AI 大模型能力排行
基于 Artificial Analysis 权威数据，每日更新
最后更新：2026/8/25 00:28:00
智能指数
综合智能评分，数值越高越好
响应速度
每秒输出 Token 数，数值越高越好
使用价格
每百万 Token 价格（美元），越低越好
首 Token 延迟
首个 Token 响应时间（秒），越低越好
大模型排名
瓦勒杉 - 综合智能指数排名
基于多项基准测试的加权评估，数值越高越好
AI 资讯
模型详细数据
| 模型名称 | 提供商 | 智能指数 | 编程指数 | 速度 | 首Token延迟 | 价格 |
|---|---|---|---|---|---|---|
 Claude Opus 5 (Adaptive Reasoning, Max Effort) | Anthropic | 63.1 | 78.0 | 52 t/s | 34.6s | $10 |
| Claude Opus 5 (Adaptive Reasoning, Xhigh Effort) | Anthropic | 62.5 | 77.0 | 50 t/s | 18.3s | $10 |
| Claude Fable 5 (Adaptive Reasoning, Max Effort, Opus 4.8 Fallback) | Anthropic | 62.1 | 76.5 | 63 t/s | 43.7s | $20 |
| Claude Opus 5 (Adaptive Reasoning, High Effort) | Anthropic | 61.5 | 76.5 | 47 t/s | 9.4s | $10 |
 GPT-5.6 Sol (max) | OpenAI | 60.9 | 77.4 | 64 t/s | 74.8s | $11 |
%20{%20.cls-1%20{%20fill:%20%23fff;%20}%20polygon%20{%20fill:%20%23000;%20}%20}%20%3c/style%3e%3c/defs%3e%3crect%20class='cls-1'%20width='1000'%20height='1000'/%3e%3cg%3e%3cpolygon%20points='226.83%20411.15%20501.31%20803.15%20623.31%20803.15%20348.82%20411.15%20226.83%20411.15'%20/%3e%3cpolygon%20points='348.72%20628.87%20226.69%20803.15%20348.77%20803.15%20409.76%20716.05%20348.72%20628.87'%20/%3e%3cpolygon%20points='651.23%20196.85%20440.28%20498.12%20501.32%20585.29%20773.31%20196.85%20651.23%20196.85'%20/%3e%3cpolygon%20points='673.31%20383.25%20673.31%20803.15%20773.31%20803.15%20773.31%20240.44%20673.31%20383.25'%20/%3e%3c/g%3e%3c/svg%3e) Grok 4.6 (high) | SpaceXAI | 60.9 | 76.8 | 56 t/s | 31.4s | $3.0 |
| Grok 4.6 (xhigh) | SpaceXAI | 60.0 | 75.9 | 56 t/s | 36.1s | $3.0 |
 Kimi K3 (max) | Kimi | 59.7 | 76.2 | 34 t/s | 2.7s | $6.0 |
 GLM-5.3 (max) | Z AI | 59.5 | 74.8 | - | - | $2.1 |
| GPT-5.6 Sol (xhigh) | OpenAI | 59.0 | 78.3 | 64 t/s | 28.5s | $11 |
| Grok 4.6 (medium) | SpaceXAI | 59.0 | 74.4 | 58 t/s | 20.9s | $3.0 |
| Claude Opus 5 (Adaptive Reasoning, Medium Effort) | Anthropic | 58.6 | 74.3 | 46 t/s | 4.1s | $10 |
 Qwen3.8 Max | Alibaba | 58.1 | 71.8 | 43 t/s | 1.7s | $3.0 |
| Qwen3.8 2.4T A95B | Alibaba | 57.7 | 71.9 | 45 t/s | 1.8s | $3.0 |
| GPT-5.6 Sol (high) | OpenAI | 57.3 | 77.2 | 57 t/s | 12.1s | $11 |
| Claude Opus 4.8 (Adaptive Reasoning, Max Effort) | Anthropic | 57.3 | 74.3 | - | - | $10 |
 Muse Spark 1.2 (xhigh) | Meta | 56.8 | 72.2 | - | - | $2.0 |
| GPT-5.6 Terra (max) | OpenAI | 56.6 | 76.7 | 111 t/s | 60.5s | $4.5 |
| GPT-5.5 (xhigh) | OpenAI | 56.3 | 74.9 | - | - | $11 |
 Gemini 3.7 Flash (high) | 56.0 | 76.1 | 327 t/s | 6.7s | $1.5 | |
| Grok 4.5 (high) | SpaceXAI | 55.8 | 72.4 | 53 t/s | 10.3s | $3.0 |
| GPT-5.6 Sol (medium) | OpenAI | 55.6 | 76.3 | 53 t/s | 5.5s | $11 |
| Claude Sonnet 5 (Adaptive Reasoning, Max Effort) | Anthropic | 55.3 | 71.5 | 77 t/s | 116.8s | $4.0 |
| Claude Opus 4.7 (Adaptive Reasoning, Max Effort) | Anthropic | 55.0 | 73.6 | - | - | $10 |
| GPT-5.5 (high) | OpenAI | 54.7 | 71.6 | - | - | $11 |
| Gemini 3.7 Flash (medium) | 53.4 | 71.5 | 301 t/s | 5.0s | $1.5 | |
 DeepSeek V4 Pro 0813 (Reasoning, Max Effort) | DeepSeek | 53.2 | 68.8 | 73 t/s | 1.4s | $2.0 |
| Muse Spark 1.1 (xhigh) | Meta | 53.2 | 71.3 | - | - | $2.0 |
| GPT-5.4 (xhigh) | OpenAI | 53.1 | 71.1 | - | - | $5.6 |
| GPT-5.6 Terra (xhigh) | OpenAI | 52.8 | 70.6 | 95 t/s | 9.7s | $4.5 |
| GLM-5.2 (max) | Z AI | 52.6 | 68.8 | - | - | $2.1 |
| Claude Opus 5 (Adaptive Reasoning, Low Effort) | Anthropic | 52.5 | 66.9 | 45 t/s | 2.9s | $10 |
| GPT-5.6 Luna (max) | OpenAI | 52.3 | 71.4 | 139 t/s | 81.1s | $0.5 |
| Qwen3.8 27B (xhigh) | Alibaba | 52.0 | 68.1 | 53 t/s | 1.1s | $1.1 |
| Gemini 3.5 Flash (high) | 52.0 | 70.1 | - | - | $3.4 | |
| DeepSeek V4 Flash 0731 (Reasoning, Max Effort) | DeepSeek | 51.8 | 69.1 | 109 t/s | 0.9s | $0.7 |
| Grok 4.6 (low) | SpaceXAI | 51.7 | 66.3 | 50 t/s | 4.8s | $3.0 |
| Gemini 3.6 Flash (high) | 51.6 | 69.2 | 196 t/s | 12.9s | $1.5 | |
| GPT-5.5 (medium) | OpenAI | 51.4 | 71.5 | - | - | $11 |
| Gemini 3.7 Flash (low) | 50.9 | 71.0 | 304 t/s | 0.9s | $1.5 | |
| GPT-5.6 Sol (low) | OpenAI | 50.7 | 69.7 | 58 t/s | 3.0s | $11 |
| GPT-5.6 Terra (high) | OpenAI | 50.1 | 67.1 | 93 t/s | 2.7s | $4.5 |
| GPT-5.6 Luna (xhigh) | OpenAI | 50.1 | 68.6 | 130 t/s | 20.3s | $0.5 |
| Claude Sonnet 4.6 (Adaptive Reasoning, Max Effort) | Anthropic | 48.4 | 63.0 | - | - | $6.0 |
| Kimi K3 (low) | Kimi | 48.3 | 72.0 | 33 t/s | 2.9s | $6.0 |
| Gemini 3.1 Pro Preview | 47.7 | 68.8 | 123 t/s | 23.0s | $4.5 | |
| Motif 3 | Motif Technologies | 47.4 | 63.5 | - | - | - |
| GPT-5.6 Luna (high) | OpenAI | 47.0 | 63.3 | 125 t/s | 7.1s | $0.5 |
| GPT-5.6 Terra (medium) | OpenAI | 46.8 | 64.7 | 86 t/s | 1.9s | $4.5 |
| Gemini 3.5 Flash (medium) | 46.7 | - | 281 t/s | 9.4s | $3.4 | |
| Qwen3.7 Max | Alibaba | 46.7 | 66.0 | - | - | $3.8 |
| GPT-5.3 Codex (xhigh) | OpenAI | 45.5 | - | 128 t/s | 34.4s | $4.8 |
 MiniMax-M3 | MiniMax | 45.4 | 58.6 | 133 t/s | 0.9s | $0.5 |
| DeepSeek V4 Pro (Reasoning, Max Effort) | DeepSeek | 45.3 | 59.4 | 70 t/s | 1.2s | $0.5 |
| Motif 3 (Beta) | Motif Technologies | 45.3 | 62.0 | - | - | - |
| Kimi K2.6 | Kimi | 45.1 | 61.8 | - | - | $1.7 |
| Claude Opus 4.6 (Adaptive Reasoning, Max Effort) | Anthropic | 44.9 | - | - | - | $10 |
| Qwen3.8 27B (medium) | Alibaba | 44.5 | 56.1 | 58 t/s | 1.2s | $1.1 |
