AI 大语言模型排行榜（Artificial Analysis Intelligence Index）
数据来源：Artificial Analysis — Model Leaderboards
以下内容根据提供的榜单文本整理。保留原始排序、模型名称、上下文窗口、开发商、Artificial Analysis Intelligence Index、每任务成本、中位数生成速度及端到端响应时间。
--表示原始数据未提供；带*的指数分值按原文保留。
| 排名 | 模型 | 上下文窗口 | 开发商 | Intelligence Index | 每任务成本（美元） | 中位数速度（Token/秒） | 首个分块延迟（秒） | 总响应时间（秒） |
|---|---|---|---|---|---|---|---|---|
| 1 | Claude Opus 5.5 (max with fallback) | 1M | Anthropic | 58 | $5.98 | 96 | 562.48 | 567.67 |
| 2 | Claude Sonnet 5.5 (max with fallback) | 1M | Anthropic | 56 | $5.46 | 133 | 343.12 | 346.89 |
| 3 | Claude Opus 5.5 (xhigh with fallback) | 1M | Anthropic | 56 | $3.46 | 85 | 97.91 | 103.80 |
| 4 | Muse Wonder Full (max) | 1M | Meta | 56 | -- | -- | -- | -- |
| 5 | Kimi K3.1 Pro | 1M | Kimi | 55 | $2.11 | 73 | 2.97 | 37.00 |
| 6 | Claude Opus 5.5 (high with fallback) | 1M | Anthropic | 54 | $1.82 | 81 | 21.55 | 27.75 |
| 7 | Claude Fable 5.1 (max with fallback) | 1M | Anthropic | 53 | $7.63 | 68 | 163.20 | 170.50 |
| 8 | Claude Fable 5.1 (xhigh with fallback) | 1M | Anthropic | 53 | $5.98 | 65 | 60.95 | 68.67 |
| 9 | GPT-6 Astra (max) | 1M | OpenAI | 53 | $3.26 | 58 | 246.99 | 255.59 |
| 10 | Gemini 4 Argon (high) | 1M | 53 | $1.99 | 79 | 57.01 | 63.33 | |
| 11 | Muse Wonder (max) | 1.05M | Meta | 52 | $4.54 | 145 | 26.53 | 43.75 |
| 12 | GPT-6 Astra (xhigh) | 1M | OpenAI | 52 | $2.31 | 54 | 132.31 | 141.52 |
| 13 | Claude Sonnet 5.5 (xhigh with fallback) | 1M | Anthropic | 52 | $2.01 | 111 | 26.07 | 30.57 |
| 14 | GPT-6.1 Sol (max) | 1M | OpenAI | 52 | $0.72 | 57 | 162.10 | 170.82 |
| 15 | Claude Opus 5.5 (medium with fallback) | 1M | Anthropic | 51 | $1.34 | 84 | 10.44 | 16.41 |
| 16 | Claude Fable 5.1 (high with fallback) | 1M | Anthropic | 51 | $3.91 | 60 | 25.55 | 33.94 |
| 17 | GPT-6.1 Sol (xhigh) | 1M | OpenAI | 51 | $0.39 | 51 | 60.09 | 69.83 |
| 18 | GPT-6 Astra (high) | 1M | OpenAI | 51 | $1.73 | 57 | 32.11 | 40.94 |
| 19 | GPT-6.1 Sol (high) | 1M | OpenAI | 50 | $0.32 | 51 | 31.33 | 41.18 |
| 20 | GPT-6 Astra (medium) | 1M | OpenAI | 50 | $1.54 | 55 | 4.04 | 13.10 |
| 21 | Claude Fable 5.1 (medium with fallback) | 1M | Anthropic | 49 | $2.98 | 59 | 3.76 | 12.28 |
| 22 | GPT-6.1 Sol (medium) | 1M | OpenAI | 48 | $0.21 | 49 | 3.96 | 14.25 |
| 23 | Claude Fable 5.1 (low with fallback) | 1M | Anthropic | 47 | $2.37 | 60 | 0.98 | 9.31 |
| 24 | Claude Sonnet 5.5 (high with fallback) | 1M | Anthropic | 47 | $0.88 | 108 | 10.34 | 14.98 |
| 25 | Grok 4.7 (xhigh) | 500k | SpaceXAI | 46 | $3.74 | 97 | 34.82 | 39.99 |
| 26 | Grok 4.7 (high) | 500k | SpaceXAI | 46 | $2.73 | 91 | 18.67 | 24.17 |
| 27 | MiMo-V2.6-Pro | 1M | Xiaomi | 46 | $0.13 | 55 | 2.66 | 48.33 |
| 28 | GPT-6 Astra (low) | 1M | OpenAI | 46 | $0.82 | 53 | 2.22 | 11.58 |
| 29 | Qwen3.8 Max (0902) | 984k | Alibaba | 45 | $5.41 | 42 | 2.99 | 62.80 |
| 30 | DeepSeek V4.1 Flash 1017 (max) | 1M | DeepSeek | 45 | $0.26 | 195 | 1.14 | 13.95 |
| 31 | GLM-5.3 (max) | 1M | Z AI | 45 | $2.01 | 85 | 2.86 | 32.29 |
| 32 | Ling 3.1 Flash 1019 | 1.05M | InclusionAI | 44 | $1.49 | -- | -- | -- |
| 33 | Step 5 Preview | 1M | StepFun | 44 | $1.03 | 229 | 1.16 | 12.08 |
| 34 | Claude Haiku 5.5 (max) | 1M | Anthropic | 43 | $0.21 | 236 | 301.64 | 303.76 |
| 35 | MiniMax-M3.1-Flash | 1M | MiniMax | 43 | $0.22 | 96 | 1.32 | 27.48 |
| 36 | Claude Opus 5.5 (low with fallback) | 1M | Anthropic | 42 | $0.55 | 81 | 6.05 | 12.21 |
| 37 | Grok 4.7 (low) | 500k | SpaceXAI | 42 | $1.25 | 78 | 4.77 | 11.17 |
| 38 | GPT-6.1 Sol (low) | 1M | OpenAI | 42 | $0.13 | 48 | 1.78 | 12.24 |
| 39 | GPT-5.6 Terra (max) | 1M | OpenAI | 42 | $1.40 | 126 | 101.00 | 104.98 |
| 40 | GLM-5.3-Flash | 1M | Z AI | 42 | $0.28 | 69 | 3.29 | 39.60 |
| 41 | Claude Haiku 5.5 (xhigh) | 1M | Anthropic | 41 | $0.12 | 190 | 55.53 | 58.16 |
| 42 | Ling 3.1 Flash | 1M | InclusionAI | 41 | $1.01 | 219 | 1.91 | 13.33 |
| 43 | Gemini 3.8 Flash (high) | 1M | 41 | $1.24 | 141 | 25.70 | 29.24 | |
| 44 | Claude Sonnet 5.5 (medium with fallback) | 1M | Anthropic | 41 | $0.48 | 104 | 1.39 | 6.20 |
| 45 | Qwen3.8 2.4T A95B | 262k | Alibaba | 40 | $2.16 | 43 | 2.94 | 61.07 |
| 46 | Qwen3.8-Flash-Next | 256k | Alibaba | 40 | $0.37 | 52 | 2.16 | 50.69 |
| 47 | Gemini 3.8 Flash (medium) | 1M | 40 | $0.93 | -- | -- | -- | |
| 48 | Muse Spark Open (xhigh) | 1.05M | Meta | 39 | $0.84 | 151 | 0.77 | 17.34 |
| 49 | Mistral Large 4 Preview | 524k | Mistral | 38 | $1.13 | 95 | 3.14 | 29.59 |
| 50 | GPT-6 Luna (max) | 1M | OpenAI | 38 | $0.07 | 137 | 64.91 | 68.56 |
