89.4% | 720 | 4567 | 637 |
| 385 | NVIDIA | NVIDIA Nemotron Nano 12B v2 VL | -64.42 | 13.8% | 90.7% | 94.6% | 826 | 4691 | 322 |
| 386 | Amazon | Nova 2.0 Omni | -64.5 | 12.9% | 88.8% | 91.8% | 772 | 4642 | 489 |
| 387 | China Mobile | JT-MINI | -64.58 | 13.9% | 91.2% | 93.8% | 836 | 4711 | 370 |
| 388 | IBM | Granite 4.1 8B | -64.6 | 12.1% | 87.3% | 90.2% | 726 | 4602 | 588 |
| 389 | Reka AI | Reka Flash 3 | -64.7 | 13.4% | 90.1% | 93.2% | 802 | 4684 | 407 |
| 390 | Mistral | Mistral 7B | -64.77 | 7.9% | 79% | 84.6% | 476 | 4362 | 922 |
| 391 | Mistral | Magistral Small 1.2 | -64.98 | 13.5% | 90.7% | 93.7% | 809 | 4708 | 376 |
| 392 | Alibaba | Qwen3 8B | -65.07 | 12.8% | 89.3% | 93.5% | 767 | 4671 | 392 |
| 393 | Alibaba | Qwen3 30B | -65.07 | 11.7% | 86.8% | 92.5% | 699 | 4603 | 451 |
| 394 | Alibaba | Qwen3.5 Omni Flash | -65.63 | 14.4% | 93.6% | 95.9% | 867 | 4805 | 248 |
| 395 | InclusionAI | Ling 2.6 Flash | -65.68 | 15.4% | 95.8% | 99.1% | 923 | 4864 | 52 |
| 396 | Gemma 3 27B | -65.88 | 12.5% | 89.5% | 95% | 749 | 4702 | 303 | |
| 397 | Alibaba | Qwen3 30B A3B 2507 | -65.92 | 15.1% | 95.4% | 97.3% | 906 | 4861 | 164 |
| 398 | Mistral | Ministral 3 14B | -66.82 | 12.3% | 90.2% | 95.5% | 738 | 4747 | 273 |
| 399 | Alibaba | Qwen3 14B | -67.38 | 12.7% | 91.8% | 95% | 764 | 4807 | 302 |
| 400 | IBM | Granite 4.1 30B | -67.78 | 14% | 95.1% | 96.9% | 839 | 4906 | 184 |
| 401 | Mistral | Ministral 3 8B | -67.83 | 11.2% | 89% | 96.8% | 671 | 4741 | 194 |
| 402 | Allen Institute for AI | Molmo2-8B | -69.12 | 11.3% | 90.7% | 93.2% | 679 | 4826 | 410 |
| 403 | Alibaba | Qwen3 Omni 30B A3B | -69.18 | 14.4% | 97.7% | 99.2% | 865 | 5016 | 50 |
| 404 | NVIDIA | NVIDIA Nemotron 3 Nano | -69.22 | 11.4% | 90.9% | 93.6% | 682 | 4835 | 385 |
| 405 | Alibaba | Qwen3 VL 4B | -69.35 | 12.2% | 92.8% | 94.3% | 730 | 4891 | 344 |
| 406 | Meta | Llama 3 8B | -69.55 | 9.8% | 88% | 92.6% | 588 | 4761 | 443 |
| 407 | LongCat | LongCat Flash Lite | -70.1 | 13.1% | 95.7% | 98% | 786 | 4992 | 122 |
| 408 | Alibaba | Qwen3.5 9B | -70.75 | 14.1% | 98.8% | 99.7% | 846 | 5091 | 19 |
| 409 | NVIDIA | NVIDIA Nemotron 3 Nano 4B | -71.53 | 9.4% | 89.4% | 91% | 567 | 4859 | 539 |
| 410 | Liquid AI | LFM2 1.2B | -71.62 | 4.3% | 79.3% | 81.8% | 257 | 4554 | 1095 |
| 411 | Sarvam | Sarvam 30B (high) | -72.05 | 12.7% | 97% | 98.2% | 760 | 5083 | 106 |
| 412 | IBM | Granite 4.0 350M | -72.15 | 3.2% | 77.8% | 82.1% | 190 | 4519 | 1072 |
| 413 | NVIDIA | NVIDIA Nemotron Nano 12B v2 VL | -72.75 | 11.4% | 94.9% | 97.4% | 682 | 5047 | 154 |
| 414 | IBM | Granite 4.0 H 1B | -73.65 | 5.3% | 83.4% | 85.1% | 320 | 4739 | 895 |
| 415 | Allen Institute for AI | Olmo 3 7B Think | -73.68 | 10.8% | 94.7% | 95.7% | 649 | 5070 | 257 |
| 416 | AI21 Labs | Jamba Reasoning 3B | -73.68 | 6.6% | 86% | 88.1% | 396 | 4817 | 713 |