| GPT-5.5 (low) | OpenAI | 44.5 | 60.9 | - | - | $11 |
| Muse Spark | Meta | 44.3 | 58.6 | - | - | - |
| Claude Opus 4.7 (Non-reasoning, High Effort) | Anthropic | 43.9 | - | - | - | $10 |
| DeepSeek V4 Pro (Reasoning, High Effort) | DeepSeek | 43.7 | 58.7 | 69 t/s | 1.3s | $0.5 |
| GPT-5.2 (xhigh) | OpenAI | 43.3 | - | - | - | $4.8 |
| Kimi K2.7 Code | Kimi | 43.0 | 60.8 | 43 t/s | 1.1s | $1.7 |
 MiMo-V2.5-Pro | Xiaomi | 42.9 | 60.2 | 44 t/s | 3.1s | $0.5 |
| Qwen3.8 27B (low) | Alibaba | 42.9 | 58.2 | 60 t/s | 1.1s | $1.1 |
| Claude Sonnet 5 (Non-reasoning, High Effort) | Anthropic | 42.6 | 66.4 | 58 t/s | 1.3s | $4.0 |
| Inkling (xhigh) | Thinking Machines | 42.3 | 52.1 | 61 t/s | 2.5s | $1.8 |
| Hy3 | Tencent | 42.2 | 58.8 | 74 t/s | 1.8s | $0.2 |
| DeepSeek V4 Flash (Reasoning, Max Effort) | DeepSeek | 42.1 | 56.2 | - | - | $0.2 |
| GPT-5.6 Sol (Non-reasoning) | OpenAI | 41.9 | 65.1 | 55 t/s | 1.3s | $11 |
| Claude Opus 4.5 (Reasoning) | Anthropic | 41.9 | - | - | - | $10 |
| Nex-N2-Pro | Nex AGI | 41.7 | 59.1 | 136 t/s | 0.9s | $1.0 |
'%3e%3c/path%3e%3cdefs%3e%3clinearGradient%20gradientUnits='userSpaceOnUse'%20id='lobe-icons-upsate-fill'%20x1='11.927'%20x2='11.927'%20y2='24'%3e%3cstop%20offset='0'%20stop-color='%23AEBCFE'%3e%3c/stop%3e%3cstop%20offset='1'%20stop-color='%23805DFA'%3e%3c/stop%3e%3c/linearGradient%3e%3c/defs%3e%3c/svg%3e) Solar Pro 4 | Upstage | 41.6 | 52.7 | 45 t/s | 1.3s | $0.5 |
| MiMo-V2-Pro | Xiaomi | 41.4 | - | - | - | - |
| GPT-5.6 Terra (low) | OpenAI | 41.3 | 58.1 | 88 t/s | 1.7s | $4.5 |
| Inkling Small | Thinking Machines | 41.2 | 52.9 | 135 t/s | 1.7s | $0.5 |
| GPT-5.2 Codex (xhigh) | OpenAI | 41.2 | - | - | - | $4.8 |
| Qwen3.6 Max Preview | Alibaba | 41.1 | - | - | - | $2.9 |
| GLM-5.1 (Reasoning) | Z AI | 41.0 | 55.8 | - | - | $2.1 |
| GPT-5.4 mini (xhigh) | OpenAI | 40.9 | 56.1 | - | - | $1.7 |
| Grok Build 0.1 0616 | SpaceXAI | 40.7 | 51.5 | - | - | $1.3 |
| Gemini 3 Pro Preview (high) | 40.6 | - | - | - | $4.5 | |
| GLM-5 (Reasoning) | Z AI | 40.6 | - | - | - | $1.6 |
| Qwen3.6 Plus | Alibaba | 40.5 | 54.5 | - | - | $1.1 |
| GPT-5.4 (low) | OpenAI | 40.2 | - | - | - | $5.6 |
| JT-4.1 Flash 236B A21B | China Mobile | 39.9 | 52.4 | - | - | - |
| Agnes 2.5 Pro Alpha | Sapiens AI | 39.7 | 58.8 | 132 t/s | 1.4s | $0.6 |
| GPT-5.4 nano (xhigh) | OpenAI | 39.7 | 56.1 | - | - | $0.5 |
| Qwen3.7 Plus | Alibaba | 39.4 | 55.9 | 57 t/s | 1.3s | $0.7 |
| GLM-5-Turbo | Z AI | 39.1 | - | - | - | - |
| DeepSeek V4 Flash (Reasoning, High Effort) | DeepSeek | 39.0 | 52.0 | - | - | $0.2 |
| GPT-5.6 Luna (medium) | OpenAI | 38.9 | 50.7 | 125 t/s | 2.3s | $0.5 |
| GPT-5.2 (medium) | OpenAI | 38.9 | - | - | - | $4.8 |
| MiniMax-M2.7 | MiniMax | 38.9 | 52.6 | - | - | $0.5 |
| Claude Opus 4.6 (Non-reasoning, High Effort) | Anthropic | 38.8 | - | - | - | $10 |
| Gemini 3 Flash Preview (Reasoning) | 38.7 | - | - | - | $1.1 | |
 Nemotron 3 Ultra 550B A55B (Reasoning) | NVIDIA | 38.3 | 49.3 | 169 t/s | 1.0s | $1.1 |
| MiMo-V2.5 | Xiaomi | 38.0 | 56.8 | 57 t/s | 3.7s | $0.2 |
| Grok 4.20 0309 v2 (Reasoning) | SpaceXAI | 38.0 | - | - | - | $1.6 |
| Grok 4.3 (high) | SpaceXAI | 37.9 | 42.2 | - | - | $1.6 |
 Ling 3.0 Flash | InclusionAI | 37.8 | 50.6 | 407 t/s | 1.6s | $0.1 |
| Qwen3.6 27B (Reasoning) | Alibaba | 37.7 | 53.7 | 58 t/s | 1.4s | $1.4 |
| GPT-5.1 (high) | OpenAI | 37.5 | 49.4 | - | - | $3.4 |
| Gemini 3.5 Flash-Lite | 37.4 | 49.3 | 352 t/s | 7.3s | $0.8 | |
| Solar Open2 250B | Upstage | 37.4 | 44.7 | - | - | - |
| Claude 4.5 Sonnet (Reasoning) | Anthropic | 37.4 | 52.1 | - | - | $6.0 |
| Grok 4.20 0309 (Reasoning) | SpaceXAI | 37.4 | - | - | - | $3.0 |
| MiMo-V2-Omni-0327 | Xiaomi | 37.3 | - | - | - | - |
| GPT-5 Codex (high) | OpenAI | 37.0 | - | - | - | $3.4 |
| Grok 4.3 (medium) | SpaceXAI | 36.9 | - | 125 t/s | 16.9s | $1.6 |
| Claude Sonnet 4.6 (Non-reasoning, High Effort) | Anthropic | 36.8 | - | - | - | $6.0 |
| Grok 4.3 (low) | SpaceXAI | 36.3 | - | 113 t/s | 8.2s | $1.6 |
| GLM-5.1 (Non-reasoning) | Z AI | 36.3 | - | - | - | $2.1 |
| Kimi K2.5 (Reasoning) | Kimi | 36.0 | 46.8 | - | - | $1.2 |
| MiMo-V2-Omni | Xiaomi | 35.9 | - | - | - | - |
| Gemini 3.5 Flash (minimal) | 35.8 | - | 244 t/s | 0.7s | $3.4 | |
| GPT-5.5 (Non-reasoning) | OpenAI | 35.8 | 56.5 | - | - | $11 |
| GPT-5.1 Codex (high) | OpenAI | 35.6 | - | - | - | $3.4 |
| Claude Opus 4.5 (Non-reasoning) | Anthropic | 35.6 | - | - | - | $10 |
| Kimi K2.6 (Non-reasoning) | Kimi | 35.4 | - | - | - | $1.7 |
| GPT-5 (high) | OpenAI | 35.3 | 37.8 | - | - | $3.4 |
| GLM 5V Turbo (Reasoning) | Z AI | 35.3 | - | - | - | - |
| Muse Glimmer (high) | Meta | 35.1 | 49.0 | 105 t/s | 0.4s | $0.6 |
| Claude Sonnet 4.6 (Non-reasoning, Low Effort) | Anthropic | 35.1 | - | 44 t/s | 1.1s | $6.0 |
| A.X-K2 | SK Telecom | 35.0 | 38.8 | - | - | - |
| GLM-5.2 (Non-reasoning) | Z AI | 34.8 | 46.5 | 110 t/s | 1.4s | $2.1 |
| GPT-5.6 Terra (Non-reasoning) | OpenAI | 34.6 | 52.3 | 87 t/s | 0.7s | $4.5 |
| GPT-5 (medium) | OpenAI | 34.6 | - | - | - | $3.4 |
| Qwen3.5 27B (Reasoning) | Alibaba | 34.6 | - | - | - | $0.8 |
| Claude 4.1 Opus (Reasoning) | Anthropic | 34.5 | - | - | - | $30 |