| 51 | GPT-5.6 Terra (xhigh) | 1M | OpenAI | 38 | $0.63 | 112 | 7.25 | 11.72 |
| 52 | MiMo-V2.6-Flash | 1M | Xiaomi | 38 | $0.06 | 59 | 4.12 | 46.19 |
| 53 | Claude Haiku 5.5 (high) | 1M | Anthropic | 38 | $0.08 | 189 | 17.81 | 20.45 |
| 54 | DeepSeek V4 Pro 0813 (max) | 1M | DeepSeek | 36 | $0.67 | 97 | 1.65 | 27.41 |
| 55 | Claude Sonnet 5.5 (low with fallback) | 1M | Anthropic | 36 | $0.35 | 104 | 1.40 | 6.21 |
| 56 | Agnes 2.5 Pro Beta | 1M | Sapiens AI | 35* | -- | 113 | 1.40 | 23.51 |
| 57 | DeepSeek V4 Flash Vision (max) | 1M | DeepSeek | 35 | $1.04 | 190 | 1.51 | 14.68 |
| 58 | GPT-6 Luna (xhigh) | 1M | OpenAI | 35 | $0.04 | 126 | 15.29 | 19.26 |
| 59 | Claude Haiku 5.5 (medium) | 1M | Anthropic | 34 | $0.05 | 180 | 7.71 | 10.49 |
| 60 | GLM-5.3 (low) | 1M | Z AI | 34 | $0.85 | 89 | 3.10 | 31.14 |
| 61 | GPT-5.6 Terra (high) | 1M | OpenAI | 34 | $0.34 | 116 | 2.15 | 6.48 |
| 62 | JT-4.1 Flash 236B A21B | 262k | China Mobile | 34 | -- | -- | -- | -- |
| 63 | Qwen3.8 27B (xhigh) | 256k | Alibaba | 34 | $1.01 | 44 | 3.88 | 61.31 |
| 64 | Motif 3 | 262k | Motif Technologies | 34* | -- | -- | -- | -- |
| 65 | Gemini 3.8 Flash (low) | 1M | 33 | -- | -- | -- | -- | |
| 66 | GPT-6 Luna (high) | 1M | OpenAI | 33 | $0.03 | 123 | 6.11 | 10.17 |
| 67 | GPT-5.3 Codex (xhigh) | 400k | OpenAI | 33* | -- | 87 | 60.62 | 66.39 |
| 68 | K2 Horizon 375B A23B | 524k | Institute of Foundation Models | 31 | $0.00 | 909 | 2.09 | 4.84 |
| 69 | Beam 1.0 | 1.05M | Reflection | 30 | $0.23 | 337 | 0.80 | 8.23 |
| 70 | GPT-5.6 Terra (medium) | 1M | OpenAI | 30 | $0.18 | 109 | 1.60 | 6.18 |
| 71 | GPT-6 Luna (medium) | 1M | OpenAI | 30 | $0.02 | 92 | 2.79 | 8.25 |
| 72 | Gemini 3.1 Pro Preview | 1M | 30 | $1.30 | 132 | 21.06 | 24.85 | |
| 73 | Claude Haiku 5.5 (low) | 1M | Anthropic | 29 | $0.02 | 176 | 6.68 | 9.52 |
| 74 | MiniMax-M3 | 1M | MiniMax | 29 | $0.51 | 89 | 1.30 | 29.48 |
| 75 | Nex-N2-Pro | 262k | Nex AGI | 28* | -- | -- | -- | -- |
| 76 | Qwen3.8 27B (medium) | 256k | Alibaba | 28 | $1.13 | 50 | 3.94 | 54.17 |
| 77 | GPT-5.6 Terra (low) | 1M | OpenAI | 27 | $0.14 | 114 | 1.39 | 5.76 |
| 78 | JT-4.1 Flash 236B A21B (non-reasoning) | 262k | China Mobile | 27* | -- | -- | -- | -- |
| 79 | Quasar 438B (max) | 1M | Multiverse Computing | 27 | -- | 172 | 1.29 | 15.86 |
| 80 | Solar Pro 4 | 512k | Upstage | 27 | $0.71 | -- | -- | -- |
| 81 | Nemotron 3.5 Super VL | 1M | NVIDIA | 26 | $0.19 | 453 | 0.71 | 6.23 |
| 82 | Apodex 1.1 | 256k | Apodex | 26 | $0.46 | -- | -- | -- |
| 83 | Qwen3.8 27B (low) | 256k | Alibaba | 26 | $1.05 | 51 | 3.97 | 52.75 |
| 84 | GPT-5.5 Instant (June 2026) | 400k | OpenAI | 26 | $0.69 | 155 | 0.89 | 17.04 |
| 85 | Kimi K2.7 Code | 256k | Kimi | 26 | $0.54 | 60 | 2.84 | 48.35 |
| 86 | Inkling Small | 1M | Thinking Machines | 26 | $0.09 | 224 | 1.62 | 12.78 |
| 87 | K2 Horizon MoVA 36B A4B | 524k | Institute of Foundation Models | 25 | -- | -- | -- | -- |
| 88 | Hy3 | 256k | Tencent | 25 | $0.07 | 82 | 2.78 | 33.12 |
| 89 | Qwen3.7 Plus | 1M | Alibaba | 25 | $0.22 | 59 | 2.28 | 44.50 |
| 90 | Inkling (xhigh) | 1M | Thinking Machines | 25 | -- | 194 | 1.77 | 14.63 |
| 91 | Solar Open2 250B | 1.05M | Upstage | 25* | -- | -- | -- | -- |
| 92 | Ling-3.0-flash-VL | 262k | InclusionAI | 25 | -- | -- | -- | -- |
| 93 | Solar Mini 4 | 1.05M | Upstage | 24 | $0.37 | 95 | 1.85 | 28.10 |
| 94 | Nemotron 3 Ultra | 262k | NVIDIA | 23 | $0.47 | 154 | 1.24 | 19.24 |
| 95 | Ling-3.0-flash-Fin | 262k | InclusionAI | 23 | -- | -- | -- | -- |
| 96 | Gemini 3.5 Flash-Lite | 1M | 22 | $0.19 | 430 | 6.94 | 8.10 | |
| 97 | G9v3-39A5B | 131k | AI9Stars | 22* | -- | -- | -- | -- |
| 98 | KAT-Coder-Pro V2 | 256k | KwaiKAT | 22* | -- | -- | -- | -- |
| 99 | GPT-6 Luna (low) | 1M | OpenAI | 22 | $0.0045 | 111 | 2.14 | 6.63 |
| 100 | A.X-K2 | 262k | SK Telecom | 21 | -- | -- | -- | -- |
| 101 | Qwen3.5 397B A17B (non-reasoning) | 262k | Alibaba | 21* | -- | 59 | 2.11 | 10.56 |
| 102 | Nemotron 3.5 Bolt 74B-A9B | 262k | NVIDIA | 21 | -- | -- | -- | -- |
| 103 | GPT-5.6 Terra (non-reasoning) | 1M | OpenAI | 21 | $0.14 | 115 | 0.73 | 5.08 |
| 104 | K2 Horizon 7B | 524k | Institute of Foundation Models | 21 | -- | -- | -- | -- |
| 105 | Qwen3.5 Omni Plus | 256k | Alibaba | 20* | -- | 112 | 1.99 | 6.44 |
| 106 | DeepSeek V4 Pro 0813 (non-reasoning) | 1M | DeepSeek | 20 | $0.36 | 96 | 1.71 | 6.90 |
| 107 | o3 | 200k | OpenAI | 20* | -- | 134 | 7.39 | 11.12 |
| 108 | Qwen3.8 27B (non-reasoning) | 256k | Alibaba | 20 | $2.49 | 48 | 3.83 | 14.30 |