| 417 | Liquid AI | LFM2.5-1.2B-Instruct | -73.78 | 6% | 84.8% | 89% | 359 | 4786 | 658 |
| 418 | Alibaba | Qwen3 8B | -74.1 | 10.3% | 94.2% | 97.7% | 621 | 5067 | 138 |
| 419 | AI21 Labs | Jamba 1.7 Mini | -74.25 | 11.4% | 96.7% | 98% | 684 | 5139 | 119 |
| 420 | Swiss AI Initiative | Apertus 8B Instruct | -74.72 | 10.3% | 94.7% | 97.2% | 615 | 5098 | 171 |
| 421 | Alibaba | Qwen3.5 4B | -74.72 | 11.4% | 97.2% | 98.2% | 685 | 5168 | 108 |
| 422 | Liquid AI | LFM2 8B A1B | -75.13 | 6.8% | 88% | 94.2% | 410 | 4918 | 347 |
| 423 | Alibaba | Qwen3 VL 4B | -75.82 | 10.3% | 96.1% | 98.5% | 621 | 5170 | 88 |
| 424 | IBM | Granite 3.3 8B | -76.62 | 8.5% | 93.1% | 98.5% | 512 | 5109 | 91 |
| 425 | Gemma 3 12B | -76.98 | 10.2% | 97.1% | 99.2% | 613 | 5232 | 48 | |
| 426 | Alibaba | Qwen3 1.7B | -77.2 | 8.4% | 93.4% | 96% | 504 | 5136 | 238 |
| 427 | Gemma 3 1B | -77.27 | 3.6% | 83.9% | 86.5% | 215 | 4851 | 811 | |
| 428 | IBM | Granite 4.1 3B | -77.38 | 9.3% | 95.5% | 96.5% | 555 | 5198 | 207 |
| 429 | Allen Institute for AI | Olmo 3 7B | -77.75 | 7.3% | 91.8% | 92.7% | 440 | 5105 | 436 |
| 430 | IBM | Granite 4.0 Micro | -77.92 | 9% | 95.5% | 97.1% | 541 | 5216 | 172 |
| 431 | InclusionAI | Ling-mini-2.0 | -78.77 | 8.7% | 95.8% | 97.5% | 522 | 5248 | 151 |
| 432 | Gemma 3n E2B | -79.58 | 6.7% | 92.4% | 95.3% | 399 | 5174 | 279 | |
| 433 | Gemma 3n E4B | -80.4 | 7.8% | 95.6% | 98.7% | 465 | 5289 | 79 | |
| 434 | Alibaba | Qwen3 0.6B | -81.78 | 5.5% | 92.4% | 94.5% | 331 | 5238 | 330 |
| 435 | IBM | Granite 4.0 1B | -81.82 | 6% | 93.5% | 95.7% | 363 | 5272 | 257 |
| 436 | LG AI Research | Exaone 4.0 1.2B | -82.43 | 6.2% | 94.5% | 96.2% | 371 | 5317 | 231 |
| 437 | LG AI Research | Exaone 4.0 1.2B | -82.58 | 4.7% | 91.5% | 93.3% | 279 | 5234 | 404 |
| 438 | Alibaba | Qwen3.5 2B | -82.83 | 7.4% | 97.4% | 98.6% | 444 | 5414 | 86 |
| 439 | Gemma 3 4B | -83.32 | 7.8% | 98.8% | 99.7% | 468 | 5467 | 20 | |
| 440 | Alibaba | Qwen3 1.7B | -83.47 | 6.6% | 96.4% | 98.8% | 396 | 5404 | 73 |
| 441 | Liquid AI | LFM2.5-VL-1.6B | -83.87 | 5.2% | 94% | 97.5% | 312 | 5344 | 149 |
| 442 | Liquid AI | LFM2.5-1.2B-Thinking | -83.87 | 6.6% | 96.9% | 98% | 397 | 5429 | 121 |
| 443 | OpenBMB | MiniCPM-V 4.6 1.3B | -84.67 | 6.2% | 96.8% | 98.3% | 371 | 5451 | 100 |
| 444 | Cohere | Tiny Aya Global | -84.72 | 5.7% | 95.8% | 97.6% | 339 | 5422 | 146 |
| 445 | Alibaba | Qwen3 0.6B | -85.53 | 4.2% | 93.7% | 95.6% | 253 | 5385 | 262 |
| 446 | IBM | Granite 4.0 H 350M | -87.25 | 3.7% | 94.4% | 94.7% | 220 | 5455 | 319 |
| 447 | Alibaba | Qwen3.5 0.8B | -89 | 4.8% | 98.6% | 99.5% | 290 | 5630 | 29 |
同时观察回答正确性与模型是否知道自己不知道。
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。
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”。