| MiniMax-M2.5 | MiniMax | 34.5 | - | - | - | $0.5 |
| GLM-4.7 (Reasoning) | Z AI | 34.5 | 45.3 | - | - | $1.0 |
| Hy3-preview (Reasoning) | Tencent | 34.4 | - | - | - | $0.1 |
| Qwen3.5 397B A17B (Reasoning) | Alibaba | 34.3 | 48.2 | 76 t/s | 1.4s | $1.4 |
| GPT-5.5 Instant (May 2026) | OpenAI | 34.3 | - | - | - | $11 |
| Grok 4 | SpaceXAI | 34.1 | - | - | - | $6.0 |
| MiMo-V2-Flash (Feb 2026) | Xiaomi | 34.0 | - | - | - | - |
| LongCat 2.0 | LongCat | 34.0 | 45.3 | 42 t/s | 1.8s | $1.3 |
| G9v3-39A5B | AI9Stars | 34.0 | 33.1 | - | - | - |
| GPT-5.6 Luna (low) | OpenAI | 33.9 | 44.2 | 124 t/s | 1.5s | $0.5 |
| Gemini 3 Pro Preview (low) | 33.9 | - | - | - | $4.5 | |
 KAT Coder Pro V2 | KwaiKAT | 33.7 | 59.5 | 108 t/s | 0.9s | $0.5 |
| Kimi K2 Thinking | Kimi | 33.5 | - | - | - | $1.1 |
| o3-pro | OpenAI | 33.3 | - | - | - | $35 |
| GLM-5 (Non-reasoning) | Z AI | 33.2 | - | - | - | $1.6 |
| Qwen3.5 122B A10B (Reasoning) | Alibaba | 32.8 | 45.7 | 134 t/s | 1.0s | $1.1 |
| DeepSeek V3.2 (Reasoning) | DeepSeek | 32.8 | 44.2 | - | - | $0.3 |
| Qwen3.5 397B A17B (Non-reasoning) | Alibaba | 32.7 | - | 74 t/s | 1.4s | $1.4 |
| Qwen3 Max Thinking | Alibaba | 32.5 | - | - | - | - |
| Qwen3.6 35B A3B (Reasoning) | Alibaba | 32.1 | 41.9 | 136 t/s | 1.0s | $0.8 |
| MiniMax-M2.1 | MiniMax | 32.1 | - | - | - | $0.5 |
| DeepSeek V4 Pro (Non-reasoning) | DeepSeek | 31.9 | - | 63 t/s | 1.4s | $0.5 |
| GPT-5 (low) | OpenAI | 31.9 | - | - | - | $3.4 |
| MiMo-V2-Flash (Reasoning) | Xiaomi | 31.9 | - | - | - | $0.1 |
| Ring-2.6-1T | InclusionAI | 31.7 | 42.8 | 124 t/s | 2.2s | $0.8 |
| Claude 4 Opus (Reasoning) | Anthropic | 31.7 | - | - | - | $30 |
| GPT-5 mini (medium) | OpenAI | 31.6 | - | - | - | $0.7 |
| DeepSeek V3.1 Terminus (Reasoning) | DeepSeek | 31.4 | 43.5 | - | - | $1.9 |
| Qwen3.5 Omni Plus | Alibaba | 31.3 | - | 50 t/s | 1.2s | $1.5 |
| Qwen3.6 27B (Non-reasoning) | Alibaba | 31.3 | 46.6 | 55 t/s | 1.5s | $1.4 |
| GPT-5.1 Codex mini (high) | OpenAI | 31.3 | - | - | - | $0.7 |
| Grok 4.1 Fast (Reasoning) | SpaceXAI | 31.3 | - | - | - | - |
| o3 | OpenAI | 31.1 | - | 139 t/s | 4.7s | $3.5 |
 K-EXAONE 2.0 0803 | LG AI Research | 31.0 | 40.6 | - | - | - |
'%3e%3cpath%20fill-rule='evenodd'%20clip-rule='evenodd'%20d='M44.024%200.471985H46.088V2.32598H48V4.26198H46.088V8.03198H44.024V4.26398H40.268V2.32398H44.024V0.471985ZM5.2%2025.214V4.21198H7.138V25.216H5.198L5.2%2025.214ZM26.046%2026.534H47.946V28.37H35.53V47.528H26.046V26.532V26.534ZM11.258%207.13798V31.85H0V40.87H20.772V16.472H41.718L41.712%207.13598L11.258%207.13798Z'%20fill='url(%23paint0_linear_2900_925)'/%3e%3c/g%3e%3cdefs%3e%3clinearGradient%20id='paint0_linear_2900_925'%20x1='3.292'%20y1='4.30398'%20x2='36.684'%20y2='44.654'%20gradientUnits='userSpaceOnUse'%3e%3cstop%20stop-color='%2301A9FF'/%3e%3cstop%20offset='1'%20stop-color='%230160FF'/%3e%3c/linearGradient%3e%3cclipPath%20id='clip0_2900_925'%3e%3crect%20width='48'%20height='48'%20fill='white'/%3e%3c/clipPath%3e%3c/defs%3e%3c/svg%3e) Step 3.7 Flash | StepFun | 30.9 | 39.6 | 131 t/s | 1.7s | $0.4 |
| GPT-5.4 nano (medium) | OpenAI | 30.8 | - | - | - | $0.5 |
| GPT-5.4 mini (medium) | OpenAI | 30.5 | - | - | - | $1.7 |
 Mistral Medium 3.5 | Mistral | 30.4 | 46.9 | 131 t/s | 0.6s | $3.0 |
| Kimi K2.5 (Non-reasoning) | Kimi | 30.1 | - | - | - | $1.2 |
| Qwen3.5 27B (Non-reasoning) | Alibaba | 30.0 | - | - | - | $0.8 |
| Claude 4.5 Haiku (Reasoning) | Anthropic | 29.9 | 43.9 | 148 t/s | 12.1s | $2.0 |
| Claude 4.5 Sonnet (Non-reasoning) | Anthropic | 29.9 | - | - | - | $6.0 |
| Qwen3.5 35B A3B (Reasoning) | Alibaba | 29.9 | - | - | - | $0.7 |
| Claude 4 Sonnet (Reasoning) | Anthropic | 29.8 | 37.6 | - | - | $6.0 |
| Gemma 4 31B (Reasoning) | 29.7 | 43.4 | 36 t/s | 1.0s | - | |
| DeepSeek V4 Flash (Non-reasoning) | DeepSeek | 29.3 | - | - | - | $0.1 |
| GLM-4.6 (Reasoning) | Z AI | 29.3 | 45.8 | - | - | $1.0 |
| GPT-5.5 Instant (June 2026) | OpenAI | 29.2 | 39.4 | 139 t/s | 0.9s | $11 |
| JT-35B-Flash | China Mobile | 29.0 | - | - | - | - |
| MiniMax-M2 | MiniMax | 28.9 | - | - | - | $0.5 |
| KAT-Coder-Pro V1 | KwaiKAT | 28.9 | - | - | - | - |
| Claude 4.1 Opus (Non-reasoning) | Anthropic | 28.8 | - | - | - | $30 |
| MiMo-V2.5-Pro (Non-reasoning) | Xiaomi | 28.4 | - | 39 t/s | 2.3s | $0.5 |
| GPT-5.4 (Non-reasoning) | OpenAI | 28.3 | - | - | - | $5.6 |
| Qwen3.5 122B A10B (Non-reasoning) | Alibaba | 28.2 | 43.3 | 144 t/s | 1.0s | $1.1 |
| Gemini 3 Flash Preview (Non-reasoning) | 27.9 | - | - | - | $1.1 | |
| Grok 4 Fast (Reasoning) | SpaceXAI | 27.9 | - | - | - | $0.3 |
| Claude 3.7 Sonnet (Reasoning) | Anthropic | 27.6 | 36.4 | - | - | - |
| GLM-4.7 (Non-reasoning) | Z AI | 27.1 | - | - | - | $1.0 |
| GPT-5.6 Luna (Non-reasoning) | OpenAI | 26.8 | 39.3 | 123 t/s | 0.8s | $0.5 |
| Hy3-preview (Non-reasoning) | Tencent | 26.6 | - | - | - | $0.1 |
| Ling-2.6-1T | InclusionAI | 26.6 | - | - | - | $0.8 |
 Doubao Seed Code | ByteDance Seed | 26.5 | - | - | - | - |
| GPT-5.2 (Non-reasoning) | OpenAI | 26.5 | - | - | - | $4.8 |
| Step 3.5 Flash 2603 | StepFun | 26.5 | - | - | - | $0.1 |
| Gemma 4 26B A4B (Reasoning) | 26.1 | 39.3 | - | - | $0.2 | |
| o4-mini (high) | OpenAI | 26.1 | - | - | - | $1.9 |