| 109 | K-EXAONE 2.0 | 262k | LG AI Research | 20* | -- | -- | -- | -- |
| 110 | LongCat 2.0 | 1M | LongCat | 19 | $0.15 | 77 | 2.72 | 35.27 |
| 111 | JT-35B-Flash | 256k | China Mobile | 19* | -- | -- | -- | -- |
| 112 | GPT-6 Luna (non-reasoning) | 1M | OpenAI | 18 | $0.01 | 111 | 0.89 | 5.39 |
| 113 | Qwen3.5 397B A17B | 262k | Alibaba | 18 | $0.47 | 58 | 2.11 | 65.33 |
| 114 | Qwen3.6 35B A3B | 262k | Alibaba | 18 | $0.48 | 136 | 1.98 | 45.20 |
| 115 | Qwen3.5 122B A10B (non-reasoning) | 262k | Alibaba | 18* | -- | 142 | 2.25 | 5.76 |
| 116 | Muse Glimmer (high) | 131k | Meta | 17 | $0.06 | 119 | 1.01 | 22.04 |
| 117 | Doubao Seed Code | 256k | ByteDance Seed | 17* | -- | -- | -- | -- |
| 118 | Gemma 4 26B A4B | 256k | 17* | -- | 81 | 0.95 | 31.64 | |
| 119 | Ring-2.6-1T | 262k | InclusionAI | 17 | $0.29 | 121 | 3.76 | 24.47 |
| 120 | K2 Horizon 3.7B | 524k | Institute of Foundation Models | 16 | -- | -- | -- | -- |
| 121 | Qwen3.5 122B A10B | 262k | Alibaba | 16 | $0.32 | 131 | 2.25 | 21.38 |
| 122 | Qwen3.6 35B A3B (non-reasoning) | 262k | Alibaba | 15* | -- | 156 | 1.97 | 5.18 |
| 123 | Gemma 4 31B | 256k | 15 | $0.00 | 36 | 1.08 | 63.28 | |
| 124 | ERNIE 5.0 Thinking Preview | 128k | Baidu | 14* | -- | -- | -- | -- |
| 125 | Mistral Medium 3.5 | 256k | Mistral | 14 | $0.50 | 159 | 2.25 | 17.96 |
| 126 | Gemma 4 12B | 262k | 14* | -- | 120 | 2.24 | 23.12 | |
| 127 | Nova 2.0 Pro Preview (medium) | 256k | Amazon | 14* | -- | 125 | 11.38 | 31.40 |
| 128 | Gemma 4 31B (non-reasoning) | 256k | 14* | -- | 52 | 1.70 | 11.41 | |
| 129 | Nova 2.0 Omni (medium) | 1M | Amazon | 14* | -- | -- | -- | -- |
| 130 | Apriel-v1.6-15B-Thinker | 128k | ServiceNow | 13* | -- | -- | -- | -- |
| 131 | Nova 2.0 Lite (high) | 1M | Amazon | 13* | -- | 182 | 16.05 | 29.78 |
| 132 | Qwen3.5 9B (non-reasoning) | 262k | Alibaba | 13* | -- | 78 | 0.81 | 7.20 |
| 133 | EXAONE 4.5 33B | 262k | LG AI Research | 13* | -- | -- | -- | -- |
| 134 | Command A+ | 192k | Cohere | 13 | $0.00 | 149 | 0.53 | 17.34 |
| 135 | Gemma 4 26B A4B (non-reasoning) | 256k | 13* | -- | 96 | 0.96 | 6.18 | |
| 136 | Qwen3.5 4B | 262k | Alibaba | 13* | -- | 17 | 0.86 | 149.52 |
| 137 | Nemotron 3.5 Lightning | 1M | NVIDIA | 13 | $0.08 | 288 | 0.64 | 9.33 |
| 138 | Nova 2.0 Pro Preview (low) | 256k | Amazon | 13* | -- | 119 | 10.64 | 31.60 |
| 139 | Granite 4.2 30B | 131k | IBM | 13 | $0.07 | 75 | 0.67 | 34.22 |
| 140 | Nova 2.0 Lite (medium) | 1M | Amazon | 12* | -- | 177 | 11.87 | 26.02 |
| 141 | Qwen3.5 Omni Flash | 256k | Alibaba | 12* | -- | 234 | 1.79 | 3.92 |
| 142 | MiniCPM5-2B | 131k | OpenBMB | 12 | -- | -- | -- | -- |
| 143 | Mercury 2.5 | 260k | Inception | 12 | $0.12 | 689 | 3.71 | 4.43 |
| 144 | JT-MINI | 128k | China Mobile | 12* | -- | -- | -- | -- |
| 145 | Magistral Medium 1.2 | 128k | Mistral | 12* | -- | -- | -- | -- |
| 146 | Nova 2.0 Lite (low) | 1M | Amazon | 12* | -- | 175 | 8.89 | 23.18 |
| 147 | HyperNova 60B 2605 (high) | 131k | Multiverse Computing | 12* | -- | -- | -- | -- |
| 148 | Nemotron Cascade 2 30B A3B | 1M | NVIDIA | 12* | -- | -- | -- | -- |
| 149 | gpt-oss-120b (high) | 131k | OpenAI | 12 | $0.11 | 208 | 0.82 | 12.84 |
| 150 | K2 Think V2 | 262k | Institute of Foundation Models | 11* | -- | -- | -- | -- |
| 151 | LongCat Flash Lite | 256k | LongCat | 11* | -- | -- | -- | -- |
| 152 | HyperCLOVA X SEED Think (32B) | 128k | Naver | 11* | -- | -- | -- | -- |
| 153 | Mistral Small 4 | 256k | Mistral | 11 | $0.02 | 183 | 0.90 | 14.55 |
| 154 | Qwen3 Next 80B A3B | 262k | Alibaba | 11* | -- | -- | -- | -- |
| 155 | Qwen3.5 9B | 262k | Alibaba | 11 | $0.21 | 59 | 1.93 | 44.43 |
| 156 | Nova 2.0 Omni (low) | 1M | Amazon | 11* | -- | -- | -- | -- |
| 157 | Granite 4.2 8B | 131k | IBM | 11 | $0.02 | 87 | 0.72 | 29.58 |
| 158 | Ling 3.0 Tiny | 262k | InclusionAI | 11 | $0.00 | 57 | 2.53 | 46.17 |
| 159 | Mi:dm K 2.5 Pro | 128k | Korea Telecom | 11* | -- | -- | -- | -- |
| 160 | G9v3-3B | 131k | AI9Stars | 11* | -- | -- | -- | -- |
| 161 | Trinity Large Thinking | 512k | Arcee AI | 11 | $0.12 | 357 | 1.25 | 8.25 |
| 162 | Qwen3.5 4B (non-reasoning) | 262k | Alibaba | 11* | -- | 11 | 1.05 | 46.01 |
| 163 | INTELLECT-3 | 131k | Prime Intellect | 11* | -- | -- | -- | -- |
| 164 | Solar Open 100B | 128k | Upstage | 10* | -- | -- | -- | -- |
| 165 | Nemotron 3 Nano Omni 30B A3B | 256k | NVIDIA | 10* | -- | 234 | 0.77 | 11.46 |
| 166 | gpt-oss-120b (low) | 131k | OpenAI | 10* | -- | 263 | 0.83 | 10.35 |
| 167 | Nova 2.0 Pro Preview (non-reasoning) | 256k | Amazon | 10* | -- | 111 | 0.98 | 5.48 |
| 168 | gpt-oss-20b (low) | 131k | OpenAI | 10* | -- | 175 | 0.87 | 15.14 |
| 169 | North Mini Code | 256k | Cohere | 10 | $0.00 | 60 | 0.45 | 41.94 |
| 170 | K2-V2 (high) | 512k | Institute of Foundation Models | 10* | -- | -- | -- | -- |
| 171 | Qwen3 Next 80B A3B | 262k | Alibaba | 10* | -- | 96 | 0.81 | 6.01 |
| 172 | DiffusionGemma 26B A4B | 256k | 10* | -- | -- | -- | -- | |
| 173 | Gemma 4 12B (non-reasoning) | 262k | 9* | -- | 127 | 2.22 | 6.14 | |
| 174 | Llama 4 Maverick | 1M | Meta | 9 | $0.04 | 52 | 1.06 | 10.65 |
| 175 | Qwen3 Coder Next | 256k | Alibaba | 9 | $0.47 | 88 | 1.05 | 6.76 |
| 176 | Motif-2-12.7B | 128k | Motif Technologies | 9* | -- | -- | -- | -- |
| 177 | Nova Premier | 1M | Amazon | 9* | -- | -- | -- | -- |
| 178 | Granite 4.2 3B | 131k | IBM | 9 | $0.01 | 99 | 0.77 | 25.92 |
| 179 | K2-V2 (medium) | 512k | Institute of Foundation Models | 9* | -- | -- | -- | -- |
| 180 | Mistral Small 4 (non-reasoning) | 256k | Mistral | 9* | -- | 171 | 0.92 | 3.84 |
| 181 | Tri-21B-Think | 32k | Trillion Labs | 9* | -- | -- | -- | -- |
| 182 | gpt-oss-20b (high) | 131k | OpenAI | 9 | $0.01 | 152 | 0.80 | 17.29 |
| 183 | Gemma 4 E4B | 128k | 9 | $0.01 | 41 | 0.94 | 62.58 | |
| 184 | Nemotron 3 Nano | 1M | NVIDIA | 9 | $0.02 | 235 | 0.93 | 11.59 |
| 185 | MiniCPM5-1B | 128k | OpenBMB | 9* | -- | -- | -- | -- |
| 186 | Sarvam 105B (high) | 128k | Sarvam | 9* | -- | 192 | 1.59 | 14.60 |
| 187 | Nova 2.0 Lite (non-reasoning) | 1M | Amazon | 9* | -- | 176 | 1.05 | 3.90 |
| 188 | MiniCPM5-1B (non-reasoning) | 128k | OpenBMB | 9* | -- | -- | -- | -- |
| 189 | Magistral Small 1.2 | 128k | Mistral | 9* | -- | 78 | 1.48 | 33.45 |
| 190 | Nanbeige4.1-3B | 256k | Nanbeige | 8* | -- | -- | -- | -- |
| 191 | LFM2.5-2.6B | 128k | Liquid AI | 8* | -- | -- | -- | -- |
| 192 | EXAONE 4.0 32B | 131k | LG AI Research | 8* | -- | -- | -- | -- |
| 193 | Nova 2.0 Omni (non-reasoning) | 1M | Amazon | 8* | -- | -- | -- | -- |
| 194 | Hermes 4 70B | 128k | Nous Research | 8* | -- | -- | -- | -- |
| 195 | Falcon-H1R-7B | 256k | TII UAE | 8* | -- | -- | -- | -- |
| 196 | Gemma 4 E2B | 128k | 8* | -- | -- | -- | -- | |
| 197 | Qwen3 Omni 30B A3B | 65.5k | Alibaba | 8* | -- | 102 | 1.84 | 26.24 |
| 198 | Step3 VL 10B | 65.5k | StepFun | 8* | -- | -- | -- | -- |
| 199 | Llama 3.3 70B | 128k | Meta | 8* | -- | 86 | 1.78 | 7.61 |
| 200 | Llama Nemotron Ultra | 128k | NVIDIA | 8* | -- | -- | -- | -- |
| 201 | ERNIE 4.5 300B A47B | 131k | Baidu | 8* | -- | -- | -- | -- |
| 202 | Hermes 4 405B | 128k | Nous Research | 7* | -- | 41 | 2.30 | 63.34 |
| 203 | NVIDIA Nemotron Nano 12B v2 VL | 128k | NVIDIA | 7* | -- | -- | -- | -- |
| 204 | Gemma 4 E4B (non-reasoning) | 128k | 7* | -- | 49 | 0.83 | 11.03 | |
| 205 | NVIDIA Nemotron Nano 9B V2 | 131k | NVIDIA | 7* | -- | 95 | 0.79 | 27.08 |
| 206 | Hermes 4 405B (non-reasoning) | 128k | Nous Research | 7* | -- | 40 | 2.28 | 14.72 |
| 207 | K2-V2 (low) | 512k | Institute of Foundation Models | 7* | -- | -- | -- | -- |
| 208 | Kimi Linear 48B A3B Instruct | 1M | Kimi | 7* | -- | -- | -- | -- |
| 209 | Llama 3.1 405B | 128k | Meta | 7* | -- | -- | -- | -- |
| 210 | LFM2.5-8B-A1B | 32.8k | Liquid AI | 7* | -- | -- | -- | -- |
| 211 | Ring-flash-2.0 | 128k | InclusionAI | 7* | -- | -- | -- | -- |
| 212 | Olmo 3.1 32B Think | 65.5k | Allen Institute for AI | 7* | -- | -- | -- | -- |
| 213 | Command A | 256k | Cohere | 7* | -- | 51 | 1.72 | 11.62 |
| 214 | Qwen3.5 2B | 262k | Alibaba | 7* | -- | -- | -- | -- |
| 215 | Nemotron 3 Nano (non-reasoning) | 1M | NVIDIA | 7* | -- | 268 | 0.81 | 2.67 |
| 216 | NVIDIA Nemotron Nano 9B V2 (non-reasoning) | 131k | NVIDIA | 7* | -- | 165 | 1.04 | 4.06 |
| 217 | Hermes 4 70B (non-reasoning) | 128k | Nous Research | 7* | -- | -- | -- | -- |
| 218 | Sarvam 30B (high) | 65.5k | Sarvam | 7* | -- | -- | -- | -- |
| 219 | Olmo 3.1 32B Instruct | 65.5k | Allen Institute for AI | 6* | -- | -- | -- | -- |
| 220 | Gemma 4 E2B (non-reasoning) | 128k | 6* | -- | -- | -- | -- | |
| 221 | Llama 4 Scout | 10M | Meta | 6 | $0.11 | 103 | 0.83 | 5.70 |
| 222 | R1 1776 | 128k | Perplexity | 6* | -- | -- | -- | -- |
| 223 | Llama 3.2 90B (Vision) | 128k | Meta | 6* | -- | -- | -- | -- |
| 224 | Celeris-1 | 131k | Celeris | 6 | $0.05 | 1,581 | 0.55 | 0.87 |
| 225 | Phi-4 Mini | 128k | Microsoft | 6* | -- | 46 | 0.82 | 11.78 |
| 226 | EXAONE 4.0 32B (non-reasoning) | 131k | LG AI Research | 6* | -- | -- | -- | -- |