| Claude 4 Opus (Non-reasoning) | Anthropic | 26.0 | - | - | - | $30 |
| Claude 4 Sonnet (Non-reasoning) | Anthropic | 26.0 | - | - | - | $6.0 |
| Step 3.5 Flash | StepFun | 26.0 | - | - | - | $0.1 |
| Gemini 2.5 Pro | 25.9 | 33.3 | - | - | $3.4 | |
| DeepSeek V3.2 Exp (Reasoning) | DeepSeek | 25.9 | - | - | - | $0.3 |
| GPT-5 mini (high) | OpenAI | 25.8 | 15.6 | - | - | $0.7 |
| Nemotron 3 Super 120B A12B (Reasoning) | NVIDIA | 25.7 | 37.7 | 143 t/s | 1.1s | $0.3 |
| Gemini 3.1 Flash-Lite | 25.6 | 34.7 | - | - | $0.6 | |
| Qwen3 Max Thinking (Preview) | Alibaba | 25.5 | - | - | - | $2.4 |
| MiMo-V2-Flash (Non-reasoning) | Xiaomi | 25.1 | 49.8 | - | - | - |
| DeepSeek V3.2 (Non-reasoning) | DeepSeek | 25.1 | - | - | - | $0.3 |
| Grok 4.3 (Non-reasoning) | SpaceXAI | 25.0 | 35.2 | 106 t/s | 0.6s | $1.6 |
| Qwen3.6 35B A3B (Non-reasoning) | Alibaba | 24.6 | 28.1 | 138 t/s | 0.9s | $0.8 |
| Ling 3.0 Tiny | InclusionAI | 24.5 | 26.5 | 203 t/s | 2.1s | - |
| Qwen3 Max | Alibaba | 24.5 | - | - | - | $2.4 |
| Qwen3.5 35B A3B (Non-reasoning) | Alibaba | 24.3 | 37.0 | 178 t/s | 1.3s | $0.7 |
| Gemini 2.5 Flash Preview (Sep '25) (Reasoning) | 24.2 | - | - | - | - | |
| gpt-oss-120b (high) | OpenAI | 24.1 | 30.4 | 185 t/s | 0.5s | $0.3 |
| Claude 4.5 Haiku (Non-reasoning) | Anthropic | 24.1 | - | 96 t/s | 0.6s | $2.0 |
| Kimi K2 0905 | Kimi | 24.0 | - | - | - | $1.1 |
| o1 | OpenAI | 23.9 | 39.7 | - | - | $26 |
| Claude 3.7 Sonnet (Non-reasoning) | Anthropic | 23.9 | - | - | - | $6.0 |
| Nemotron 3.5 Lightning | NVIDIA | 23.6 | 26.8 | 308 t/s | 1.0s | $0.1 |
| Gemini 2.5 Pro Preview (Mar' 25) | 23.4 | 46.7 | - | - | - | |
| GLM-4.6 (Non-reasoning) | Z AI | 23.4 | - | - | - | $1.0 |
| GLM-4.7-Flash (Reasoning) | Z AI | 23.3 | - | - | - | $0.2 |
| Grok 4.20 0309 (Non-reasoning) | SpaceXAI | 22.9 | - | - | - | $3.0 |
| Grok 3 mini Reasoning (high) | SpaceXAI | 22.9 | - | - | - | $0.3 |
 Command A+ | Cohere | 22.8 | 27.8 | 228 t/s | 0.1s | - |
| Gemini 2.5 Pro Preview (May' 25) | 22.7 | - | - | - | $3.4 | |
| DeepSeek V3.2 Speciale | DeepSeek | 22.6 | - | - | - | - |
| K-EXAONE (Reasoning) | LG AI Research | 22.5 | 32.1 | - | - | - |
| Gemma 4 31B (Non-reasoning) | 22.3 | 33.2 | 106 t/s | 0.5s | $0.2 | |
 ERNIE 5.0 Thinking Preview | Baidu | 22.3 | - | - | - | - |
| Gemma 4 12B (Reasoning) | 22.2 | 31.0 | 137 t/s | 1.3s | $0.1 | |
| Grok 4.20 0309 v2 (Non-reasoning) | SpaceXAI | 22.2 | - | - | - | $1.6 |
 Nova 2.0 Pro Preview (medium) | Amazon | 22.1 | 34.0 | 118 t/s | 20.4s | $3.4 |
| Grok Code Fast 1 | SpaceXAI | 22.0 | - | - | - | - |
| Mercury 2 | Inception | 21.9 | 31.1 | 1164 t/s | 4.2s | $0.4 |
| Qwen3.5 9B (Reasoning) | Alibaba | 21.8 | 28.7 | 87 t/s | 1.0s | $0.2 |
| DeepSeek V3.2 Exp (Non-reasoning) | DeepSeek | 21.7 | - | - | - | $0.3 |
| DeepSeek V3.1 Terminus (Non-reasoning) | DeepSeek | 21.7 | - | - | - | $0.5 |
| Apriel-v1.5-15B-Thinker | ServiceNow | 21.6 | - | - | - | - |
| DeepSeek V3.1 (Non-reasoning) | DeepSeek | 21.4 | - | - | - | $0.8 |
| Nova 2.0 Omni (medium) | Amazon | 21.3 | - | - | - | $0.8 |
| Qwen3 Coder Next | Alibaba | 21.3 | 36.2 | 131 t/s | 0.9s | $0.6 |
| DeepSeek V3.1 (Reasoning) | DeepSeek | 21.0 | - | - | - | $0.9 |
| Qwen3 VL 235B A22B (Reasoning) | Alibaba | 20.9 | - | - | - | $1.3 |
| Nova 2.0 Lite (high) | Amazon | 20.8 | 23.0 | 194 t/s | 9.7s | $0.8 |
| Apriel-v1.6-15B-Thinker | ServiceNow | 20.8 | - | - | - | - |
| GPT-5.1 (Non-reasoning) | OpenAI | 20.7 | - | - | - | $3.4 |
| Qwen3.5 9B (Non-reasoning) | Alibaba | 20.6 | 23.5 | 98 t/s | 0.6s | $0.2 |
| EXAONE 4.5 33B | LG AI Research | 20.5 | 23.6 | - | - | - |
| Gemma 4 26B A4B (Non-reasoning) | 20.4 | - | 91 t/s | 0.5s | $0.2 | |
| Qwen3.5 4B (Reasoning) | Alibaba | 20.4 | 22.6 | 15 t/s | 0.6s | $0.1 |
| DeepSeek R1 0528 (May '25) | DeepSeek | 20.4 | - | - | - | $1.8 |
| Gemini 2.5 Flash (Reasoning) | 20.3 | - | - | - | $0.8 | |
| North Mini Code | Cohere | 20.2 | 36.5 | 28 t/s | 0.3s | - |
| GPT-5 nano (high) | OpenAI | 20.1 | - | - | - | $0.1 |
| Qwen3 235B A22B 2507 (Reasoning) | Alibaba | 19.9 | 22.1 | - | - | $0.7 |
| Nova 2.0 Pro Preview (low) | Amazon | 19.8 | 25.9 | 129 t/s | 4.4s | $3.4 |
| Mistral Small 4 (Reasoning) | Mistral | 19.7 | 26.6 | 154 t/s | 0.6s | $0.3 |
| Kimi K2 | Kimi | 19.7 | - | - | - | $1.0 |
| GLM-4.5 (Reasoning) | Z AI | 19.7 | - | - | - | - |
| GPT-4.1 | OpenAI | 19.6 | - | - | - | $3.5 |
| Qwen3 Max (Preview) | Alibaba | 19.4 | - | - | - | $2.4 |
| Devstral 2 | Mistral | 19.2 | 31.3 | 47 t/s | 0.8s | - |
| Nova 2.0 Lite (medium) | Amazon | 19.2 | - | 192 t/s | 13.4s | $0.8 |
| Qwen3.5 Omni Flash | Alibaba | 19.2 | - | 225 t/s | 1.0s | $0.3 |
| o3-mini | OpenAI | 19.2 | - | - | - | $1.9 |
| GPT-5 nano (medium) | OpenAI | 19.2 | - | - | - | $0.1 |
| o1-pro | OpenAI | 19.1 | - | - | - | $263 |
| Gemini 2.5 Flash Preview (Sep '25) (Non-reasoning) | 19.1 | - | - | - | - | |
| JT-MINI | China Mobile | 18.8 | - | - | - | - |
| Trinity Large Thinking | Arcee AI | 18.7 | 25.8 | 198 t/s | 0.7s | $0.4 |
| DeepSeek R1 (Jan '25) | DeepSeek | 18.6 | 24.6 | - | - | $2.5 |
| Grok 3 | SpaceXAI | 18.6 | - | - | - | $8.0 |
| Seed-OSS-36B-Instruct | ByteDance Seed | 18.5 | - | - | - | $0.3 |