| 227 | Qwen3.5 2B (non-reasoning) | 262k | Alibaba | 6* | -- | -- | -- | -- |
| 228 | Qwen3.5 0.8B | 262k | Alibaba | 6* | -- | -- | -- | -- |
| 229 | DeepHermes 3 - Mistral 24B (non-reasoning) | 32k | Nous Research | 6* | -- | -- | -- | -- |
| 230 | Jamba 1.7 Large | 256k | AI21 Labs | 6* | -- | -- | -- | -- |
| 231 | Granite 4.0 H Small | 128k | IBM | 6* | -- | -- | -- | -- |
| 232 | Ministral 3 14B | 256k | Mistral | 6 | $0.02 | 83 | 0.95 | 6.98 |
| 233 | Qwen3 Omni 30B A3B | 65.5k | Alibaba | 6* | -- | 96 | 1.78 | 6.97 |
| 234 | LFM2 24B A2B | 32.8k | Liquid AI | 6* | -- | -- | -- | -- |
| 235 | Phi-4 | 16k | Microsoft | 6* | -- | 38 | 2.67 | 15.83 |
| 236 | Nova Micro | 130k | Amazon | 6* | -- | 280 | 0.86 | 2.65 |
| 237 | NVIDIA Nemotron Nano 12B v2 VL (non-reasoning) | 128k | NVIDIA | 6* | -- | 202 | 1.12 | 3.59 |
| 238 | Phi-4 Multimodal | 128k | Microsoft | 6* | -- | -- | -- | -- |
| 239 | MiniCPM-V 4.6 1.3B | 262k | OpenBMB | 6* | -- | -- | -- | -- |
| 240 | Jamba Reasoning 3B | 262k | AI21 Labs | 6* | -- | -- | -- | -- |
| 241 | Reka Flash 3 | 128k | Reka AI | 6* | -- | 88 | 1.43 | 29.84 |
| 242 | Olmo 3 7B Think | 65.5k | Allen Institute for AI | 6* | -- | -- | -- | -- |
| 243 | Molmo 7B-D | 4.1k | Allen Institute for AI | 6* | -- | -- | -- | -- |
| 244 | Ministral 3 8B | 256k | Mistral | 5 | $0.01 | 113 | 0.86 | 5.28 |
| 245 | Llama 3.2 11B (Vision) | 128k | Meta | 5* | -- | 23 | 1.28 | 23.14 |
| 246 | Qwen3.5 0.8B (non-reasoning) | 262k | Alibaba | 5* | -- | -- | -- | -- |
| 247 | Exaone 4.0 1.2B | 64k | LG AI Research | 5* | -- | -- | -- | -- |
| 248 | Olmo 3 7B | 65.5k | Allen Institute for AI | 5* | -- | -- | -- | -- |
| 249 | Exaone 4.0 1.2B (non-reasoning) | 64k | LG AI Research | 5* | -- | -- | -- | -- |
| 250 | LFM2.5-1.2B-Thinking | 32k | Liquid AI | 5* | -- | -- | -- | -- |
| 251 | Jamba 1.7 Mini | 258k | AI21 Labs | 5* | -- | -- | -- | -- |
| 252 | LFM2.5-1.2B-Instruct | 32k | Liquid AI | 5* | -- | -- | -- | -- |
| 253 | Granite 4.0 H 1B | 128k | IBM | 5* | -- | -- | -- | -- |
| 254 | Gemma 3 270M | 32k | 5* | -- | -- | -- | -- | |
| 255 | Apertus 70B Instruct | 65.5k | Swiss AI Initiative | 5* | -- | -- | -- | -- |
| 256 | Granite 4.0 Micro | 128k | IBM | 5* | -- | -- | -- | -- |
| 257 | DeepHermes 3 - Llama-3.1 8B (non-reasoning) | 128k | Nous Research | 5* | -- | -- | -- | -- |
| 258 | Molmo2-8B | 36.9k | Allen Institute for AI | 5* | -- | -- | -- | -- |
| 259 | Ministral 3 3B | 256k | Mistral | 5 | $0.01 | 207 | 0.68 | 3.09 |
| 260 | LFM2.5-VL-1.6B | 32k | Liquid AI | 5* | -- | -- | -- | -- |
| 261 | Granite 4.0 350M | 32.8k | IBM | 5* | -- | -- | -- | -- |
| 262 | Tiny Aya Global | 8.19k | Cohere | 5* | -- | -- | -- | -- |
| 263 | Apertus 8B Instruct | 65.5k | Swiss AI Initiative | 5* | -- | 131 | 3.22 | 7.04 |
| 264 | Granite 4.0 H 350M | 32.8k | IBM | 5* | -- | -- | -- | -- |
| 265 | Nemotron 3 Nano 4B | 262k | NVIDIA | 5 | -- | -- | -- | -- |
| 266 | K2 Horizon 0.9B | 131k | Institute of Foundation Models | 3 | -- | -- | -- | -- |
| 267 | EXAONE 4.5 33B (non-reasoning) | 262k | LG AI Research | -- | -- | -- | -- | -- |
| 268 | Nemotron 3.5 Bolt E4B 4GB | 131k | NVIDIA | -- | -- | -- | -- | -- |
| 269 | GPT-6 Sol (Daybreak Blue, max) | 1.05M | OpenAI | -- | -- | -- | -- | -- |
| 270 | Gemini 3 Deep Think | 128k | -- | -- | -- | -- | -- | |
| 271 | GPT-5.5 Pro (xhigh) | 922k | OpenAI | -- | -- | -- | -- | -- |
| 272 | Cogito v2.1 | 128k | Deep Cogito | -- | -- | -- | -- | -- |
指标说明
- Intelligence Index：榜单列出的综合智能指数。
- 每任务成本：按原榜单显示的美元成本记录。
- 中位数速度：Token/秒。
- 首个分块延迟：首个输出分块的延迟，单位为秒。
- 总响应时间：端到端总响应时间，单位为秒。
- 上下文窗口：按原榜单的
k、M等标记保留。
模型条目数：272。
AI 视频编辑模型排行榜（Video Editing）
数据来源：Artificial Analysis — Video Editing Leaderboard
排名依据：Elo 评分。数据包括模型排名、评分误差、95% 置信区间、样本数、发布日期及 API 价格。
| 排名 | 模型 | Elo | 误差范围 | 95% 置信区间 | 样本数 | 发布日期 | API 价格（美元/分钟） |
|---|---|---|---|---|---|---|---|
| 1 | Wan 3.0 | 1198 | ±10 | 1188–1208 | 8,290 | 2026-08 | 12.00 |
| 2 | Dreamina Seedance 2.5 | 1155 | ±12 | 1143–1167 | 6,130 | 2026-07 | 40.82 |
| 3 | Gemini Omni Flash 1.1 | 1148 | ±11 | 1137–1159 | 7,316 | 2026-08 | 9.64 |
| 4 | MiniMax H3 (768p) | 1124 | ±10 | 1114–1134 | 8,005 | 2026-07 | 9.60 |
| 5 | HappyHorse-1.0 | 1090 | ±8 | 1082–1098 | 24,770 | 2026-04 | 27.04 |
| 6 | Dreamina Seedance 2.0 | 1082 | ±11 | 1071–1093 | 8,113 | 2026-03 | 27.42 |
| 7 | Wan 2.7 | 1077 | ±8 | 1069–1085 | 21,021 | 2026-04 | 16.90 |
| 8 | Aleph 2.0 | 1012 | ±8 | 1004–1020 | 21,351 | 2026-05 | 16.80 |
| 9 | Kling 3.0 Omni 1080p (Pro) | 1000 | 锚点 | 固定参照点 | 19,250 | 2026-02 | 10.91 |
说明： API 价格按原始榜单记录，单位为美元/分钟。实际成本可能因视频时长、分辨率、编辑模式及服务商计费规则而异。
AI 视频生成模型排行榜（图像转视频）
数据来源：Artificial Analysis — Image-to-Video Leaderboard
排名依据：Elo 评分。数据包括模型排名、评分误差、95% 置信区间、样本数、发布日期及 API 价格。
| 排名 | 模型 | Elo | 误差范围 | 95% 置信区间 | 样本数 | 发布日期 | API 价格（美元/分钟） |
|---|---|---|---|---|---|---|---|
| 1 | MiniMax H3 (768p) | 1140 | ±16 | 1124–1156 | 4,209 | 2026-07 | 4.80 |
| 2 | FLUX 3 Max (720p) | 1138 | ±15 | 1123–1153 | 5,054 | 2026-10 | 2.55 |
| 3 | Dreamina Seedance 2.5 | 1138 | ±17 | 1121–1155 | 3,688 | 2026-07 | 34.12 |
| 4 | Gemini Omni Flash 1.1 | 1132 | ±17 | 1115–1149 | 3,368 | 2026-08 | 9.12 |
| 5 | Wan 3.0 | 1127 | ±16 | 1111–1143 | 5,144 | 2026-08 | 12.00 |
| 6 | MiniMax H3 Max | 1125 | ±17 | 1108–1142 | 3,476 | 2026-08 | 5.76 |
| 7 | Dreamina Seedance 2.0 | 1121 | ±18 | 1103–1139 | 2,799 | 2026-03 | 22.45 |
| 8 | FLUX 3 | 1086 | ±16 | 1070–1102 | 4,170 | 2026-07 | 17.40 |
| 9 | Grok Imagine Video 1.5 | 1067 | ±17 | 1050–1084 | 3,463 | 2026-05 | 15.00 |
| 10 | HappyHorse-1.1 | 1061 | ±17 | 1044–1078 | 4,597 | 2026-06 | 10.80 |
| 11 | MAGI-2 Preview | 1046 | ±16 | 1030–1062 | 4,153 | 2026-08 | 即将推出 |
| 12 | HappyHorse-1.0 | 1042 | ±17 | 1025–1059 | 3,501 | 2026-04 | 14.40 |
| 13 | Veo 3.1 | 1028 | ±17 | 1011–1045 | 3,499 | 2026-01 | 24.00 |
| 14 | PixVerse V6 | 1024 | ±17 | 1007–1041 | 3,451 | 2026-03 | 6.90 |
| 15 | Veo 3.1 Fast | 1017 | ±15 | 1002–1032 | 5,042 | 2026-01 | 7.20 |
| 16 | Vidu Q3 Pro | 1016 | ±15 | 1001–1031 | 4,953 | 2026-01 | 7.20 |
| 17 | Kling 3.0 Omni 1080p (Pro) | 1004 | ±17 | 987–1021 | 3,077 | 2026-02 | 8.40 |
| 18 | Kling 3.0 1080p (Pro) | 1000 | 锚点 | 固定参照点 | 5,211 | 2026-02 | 10.08 |
| 19 | LTX-2.5 Fast | 984 | ±19 | 965–1003 | 2,322 | 2026-08 | 7.80 |
| 20 | Vidu Q3 Turbo | 967 | ±19 | 948–986 | 2,439 | 2026-02 | 3.90 |
| 21 | LTX-2.5 Pro | 958 | ±19 | 939–977 | 2,357 | 2026-08 | 10.20 |
| 22 | LTX-2.3 Pro | 903 | ±18 | 885–921 | 3,126 | 2026-03 | 4.80 |
| 23 | LTX-2.3 Fast | 888 | ±18 | 870–906 | 3,165 | 2026-03 | 3.60 |
说明： API 价格按原始榜单记录，单位为美元/分钟。实际成本可能受分辨率、视频时长、生成模式和服务商计费规则影响。
AI 图像编辑模型排行榜（Image Editing）
数据来源：Artificial Analysis — Image Editing Leaderboard
排名依据：Elo 评分。数据包括模型排名、评分误差、95% 置信区间、样本数、发布日期及 API 价格。
| 排名 | 模型 | Elo | 误差范围 | 95% 置信区间 | 样本数 | 发布日期 | API 价格（美元/1000 张） |
|---|---|---|---|---|---|---|---|
| 1 | GPT Image 2.5 Sunburst (max) | 1199 | ±11 | 1188–1210 | 18,250 | 2026-09 | 210.7 |
| 2 | GPT Image 2.5 Flare (max) | 1182 | ±11 | 1171–1193 | 19,136 | 2026-09 | 210.7 |
| 3 | MAI-Image-2.6 | 1151 | ±12 | 1139–1163 | 16,644 | 2026-08 | 38.9 |
| 4 | Krea 3 (Ultra) | 1151 | ±14 | 1137–1165 | 5,505 | 2026-10 | 270.0 |
| 5 | Nano Banana 2.1 | 1151 | ±13 | 1138–1164 | 6,868 | 2026-10 | 33.6 |
| 6 | MAI-Image-2.6-Flash | 1144 | ±13 | 1131–1157 | 10,753 | 2026-09 | 19.5 |
| 7 | GPT Image 2 (high) | 1136 | ±13 | 1123–1149 | 14,570 | 2026-04 | 211.0 |
| 8 | Seedream 5.0 Pro | 1131 | ±13 | 1118–1144 | 13,380 | 2026-07 | 90.0 |
| 9 | Muse Image | 1129 | ±12 | 1117–1141 | 13,358 | 2026-07 | 10.0 |
| 10 | MAI-Image-2.5 | 1127 | ±13 | 1114–1140 | 13,545 | 2026-06 | 48.1 |
| 11 | Krea 3 (Fast) | 1126 | ±14 | 1112–1140 | 5,203 | 2026-10 | 70.0 |
| 12 | Nano Banana 2 (Gemini 3.1 Flash Image) | 1123 | ±12 | 1111–1135 | 16,658 | 2026-02 | 67.0 |
| 13 | MAI-Image-2.5-Pro | 1123 | ±12 | 1111–1135 | 15,654 | 2026-07 | 108.5 |
| 14 | FLUX 3 Image | 1122 | ±14 | 1108–1136 | 4,900 | 2026-10 | 48.0 |
| 15 | GPT Image 1.5 (high) | 1119 | ±13 | 1106–1132 | 13,507 | 2025-12 | 133.0 |
| 16 | Grok Imagine Image 2.0 | 1117 | ±13 | 1104–1130 | 8,692 | 2026-08 | 60.0 |
| 17 | Nano Banana Pro (Gemini 3 Pro Image) | 1108 | ±12 | 1096–1120 | 15,558 | 2025-11 | 134.0 |
| 18 | grok-imagine-image-quality | 1095 | ±12 | 1083–1107 | 8,787 | 2026-05 | 50.0 |
| 19 | HunyuanImage 3.5 (Preview) | 1093 | ±13 | 1080–1106 | 6,405 | 2026-09 | 24.0 |
| 20 | Qwen-Image-3.0-Pro | 1091 | ±13 | 1078–1104 | 13,533 | 2026-07 | 40.0 |
| 21 | MAI-Image-2.5-Flash | 1091 | ±13 | 1078–1104 | 13,364 | 2026-06 | 20.0 |
| 22 | Fibo 2.0 | 1088 | ±14 | 1074–1102 | 5,009 | 2026-10 | 32.0 |
| 23 | Qwen-Image-2.1 | 1081 | ±13 | 1068–1094 | 6,409 | 2026-09 | 暂无 API |
| 24 | Luma UNI 1 Max | 1079 | ±13 | 1066–1092 | 6,473 | 2026-05 | 100.0 |
| 25 | grok-imagine-image | 1073 | ±13 | 1060–1086 | 6,100 | 2026-01 | 20.0 |
| 26 | FLUX 3 Image Klein | 1072 | ±14 | 1058–1086 | 4,904 | 2026-10 | 9.0 |
| 27 | Ideogram 4.5 (High) | 1070 | ±15 | 1055–1085 | 3,325 | 2026-09 | 220.0 |