| Qwen3 235B A22B 2507 Instruct | Alibaba | 18.4 | - | - | - | $0.4 |
| HyperNova 60B 2605 | Multiverse Computing | 18.3 | 23.2 | - | - | - |
| Qwen3 Coder 480B A35B Instruct | Alibaba | 18.2 | - | - | - | $3.0 |
| Qwen3 VL 32B (Reasoning) | Alibaba | 18.1 | - | - | - | $0.3 |
| Magistral Medium 1.2 | Mistral | 18.0 | 21.3 | - | - | - |
| Nova 2.0 Lite (low) | Amazon | 18.0 | - | 194 t/s | 4.5s | $0.8 |
 Sonar Reasoning Pro | Perplexity | 18.0 | - | - | - | - |
| MiniMax M1 80k | MiniMax | 17.9 | - | - | - | $1.0 |
| Nemotron Cascade 2 30B A3B | NVIDIA | 17.8 | 25.3 | - | - | - |
| GPT-5.4 nano (Non-Reasoning) | OpenAI | 17.8 | - | - | - | $0.5 |
| Devstral Small 2 | Mistral | 17.7 | 29.3 | 115 t/s | 0.6s | - |
| Gemini 2.5 Flash Preview (Reasoning) | 17.7 | - | - | - | - | |
| K2 Think V2 | MBZUAI Institute of Foundation Models | 17.4 | 21.0 | - | - | - |
| LongCat Flash Lite | LongCat | 17.4 | - | - | - | - |
| GPT-5 (minimal) | OpenAI | 17.3 | - | - | - | $3.4 |
 HyperCLOVA X SEED Think (32B) | Naver | 17.2 | - | - | - | - |
| o1-preview | OpenAI | 17.2 | 34.0 | - | - | $29 |
| Grok 4.1 Fast (Non-reasoning) | SpaceXAI | 17.0 | - | - | - | - |
| K-EXAONE (Non-reasoning) | LG AI Research | 16.9 | - | - | - | - |
| Qwen3 Next 80B A3B (Reasoning) | Alibaba | 16.9 | 17.4 | 205 t/s | 1.2s | $0.4 |
| GLM-4.6V (Reasoning) | Z AI | 16.9 | - | - | - | $0.5 |
| GPT-5.4 mini (Non-Reasoning) | OpenAI | 16.8 | - | - | - | $1.7 |
| Nova 2.0 Omni (low) | Amazon | 16.7 | - | - | - | $0.8 |
| GLM-4.5-Air | Z AI | 16.7 | - | - | - | $0.4 |
 Mi:dm K 2.5 Pro | Korea Telecom | 16.6 | - | - | - | - |
| Grok 4 Fast (Non-reasoning) | SpaceXAI | 16.6 | - | - | - | $0.3 |
| Ring-1T | InclusionAI | 16.3 | - | - | - | - |
| G9v3-3B | AI9Stars | 16.2 | 9.9 | - | - | - |
| Qwen3.5 4B (Non-reasoning) | Alibaba | 16.1 | 20.3 | 13 t/s | 0.6s | $0.1 |
| Mistral Large 3 | Mistral | 15.9 | 20.1 | 54 t/s | 0.6s | $0.8 |
| INTELLECT-3 | Prime Intellect | 15.7 | - | - | - | - |
| o3-mini (high) | OpenAI | 15.7 | 16.3 | - | - | $1.9 |
| GLM-4.7-Flash (Non-reasoning) | Z AI | 15.6 | - | - | - | $0.2 |
| GPT-5 (ChatGPT) | OpenAI | 15.4 | - | - | - | - |
| gpt-oss-20b (high) | OpenAI | 15.2 | 20.7 | 114 t/s | 0.7s | $0.1 |
| Solar Open 100B (Reasoning) | Upstage | 15.2 | - | - | - | - |
| Gemini 2.5 Flash-Lite Preview (Sep '25) (Reasoning) | 15.2 | - | - | - | $0.2 | |
| DeepSeek V3 0324 | DeepSeek | 15.2 | 21.2 | - | - | $0.5 |
| Grok 3 Reasoning Beta | SpaceXAI | 15.2 | - | - | - | - |
| Nemotron 3 Nano Omni 30B A3B Reasoning | NVIDIA | 15.0 | 13.8 | 330 t/s | 0.5s | $0.1 |
| gpt-oss-120b (low) | OpenAI | 14.9 | 21.2 | 246 t/s | 0.6s | $0.3 |
| Mistral Small 3.1 | Mistral | 14.9 | 26.3 | - | - | $0.1 |
| GPT-4.1 mini | OpenAI | 14.8 | 20.2 | - | - | $0.7 |
| Mistral Medium 3.1 | Mistral | 14.7 | 20.5 | - | - | $0.8 |
| Qwen3 30B A3B 2507 (Reasoning) | Alibaba | 14.6 | 12.1 | - | - | $0.8 |
| Llama 4 Maverick | Meta | 14.5 | 16.3 | 101 t/s | 0.6s | $0.4 |
| Solar Pro 3 | Upstage | 14.5 | 16.2 | 159 t/s | 1.1s | $0.3 |
| NVIDIA Nemotron 3 Nano 30B A3B (Reasoning) | NVIDIA | 14.5 | 14.4 | 298 t/s | 0.6s | $0.1 |
| MiniMax M1 40k | MiniMax | 14.5 | - | - | - | - |
| gpt-oss-20b (low) | OpenAI | 14.4 | - | 136 t/s | 0.8s | $0.1 |
| Nova 2.0 Pro Preview (Non-reasoning) | Amazon | 14.4 | 20.9 | 132 t/s | 0.7s | $3.4 |
| Qwen3 VL 235B A22B Instruct | Alibaba | 14.4 | - | - | - | $0.7 |
| GPT-5 mini (minimal) | OpenAI | 14.3 | - | - | - | $0.7 |
| K2-V2 (high) | MBZUAI Institute of Foundation Models | 14.2 | - | - | - | - |
| Gemini 2.5 Flash (Non-reasoning) | 14.2 | - | - | - | $0.8 | |
| DeepSeek V3 (Dec '24) | DeepSeek | 14.2 | 23.0 | - | - | $0.5 |
| Ling 2.6 Flash | InclusionAI | 14.2 | 25.3 | - | - | $0.1 |
| o1-mini | OpenAI | 14.0 | - | - | - | - |
| Qwen3 Next 80B A3B Instruct | Alibaba | 13.8 | - | 189 t/s | 1.0s | $0.4 |
| GPT-4.5 (Preview) | OpenAI | 13.6 | - | - | - | - |
| Tri-21B-think Preview | Trillion Labs | 13.6 | - | - | - | - |
| Qwen3 Coder 30B A3B Instruct | Alibaba | 13.6 | - | - | - | $0.9 |
| DiffusionGemma 26B A4B | 13.5 | 19.7 | - | - | - | |
| Qwen3 235B A22B (Reasoning) | Alibaba | 13.5 | - | - | - | $2.6 |
| Qwen3 VL 30B A3B (Reasoning) | Alibaba | 13.4 | - | - | - | $0.8 |
| QwQ 32B | Alibaba | 13.4 | - | - | - | $0.7 |
| Gemini 2.0 Flash Thinking Experimental (Jan '25) | 13.3 | 24.1 | - | - | - | |
| Gemma 4 12B (Non-reasoning) | 13.2 | - | 140 t/s | 1.3s | $0.1 | |
| Gemini 2.5 Flash-Lite Preview (Sep '25) (Non-reasoning) | 13.1 | - | - | - | $0.2 | |
| Motif-2-12.7B-Reasoning | Motif Technologies | 12.8 | - | - | - | - |
| Nova Premier | Amazon | 12.7 | - | 48 t/s | 1.1s | $5.0 |
| Ling-1T | InclusionAI | 12.7 | - | - | - | - |
| Mistral Medium 3 | Mistral | 12.5 | - | - | - | $0.8 |
| Magistral Medium 1 | Mistral | 12.5 | - | - | - | - |
| Solar Pro 2 (Preview) (Reasoning) | Upstage | 12.5 | - | - | - | - |
| Llama Nemotron Super 49B v1.5 (Reasoning) | NVIDIA | 12.4 | - | 122 t/s | 3.3s | $0.4 |
| K2-V2 (medium) | MBZUAI Institute of Foundation Models | 12.4 | - | - | - | - |
| Celeris-1 | Celeris | 12.4 | 14.4 | 2067 t/s | 0.5s | $0.3 |
| Devstral Medium | Mistral | 12.4 | - | - | - | - |
| Mistral Small 4 (Non-reasoning) | Mistral | 12.3 | - | 140 t/s | 0.5s | $0.3 |