| 28 | Qwen-Image-2.1-Pro | 1069 | ±13 | 1056–1082 | 5,382 | 2026-10 | 35.3 |
| 29 | Qwen-Image-3.0 | 1068 | ±13 | 1055–1081 | 8,072 | 2026-07 | 30.0 |
| 30 | Qwen-Image-2.1-Turbo | 1064 | ±14 | 1050–1078 | 5,023 | 2026-10 | 16.0 |
| 31 | HunyuanImage 3.0 Instruct | 1063 | ±13 | 1050–1076 | 6,123 | 2026-01 | 暂无 API |
| 32 | Seedream 5.0 Lite | 1062 | ±13 | 1049–1075 | 6,471 | 2026-02 | 35.0 |
| 33 | Qwen Image 2.0 Pro (2026-04-22) | 1058 | ±17 | 1041–1075 | 2,888 | 2026-04 | 75.0 |
| 34 | Nano Banana 2 Lite (Gemini 3.1 Flash Lite Image) | 1054 | ±12 | 1042–1066 | 8,104 | 2026-06 | 33.6 |
| 35 | Wan 2.7 Pro | 1050 | ±13 | 1037–1063 | 6,331 | 2026-04 | 75.0 |
| 36 | DreamImage 2.0 | 1048 | ±13 | 1035–1061 | 6,249 | 2026-07 | 80.0 |
| 37 | Wan 2.6 Image | 1044 | ±13 | 1031–1057 | 6,267 | 2025-12 | 30.0 |
| 38 | Seedream 4.5 | 1043 | ±13 | 1030–1056 | 5,473 | 2025-12 | 40.0 |
| 39 | Luma UNI 1 | 1043 | ±13 | 1030–1056 | 9,459 | 2026-05 | 40.4 |
| 40 | Wan 2.7 | 1042 | ±13 | 1029–1055 | 7,613 | 2026-04 | 26.0 |
| 41 | HunyuanImage 3.0 (Tencent Cloud) | 1029 | ±16 | 1013–1045 | 3,093 | 2025-09 | 32.0 |
| 42 | FLUX.2 [dev] Turbo | 1019 | ±18 | 1001–1037 | 2,545 | 2025-12 | 8.0 |
| 43 | FLUX.2 [klein] 9B | 1015 | ±13 | 1002–1028 | 7,176 | 2026-01 | 17.0 |
| 44 | FLUX.2 [dev] Flash | 1012 | ±18 | 994–1030 | 2,593 | 2025-12 | 5.0 |
| 45 | FLUX.2 [pro] | 1012 | ±15 | 997–1027 | 4,229 | 2025-11 | 45.0 |
| 46 | Step Image Edit 2 | 1009 | ±13 | 996–1022 | 5,496 | 2026-05 | 3.0 |
| 47 | FLUX.2 [dev] | 1000 | 锚点 | 固定参照点 | 9,012 | 2025-11 | 24.0 |
| 48 | FLUX.2 [flex] | 999 | ±17 | 982–1016 | 2,739 | 2025-11 | 120.0 |
| 49 | FLUX.2 [max] | 999 | ±13 | 986–1012 | 6,940 | 2025-12 | 140.0 |
| 50 | GPT Image 1 (high) | 995 | ±18 | 977–1013 | 2,619 | 2025-04 | 167.0 |
| 51 | Qwen Image Edit Plus 2511 | 991 | ±18 | 973–1009 | 2,724 | 2025-11 | 60.0 |
| 52 | Reve V1 (December) | 987 | ±18 | 969–1005 | 2,797 | 2025-12 | 40.0 |
| 53 | P-Image-Edit | 985 | ±14 | 971–999 | 5,157 | 2025-11 | 10.0 |
| 54 | Nano Banana (Gemini 2.5 Flash Image) | 984 | ±14 | 970–998 | 5,213 | 2025-08 | 39.0 |
| 55 | Qwen Image Edit Max 2601 | 982 | ±18 | 964–1000 | 2,789 | 2026-01 | 75.0 |
| 56 | Qwen Image 2.0 (2026-03-03) | 966 | ±18 | 948–984 | 2,767 | 2026-03 | 35.0 |
| 57 | Qwen Image Edit Plus 2509 | 965 | ±18 | 947–983 | 2,587 | 2025-09 | 30.0 |
| 58 | HiDream-O1-Image | 954 | ±14 | 940–968 | 5,193 | 2026-05 | 10.0 |
| 59 | Kling Image 3.0 Omni | 952 | ±14 | 938–966 | 5,180 | 2026-02 | 28.0 |
| 60 | Vidu Q2 | 951 | ±18 | 933–969 | 2,612 | 2025-11 | 40.0 |
| 61 | Kling Image O1 | 944 | ±14 | 930–958 | 5,098 | 2025-12 | 28.0 |
| 62 | FLUX.2 [klein] 4B | 938 | ±17 | 921–955 | 3,635 | 2026-01 | 15.0 |
| 63 | GPT Image 1 Mini (medium) | 934 | ±19 | 915–953 | 2,428 | 2025-10 | 11.0 |
| 64 | FLUX.1 Kontext [max] | 932 | ±19 | 913–951 | 2,622 | 2025-05 | 80.0 |
| 65 | FLUX.1 Kontext [pro] | 916 | ±19 | 897–935 | 2,511 | 2025-05 | 40.0 |
| 66 | FIBO Edit | 915 | ±19 | 896–934 | 2,591 | 2026-01 | 40.0 |
| 67 | SeedEdit 3.0 | 901 | ±20 | 881–921 | 2,515 | 2025-06 | 0.0 |
| 68 | LongCat Image | 897 | ±20 | 877–917 | 2,472 | 2025-12 | 130.0 |
| 69 | Qwen Image Edit | 889 | ±20 | 869–909 | 2,589 | 2025-08 | 30.0 |
| 70 | FLUX.1 Kontext [dev] | 883 | ±21 | 862–904 | 2,356 | 2025-06 | 25.0 |
| 71 | HiDream-O1-Image-Dev | 880 | ±15 | 865–895 | 5,177 | 2026-05 | 5.0 |
| 72 | GLM-Image | 824 | ±22 | 802–846 | 2,532 | 2026-01 | 50.0 |
| 73 | Bagel | 792 | ±25 | 767–817 | 2,367 | 2025-05 | 100.0 |
| 74 | OmniGen V2 | 785 | ±23 | 762–808 | 2,603 | 2025-06 | 150.0 |
| 75 | HiDream-E1.1 | 768 | ±26 | 742–794 | 2,271 | 2025-07 | 60.0 |
| 76 | Step1X-Edit | 601 | ±39 | 562–640 | 2,466 | 2025-04 | 30.0 |
说明： API 价格采用原始榜单提供的每 1,000 张图片价格。榜单中不同模型的计费方式、编辑能力和输出质量可能不同，价格不等于单次编辑的实际成本。
AI 图像生成模型排行榜（文生图）
数据来源：Artificial Analysis — Text to Image Leaderboard
排名依据：Elo 评分。样本数、95% 置信区间、发布日期及 API 价格以榜单数据为准。
| 排名 | 模型 | Elo | 误差范围 | 95% 置信区间 | 样本数 | 发布日期 | API 价格（美元/1000 张） |
|---|---|---|---|---|---|---|---|
| 1 | GPT Image 2.5 Sunburst (max) | 1222 | ±10 | 1212–1232 | 13,014 | 2026-09 | 210.7 |
| 2 | GPT Image 2.5 Flare (max) | 1207 | ±10 | 1197–1217 | 12,364 | 2026-09 | 210.7 |
| 3 | GPT Image 2 (high) | 1186 | ±11 | 1175–1197 | 12,835 | 2026-04 | 211.0 |
| 4 | Nano Banana 2.1 | 1176 | ±12 | 1164–1188 | 6,210 | 2026-10 | 33.6 |
| 5 | MAI-Image-2.6 | 1166 | ±12 | 1154–1178 | 8,342 | 2026-08 | 38.9 |
| 6 | Grok Imagine Image 2.0 | 1166 | ±12 | 1154–1178 | 7,481 | 2026-08 | 60.0 |