| Tri-21B-Think | Trillion Labs | 12.3 | - | - | - | - |
| GPT-4o (March 2025, chatgpt-4o-latest) | OpenAI | 12.3 | - | - | - | - |
| Gemma 4 E4B (Reasoning) | 12.2 | 9.4 | 59 t/s | 0.3s | $0.0 | |
| Gemini 2.0 Flash (Feb '25) | 12.2 | - | - | - | - | |
| Claude 3.5 Haiku | Anthropic | 12.2 | 15.9 | - | - | - |
| Llama 3.3 Nemotron Super 49B v1 (Reasoning) | NVIDIA | 12.2 | - | - | - | - |
| Sarvam 105B (high) | Sarvam | 11.9 | - | - | - | $0.1 |
| MiniCPM5-1B (Reasoning) | OpenBMB | 11.9 | - | - | - | - |
| Qwen3 4B 2507 (Reasoning) | Alibaba | 11.9 | - | - | - | - |
| Nova 2.0 Lite (Non-reasoning) | Amazon | 11.8 | - | 202 t/s | 0.7s | $0.8 |
| Gemini 2.0 Pro Experimental (Feb '25) | 11.8 | 25.5 | - | - | - | |
| Claude 3 Opus | Anthropic | 11.8 | 19.5 | - | - | $30 |
| Devstral Small (May '25) | Mistral | 11.8 | - | - | - | - |
| MiniCPM5-1B (Non-reasoning) | OpenBMB | 11.7 | - | - | - | - |
| Gemini 2.5 Flash Preview (Non-reasoning) | 11.6 | - | - | - | - | |
| Sonar Reasoning | Perplexity | 11.6 | - | - | - | - |
| Magistral Small 1.2 | Mistral | 11.5 | 14.7 | - | - | $0.8 |
| Gemini 2.5 Flash-Lite (Reasoning) | 11.4 | - | - | - | $0.2 | |
| Qwen3 32B (Reasoning) | Alibaba | 11.4 | 15.3 | - | - | $0.3 |
| Ministral 3 14B | Mistral | 11.2 | 14.4 | 104 t/s | 0.5s | $0.2 |
| GPT-4o (Nov '24) | OpenAI | 11.1 | - | - | - | $4.4 |
| LFM2.5-2.6B | Liquid AI | 11.0 | 7.7 | - | - | - |
| Nanbeige4.1-3B | Nanbeige | 11.0 | 9.6 | - | - | - |
| DeepSeek R1 Distill Qwen 32B | DeepSeek | 11.0 | - | - | - | - |
| Qwen3 VL 32B Instruct | Alibaba | 11.0 | - | - | - | $0.3 |
| GLM-4.6V (Non-reasoning) | Z AI | 10.9 | - | - | - | $0.5 |
| Qwen3 235B A22B (Non-reasoning) | Alibaba | 10.8 | - | - | - | $1.2 |
| Mistral Small 3.2 | Mistral | 10.7 | 12.5 | - | - | $0.1 |
| Gemini 2.0 Flash (experimental) | 10.6 | - | - | - | - | |
| Magistral Small 1 | Mistral | 10.6 | - | - | - | - |
| EXAONE 4.0 32B (Reasoning) | LG AI Research | 10.5 | - | - | - | - |
| Qwen3 VL 8B (Reasoning) | Alibaba | 10.5 | - | - | - | $0.7 |
| Nova 2.0 Omni (Non-reasoning) | Amazon | 10.4 | - | - | - | $0.8 |
| Qwen3 14B (Reasoning) | Alibaba | 10.4 | 13.8 | - | - | $1.3 |
| Llama 4 Scout | Meta | 10.3 | 8.2 | 138 t/s | 0.5s | $0.3 |
| DeepSeek R1 0528 Qwen3 8B | DeepSeek | 10.3 | - | - | - | - |
| Qwen2.5 Max | Alibaba | 10.1 | - | - | - | - |
 Hermes 4 - Llama-3.1 70B (Reasoning) | Nous Research | 9.9 | - | 89 t/s | 0.6s | $0.2 |
| Gemini 1.5 Pro (Sep '24) | 9.9 | 23.6 | - | - | - | |
| Solar Pro 2 (Preview) (Non-reasoning) | Upstage | 9.9 | - | - | - | - |
| Qwen3 VL 30B A3B Instruct | Alibaba | 9.9 | - | - | - | $0.3 |
| Claude 3.5 Sonnet (Oct '24) | Anthropic | 9.8 | 30.2 | - | - | $6.0 |
| DeepSeek R1 Distill Llama 70B | DeepSeek | 9.8 | - | - | - | $0.8 |
 Falcon-H1R-7B | TII UAE | 9.7 | - | - | - | - |
| DeepSeek R1 Distill Qwen 14B | DeepSeek | 9.7 | - | - | - | - |
| GPT-4.1 nano | OpenAI | 9.6 | 11.1 | - | - | $0.2 |
| Ling-flash-2.0 | InclusionAI | 9.6 | - | - | - | $0.2 |
| Gemma 4 E2B (Reasoning) | 9.5 | 7.2 | - | - | - | |
| Qwen3 Omni 30B A3B (Reasoning) | Alibaba | 9.5 | - | 109 t/s | 0.9s | $0.4 |
| GPT-4o (Aug '24) | OpenAI | 9.4 | - | - | - | $4.4 |
| Sonar | Perplexity | 9.4 | - | - | - | - |
| Qwen2.5 Instruct 72B | Alibaba | 9.4 | - | - | - | $0.5 |
| Llama 3.3 Instruct 70B | Meta | 9.3 | 11.9 | 86 t/s | 0.6s | $0.7 |
| Step3 VL 10B | StepFun | 9.3 | - | - | - | - |
| Qwen3 30B A3B (Reasoning) | Alibaba | 9.2 | - | - | - | $0.8 |
| Devstral Small (Jul '25) | Mistral | 9.1 | - | - | - | - |
| Sonar Pro | Perplexity | 9.1 | - | - | - | - |
| QwQ 32B-Preview | Alibaba | 9.1 | - | - | - | - |
| Ministral 3 8B | Mistral | 9.0 | 9.7 | 127 t/s | 0.5s | $0.1 |
| Mistral Large 2 (Nov '24) | Mistral | 9.0 | - | - | - | - |
| GLM-4.5V (Reasoning) | Z AI | 9.0 | - | - | - | $0.9 |
| Llama 3.1 Nemotron Ultra 253B v1 (Reasoning) | NVIDIA | 8.9 | - | 53 t/s | 0.7s | $0.9 |
| ERNIE 4.5 300B A47B | Baidu | 8.9 | - | - | - | $0.5 |
| Qwen3 30B A3B 2507 Instruct | Alibaba | 8.9 | - | - | - | $0.3 |
| NVIDIA Nemotron Nano 12B v2 VL (Reasoning) | NVIDIA | 8.8 | - | 83 t/s | 3.4s | $0.3 |
| Hermes 4 - Llama-3.1 405B (Reasoning) | Nous Research | 8.8 | - | 30 t/s | 0.9s | $1.5 |
| Solar Pro 2 (Reasoning) | Upstage | 8.8 | - | - | - | - |
| Gemma 4 E4B (Non-reasoning) | 8.7 | - | 59 t/s | 0.3s | $0.0 | |
| NVIDIA Nemotron Nano 9B V2 (Reasoning) | NVIDIA | 8.7 | - | 228 t/s | 2.5s | $0.1 |
 Granite 4.1 30B | IBM | 8.7 | 10.4 | - | - | - |
| NVIDIA Nemotron 3 Nano 4B | NVIDIA | 8.6 | 8.0 | - | - | - |
| Hermes 4 - Llama-3.1 405B (Non-reasoning) | Nous Research | 8.6 | - | 33 t/s | 0.8s | $1.5 |
| Gemini 2.0 Flash-Lite (Feb '25) | 8.6 | - | - | - | - | |
| Llama Nemotron Super 49B v1.5 (Non-reasoning) | NVIDIA | 8.5 | - | 118 t/s | 3.7s | $0.4 |
| Qwen3 32B (Non-reasoning) | Alibaba | 8.5 | - | - | - | $0.3 |
| Kimi Linear 48B A3B Instruct | Kimi | 8.4 | - | - | - | - |
| K2-V2 (low) | MBZUAI Institute of Foundation Models | 8.4 | - | - | - | - |
| GPT-4o (May '24) | OpenAI | 8.4 | 24.2 | - | - | $7.5 |
| Gemini 2.0 Flash-Lite (Preview) | 8.4 | - | - | - | - | |
| Llama 3.1 Nemotron Nano 4B v1.1 (Reasoning) | NVIDIA | 8.4 | - | - | - | - |
| Llama 3.1 Instruct 405B | Meta | 8.3 | - | - | - | - |
| Llama 3.3 Nemotron Super 49B v1 (Non-reasoning) | NVIDIA | 8.3 | - | - | - | - |
| Qwen3 8B (Reasoning) | Alibaba | 8.3 | 9.0 | - | - | $0.7 |
| Qwen3 4B (Reasoning) | Alibaba | 8.2 | - | - | - | - |
| Qwen3 VL 8B Instruct | Alibaba | 8.2 | - | - | - | $0.3 |
| LFM2.5-8B-A1B | Liquid AI | 8.1 | - | 346 t/s | 1.1s | - |
| GPT-4o (ChatGPT) | OpenAI | 8.1 | - | - | - | - |
| Claude 3.5 Sonnet (June '24) | Anthropic | 8.1 | 26.0 | - | - | $6.0 |
| Llama 3.1 Tulu3 405B | Allen Institute for AI | 8.1 | - | - | - | - |
| Ring-flash-2.0 | InclusionAI | 8.0 | - | - | - | $0.2 |
| Pixtral Large | Mistral | 8.0 | - | - | - | - |
| Olmo 3.1 32B Think | Allen Institute for AI | 7.9 | - | - | - | - |
| GPT-5 nano (minimal) | OpenAI | 7.8 | - | - | - | $0.1 |
| Gemini 1.5 Flash (Sep '24) | 7.8 | - | - | - | - | |
| Grok 2 (Dec '24) | SpaceXAI | 7.8 | - | - | - | - |
| GPT-4 Turbo | OpenAI | 7.7 | 21.5 | - | - | $15 |
| Qwen3 VL 4B (Reasoning) | Alibaba | 7.7 | - | - | - | - |
| Solar Pro 2 (Non-reasoning) | Upstage | 7.6 | - | - | - | - |
| Command A | Cohere | 7.5 | - | 70 t/s | 0.3s | $4.4 |
| Nova Pro | Amazon | 7.5 | - | - | - | $1.4 |
| Llama 3.1 Nemotron Instruct 70B | NVIDIA | 7.4 | - | 95 t/s | 2.7s | $1.2 |
| Qwen3.5 2B (Reasoning) | Alibaba | 7.4 | 2.9 | - | - | - |
| Llama 3.1 Instruct 8B | Meta | 7.4 | 5.4 | - | - | $0.0 |
| Gemma 3 27B Instruct | 7.4 | 10.1 | - | - | - | |
| Grok Beta | SpaceXAI | 7.3 | - | - | - | - |
| NVIDIA Nemotron 3 Nano 30B A3B (Non-reasoning) | NVIDIA | 7.2 | - | 257 t/s | 0.4s | $0.1 |
| NVIDIA Nemotron Nano 9B V2 (Non-reasoning) | NVIDIA | 7.2 | - | 174 t/s | 1.2s | $0.1 |
| Qwen2.5 Instruct 32B | Alibaba | 7.2 | - | - | - | - |
| Ministral 3 3B | Mistral | 7.1 | 4.8 | 213 t/s | 0.6s | $0.1 |
| Mistral Large 2 (Jul '24) | Mistral | 7.0 | - | - | - | $3.0 |
| Qwen2.5 Coder Instruct 32B | Alibaba | 6.9 | - | - | - | - |
| Qwen3 4B 2507 Instruct | Alibaba | 6.9 | - | - | - | - |
| GPT-4 | OpenAI | 6.8 | 13.1 | - | - | $38 |
| GLM-4.5V (Non-reasoning) | Z AI | 6.8 | - | - | - | $0.9 |
| Qwen3 14B (Non-reasoning) | Alibaba | 6.8 | - | - | - | $0.6 |
| Hermes 4 - Llama-3.1 70B (Non-reasoning) | Nous Research | 6.7 | - | 91 t/s | 0.6s | $0.2 |
| GPT-4o mini | OpenAI | 6.7 | 11.4 | - | - | $0.3 |
| Gemini 2.5 Flash-Lite (Non-reasoning) | 6.7 | - | - | - | $0.2 | |
| Mistral Small 3 | Mistral | 6.7 | - | - | - | $0.1 |
| Nova Lite | Amazon | 6.7 | - | - | - | $0.1 |
| Qwen3 30B A3B (Non-reasoning) | Alibaba | 6.6 | - | - | - | $0.3 |
| Llama 3.1 Instruct 70B | Meta | 6.5 | - | - | - | $0.6 |
| DeepSeek-V2.5 (Dec '24) | DeepSeek | 6.5 | - | - | - | - |
| Qwen3 4B (Non-reasoning) | Alibaba | 6.5 | - | - | - | - |
| Granite 4.1 8B | IBM | 6.4 | 9.5 | 127 t/s | 0.4s | $0.1 |
| Sarvam 30B (high) | Sarvam | 6.4 | - | - | - | $0.0 |
| Gemini 2.0 Flash Thinking Experimental (Dec '24) | 6.4 | - | - | - | - | |
| DeepSeek-V2.5 | DeepSeek | 6.4 | - | - | - | - |
| Gemma 4 E2B (Non-reasoning) | 6.2 | - | - | - | - | |
| Olmo 3.1 32B Instruct | Allen Institute for AI | 6.2 | - | - | - | - |
| Mistral Saba | Mistral | 6.2 | - | - | - | - |
| DeepSeek R1 Distill Llama 8B | DeepSeek | 6.2 | - | - | - | - |
| Gemini 1.5 Pro (May '24) | 6.1 | 19.8 | - | - | - | |
| Olmo 3 32B Think | Allen Institute for AI | 6.1 | - | - | - | - |
| Llama 3.2 Instruct 90B (Vision) | Meta | 6.0 | - | - | - | - |
| R1 1776 | Perplexity | 6.0 | - | - | - | - |
| Solar Mini | Upstage | 6.0 | - | - | - | $0.1 |
 Reka Flash (Sep '24) | Reka AI | 6.0 | - | - | - | $0.3 |
| Qwen2.5 Turbo | Alibaba | 6.0 | - | - | - | $0.1 |
| Grok-1 | SpaceXAI | 5.8 | - | - | - | - |
 Phi-4 Mini Instruct | Microsoft | 5.7 | 3.8 | 45 t/s | 0.3s | - |
| EXAONE 4.0 32B (Non-reasoning) | LG AI Research | 5.7 | - | - | - | - |
| Qwen2 Instruct 72B | Alibaba | 5.7 | - | - | - | - |
| Gemma 3 12B Instruct | 5.5 | 5.8 | - | - | - | |
| Qwen3.5 2B (Non-reasoning) | Alibaba | 5.3 | 2.4 | - | - | - |
| Qwen3.5 0.8B (Reasoning) | Alibaba | 5.2 | - | - | - | - |
| Gemini 1.5 Flash-8B | 5.2 | - | - | - | - | |
| DeepHermes 3 - Mistral 24B Preview (Non-reasoning) | Nous Research | 5.0 | - | - | - | - |
 Jamba 1.7 Large | AI21 Labs | 5.0 | - | - | - | - |
| Granite 4.0 H Small | IBM | 4.9 | - | 28 t/s | 13.1s | $0.1 |
| Qwen3 Omni 30B A3B Instruct | Alibaba | 4.8 | - | 108 t/s | 0.9s | $0.4 |
| Hermes 3 - Llama-3.1 70B | Nous Research | 4.8 | - | - | - | $0.7 |
| Jamba 1.5 Large | AI21 Labs | 4.8 | - | - | - | $3.5 |
| Qwen3 8B (Non-reasoning) | Alibaba | 4.8 | - | - | - | $0.3 |
| DeepSeek-Coder-V2 | DeepSeek | 4.7 | - | - | - | - |
| OLMo 2 32B | Allen Institute for AI | 4.7 | - | - | - | - |
| Jamba 1.6 Large | AI21 Labs | 4.7 | - | - | - | - |
| Phi-4 | Microsoft | 4.6 | - | 45 t/s | 0.9s | $0.2 |
| LFM2 24B A2B | Liquid AI | 4.6 | - | - | - | - |
| Gemini 1.5 Flash (May '24) | 4.6 | - | - | - | - | |
| Nova Micro | Amazon | 4.4 | - | 273 t/s | 0.6s | $0.1 |
| Granite 4.1 3B | IBM | 4.4 | 4.7 | - | - | - |
| Claude 3 Sonnet | Anthropic | 4.4 | - | - | - | - |
| Gemini 1.0 Ultra | 4.3 | 17.6 | - | - | - | |
| Mistral Small (Sep '24) | Mistral | 4.3 | - | - | - | $0.3 |
| Phi-3 Mini Instruct 3.8B | Microsoft | 4.3 | - | - | - | - |
| Phi-4 Multimodal Instruct | Microsoft | 4.2 | - | 18 t/s | 0.4s | - |
| NVIDIA Nemotron Nano 12B v2 VL (Non-reasoning) | NVIDIA | 4.2 | - | 164 t/s | 1.3s | $0.3 |
| Gemma 3n E4B Instruct Preview (May '25) | 4.2 | - | - | - | - | |
| Mistral Large (Feb '24) | Mistral | 4.1 | - | - | - | $6.0 |
| Qwen2.5 Coder Instruct 7B | Alibaba | 4.1 | - | - | - | - |
| Mixtral 8x22B Instruct | Mistral | 4.0 | - | - | - | - |
| Llama 3.2 Instruct 3B | Meta | 3.9 | - | - | - | - |
| Llama 2 Chat 7B | Meta | 3.9 | - | - | - | $0.1 |
| MiniCPM-V 4.6 1.3B | OpenBMB | 3.8 | 0.7 | - | - | - |
| Jamba Reasoning 3B | AI21 Labs | 3.8 | - | - | - | - |
| Reka Flash 3 | Reka AI | 3.7 | - | 90 t/s | 1.1s | $0.3 |
| Qwen1.5 Chat 110B | Alibaba | 3.7 | - | - | - | - |
| Qwen3 VL 4B Instruct | Alibaba | 3.7 | - | - | - | - |
| Olmo 3 7B Think | Allen Institute for AI | 3.6 | - | - | - | - |
| Claude 3 Haiku | Anthropic | 3.5 | - | - | - | $0.5 |
| Claude 2.1 | Anthropic | 3.5 | 14.0 | - | - | - |
| OLMo 2 7B | Allen Institute for AI | 3.5 | - | - | - | - |
| Molmo 7B-D | Allen Institute for AI | 3.4 | - | - | - | - |
| Ling-mini-2.0 | InclusionAI | 3.4 | - | - | - | - |
| Claude 2.0 | Anthropic | 3.3 | 12.9 | - | - | - |
| DeepSeek R1 Distill Qwen 1.5B | DeepSeek | 3.3 | - | - | - | - |
| DeepSeek-V2-Chat | DeepSeek | 3.3 | - | - | - | - |
| GPT-3.5 Turbo | OpenAI | 3.2 | 10.7 | - | - | $0.8 |
| Mistral Small (Feb '24) | Mistral | 3.2 | - | - | - | $0.3 |
| Mistral Medium | Mistral | 3.2 | - | - | - | $3.0 |
| Llama 3 Instruct 70B | Meta | 3.1 | - | - | - | $1.2 |
| Llama 3.2 Instruct 11B (Vision) | Meta | 3.0 | - | 65 t/s | 0.4s | $0.3 |
| LFM 40B | Liquid AI | 3.0 | - | - | - | - |
 Arctic Instruct | Snowflake | 3.0 | - | - | - | - |
| Qwen Chat 72B | Alibaba | 3.0 | - | - | - | - |
| Qwen3.5 0.8B (Non-reasoning) | Alibaba | 2.9 | 1.2 | - | - | - |
| PALM-2 | 2.8 | 4.6 | - | - | - | |
| Gemini 1.0 Pro | 2.7 | - | - | - | - | |
| DeepSeek Coder V2 Lite Instruct | DeepSeek | 2.7 | - | - | - | - |
| Llama 2 Chat 70B | Meta | 2.6 | - | - | - | - |
| Llama 2 Chat 13B | Meta | 2.6 | - | - | - | - |
| DeepSeek LLM 67B Chat (V1) | DeepSeek | 2.6 | - | - | - | - |
 OpenChat 3.5 (1210) | OpenChat | 2.6 | - | - | - | - |
 DBRX Instruct | Databricks | 2.6 | - | - | - | - |
| Sarvam M (Reasoning) | Sarvam | 2.6 | - | - | - | - |
| Command-R+ (Apr '24) | Cohere | 2.6 | - | - | - | $6.0 |
| Exaone 4.0 1.2B (Reasoning) | LG AI Research | 2.5 | - | - | - | - |
| Olmo 3 7B Instruct | Allen Institute for AI | 2.4 | - | - | - | $0.1 |
| Exaone 4.0 1.2B (Non-reasoning) | LG AI Research | 2.4 | - | - | - | - |
| LFM2 2.6B | Liquid AI | 2.3 | - | - | - | - |
| LFM2.5-1.2B-Thinking | Liquid AI | 2.3 | - | - | - | - |
| LFM2.5-1.2B-Instruct | Liquid AI | 2.3 | - | - | - | - |
| Jamba 1.7 Mini | AI21 Labs | 2.3 | - | - | - | - |
| Jamba 1.5 Mini | AI21 Labs | 2.3 | - | - | - | $0.3 |
| Granite 4.0 H 1B | IBM | 2.2 | - | - | - | - |
| Qwen3 1.7B (Reasoning) | Alibaba | 2.2 | - | - | - | - |
| Jamba 1.6 Mini | AI21 Labs | 2.1 | - | - | - | - |
| Gemma 3 270M | 2.0 | - | - | - | - | |
| Granite 4.0 Micro | IBM | 2.0 | - | - | - | - |
| Apertus 70B Instruct | Swiss AI Initiative | 2.0 | - | - | - | $1.3 |
| Mixtral 8x7B Instruct | Mistral | 2.0 | - | - | - | $0.5 |
| DeepHermes 3 - Llama-3.1 8B Preview (Non-reasoning) | Nous Research | 1.9 | - | - | - | - |
| Llama 65B | Meta | 1.7 | - | - | - | - |
| Qwen Chat 14B | Alibaba | 1.7 | - | - | - | - |
| Claude Instant | Anthropic | 1.7 | 7.8 | - | - | - |
| Mistral 7B Instruct | Mistral | 1.7 | - | - | - | $0.3 |
| Command-R (Mar '24) | Cohere | 1.7 | - | - | - | $0.8 |
| Molmo2-8B | Allen Institute for AI | 1.6 | - | - | - | - |
| Granite 4.0 1B | IBM | 1.6 | - | - | - | - |
| LFM2 8B A1B | Liquid AI | 1.3 | - | - | - | - |
| Granite 3.3 8B (Non-reasoning) | IBM | 1.3 | - | - | - | $0.1 |
| Qwen3 1.7B (Non-reasoning) | Alibaba | 1.1 | - | - | - | - |
| LFM2.5-VL-1.6B | Liquid AI | 1.0 | - | 359 t/s | 1.1s | - |
| Granite 4.0 H 350M | IBM | 1.0 | - | - | - | - |
| Granite 4.0 350M | IBM | 1.0 | - | - | - | - |
| Apertus 8B Instruct | Swiss AI Initiative | 1.0 | - | - | - | $0.1 |
| Tiny Aya Global | Cohere | 1.0 | - | 132 t/s | 0.3s | - |
| Llama 3 Instruct 8B | Meta | 1.0 | - | - | - | $0.1 |
| Llama 3.2 Instruct 1B | Meta | 1.0 | - | - | - | - |
| Gemma 3n E2B Instruct | 1.0 | - | - | - | - | |
| Gemma 3 1B Instruct | 1.0 | - | - | - | - | |
| Gemma 3n E4B Instruct | 1.0 | 3.2 | - | - | $0.1 | |
| Gemma 3 4B Instruct | 1.0 | 2.7 | - | - | - | |
| LFM2 1.2B | Liquid AI | 1.0 | - | - | - | - |
| Qwen3 0.6B (Non-reasoning) | Alibaba | 1.0 | - | - | - | - |
| Qwen3 0.6B (Reasoning) | Alibaba | 1.0 | - | - | - | - |
| Claude Sonnet 5 (Adaptive Reasoning, Low Effort) | Anthropic | - | - | 60 t/s | 1.6s | $4.0 |
| Claude Sonnet 5 (Adaptive Reasoning, Medium Effort) | Anthropic | - | - | 61 t/s | 1.9s | $4.0 |
| Claude Sonnet 5 (Adaptive Reasoning, High Effort) | Anthropic | - | - | 60 t/s | 6.5s | $4.0 |
| Claude Sonnet 5 (Adaptive Reasoning, Xhigh Effort) | Anthropic | - | - | 73 t/s | 17.5s | $4.0 |
数据来源：Artificial Analysis · 每日自动更新 · 仅供参考