| 7 | Nano Banana 2 (Gemini 3.1 Flash Image) | 1140 | ±10 | 1130–1150 | 16,075 | 2026-02 | 67.0 |
| 8 | Muse Image | 1127 | ±11 | 1116–1138 | 8,479 | 2026-07 | 10.0 |
| 9 | GPT Image 1.5 (high) | 1123 | ±11 | 1112–1134 | 12,830 | 2025-12 | 133.0 |
| 10 | MAI-Image-2.6-Flash | 1119 | ±12 | 1107–1131 | 5,482 | 2026-09 | 19.5 |
| 11 | Nano Banana Pro (Gemini 3 Pro Image) | 1118 | ±10 | 1108–1128 | 15,174 | 2025-11 | 134.0 |
| 12 | MAI-Image-2.5 | 1118 | ±11 | 1107–1129 | 13,789 | 2026-06 | 48.1 |
| 13 | FLUX 3 Image | 1117 | ±12 | 1105–1129 | 6,115 | 2026-10 | 48.0 |
| 14 | MAI-Image-2.5-Pro | 1114 | ±10 | 1104–1124 | 12,200 | 2026-07 | 108.5 |
| 15 | Nano Banana 2 Lite (Gemini 3.1 Flash Lite Image) | 1107 | ±10 | 1097–1117 | 15,080 | 2026-06 | 33.6 |
| 16 | Seedream 5.0 Pro | 1100 | ±11 | 1089–1111 | 11,633 | 2026-07 | 90.0 |
| 17 | Qwen-Image-3.0-Pro | 1095 | ±12 | 1083–1107 | 6,332 | 2026-07 | 40.0 |
| 18 | Qwen-Image-3.0 | 1070 | ±12 | 1058–1082 | 6,090 | 2026-07 | 30.0 |
| 19 | Qwen-Image-2.1-Pro | 1056 | ±12 | 1044–1068 | 4,970 | 2026-10 | 35.3 |
| 20 | grok-imagine-image-quality | 1049 | ±9 | 1040–1058 | 16,728 | 2026-05 | 50.0 |
| 21 | FLUX 3 Image Klein | 1043 | ±12 | 1031–1055 | 5,184 | 2026-10 | 9.0 |
| 22 | Qwen-Image-2.1 | 1042 | ±12 | 1030–1054 | 5,712 | 2026-09 | 暂无 API |
| 23 | Fibo 2.0 | 1039 | ±12 | 1027–1051 | 5,113 | 2026-10 | 32.0 |
| 24 | Krea 2 Large | 1034 | ±9 | 1025–1043 | 14,513 | 2026-05 | 60.0 |
| 25 | Qwen Image 2.0 Pro (2026-04-22) | 1030 | ±11 | 1019–1041 | 6,138 | 2026-04 | 75.0 |
| 26 | MAI-Image-2.5-Flash | 1029 | ±11 | 1018–1040 | 5,855 | 2026-06 | 20.0 |
| 27 | Qwen-Image-2.1-Turbo | 1029 | ±12 | 1017–1041 | 4,625 | 2026-10 | 16.0 |
| 28 | Wan 2.6 Image | 1027 | ±12 | 1015–1039 | 5,370 | 2025-12 | 30.0 |
| 29 | Recraft V4.1 Utility | 1026 | ±11 | 1015–1037 | 7,452 | 2026-05 | 35.0 |
| 30 | Seedream 4.5 | 1026 | ±12 | 1014–1038 | 6,392 | 2025-12 | 40.0 |
| 31 | FLUX.2 [flex] | 1026 | ±11 | 1015–1037 | 6,082 | 2025-11 | 60.0 |
| 32 | Ideogram 4.0 (Quality) | 1026 | ±10 | 1016–1036 | 10,688 | 2026-06 | 100.0 |
| 33 | Luma UNI 1 Max | 1023 | ±12 | 1011–1035 | 5,539 | 2026-05 | 100.0 |
| 34 | grok-imagine-image | 1022 | ±11 | 1011–1033 | 6,330 | 2026-01 | 20.0 |
| 35 | FLUX.2 [max] | 1021 | ±10 | 1011–1031 | 11,097 | 2025-12 | 70.0 |
| 36 | Krea 2 Medium Turbo | 1020 | ±11 | 1009–1031 | 6,060 | 2026-06 | 15.0 |
| 37 | Recraft V4.1 Utility Pro | 1017 | ±9 | 1008–1026 | 15,414 | 2026-05 | 210.0 |
| 38 | HiDream-O1-Image-1.5 | 1017 | ±11 | 1006–1028 | 12,168 | 2026-06 | 80.0 |
| 39 | Seedream 5.0 Lite | 1017 | ±12 | 1005–1029 | 5,605 | 2026-02 | 35.0 |
| 40 | Ideogram 4.5 (High) | 1015 | ±15 | 1000–1030 | 2,888 | 2026-09 | 100.0 |
| 41 | Wan2.6 Text to Image | 1014 | ±12 | 1002–1026 | 5,561 | 2026-01 | 30.0 |
| 42 | GPT Image 1 (high) | 1013 | ±12 | 1001–1025 | 5,518 | 2025-04 | 167.0 |
| 43 | Krea 2 Medium | 1012 | ±9 | 1003–1021 | 14,449 | 2026-05 | 30.0 |
| 44 | FLUX.2 [pro] | 1003 | ±10 | 993–1013 | 10,826 | 2025-11 | 30.0 |
| 45 | Ideogram 4.0 | 1002 | ±10 | 992–1012 | 10,102 | 2026-06 | 60.0 |
| 46 | FLUX.2 [dev] | 1000 | 锚点 | 固定参照点 | 18,040 | 2025-11 | 12.0 |
| 47 | Ming-Image-0.1-Design | 999 | ±9 | 990–1008 | 20,241 | 2026-09 | 30.0 |
| 48 | P-Image-Ideogram (High) | 995 | ±14 | 981–1009 | 3,572 | 2026-07 | 15.0 |
| 49 | FLUX.2 [dev] Turbo | 994 | ±11 | 983–1005 | 6,274 | 2025-12 | 8.0 |
| 50 | Cosmos3-Super-Text2Image (agentic) | 993 | ±10 | 983–1003 | 11,549 | 2026-05 | 暂无 API |
| 51 | Luma UNI 1 | 992 | ±12 | 980–1004 | 5,559 | 2026-05 | 40.4 |
| 52 | P-Image-Ideogram (Very High) | 992 | ±15 | 977–1007 | 2,776 | 2026-07 | 33.0 |
| 53 | Qwen Image 2.0 (2026-03-03) | 990 | ±16 | 974–1006 | 2,633 | 2026-03 | 35.0 |
| 54 | HunyuanImage 3.0 (Tencent Cloud) | 989 | ±16 | 973–1005 | 2,531 | 2025-09 | 32.0 |
| 55 | FLUX.2 [dev] Flash | 988 | ±12 | 976–1000 | 5,344 | 2025-12 | 5.0 |
| 56 | HunyuanImage 3.0 Instruct | 987 | ±12 | 975–999 | 5,683 | 2026-01 | 暂无 API |
| 57 | Wan 2.7 | 986 | ±12 | 974–998 | 5,491 | 2026-04 | 26.0 |
| 58 | Cosmos3-Super-Text2Image | 985 | ±10 | 975–995 | 10,887 | 2026-05 | 40.0 |
| 59 | Nano Banana (Gemini 2.5 Flash Image) | 985 | ±12 | 973–997 | 5,312 | 2025-08 | 39.0 |
| 60 | Ideogram 4.0 Fast (Quality) | 982 | ±12 | 970–994 | 5,946 | 2026-07 | 17.5 |
| 61 | Ideogram 4.0 Instant | 981 | ±12 | 969–993 | 5,492 | 2026-07 | 7.5 |
| 62 | Wan 2.7 Pro | 979 | ±12 | 967–991 | 5,498 | 2026-04 | 75.0 |
| 63 | HiDream-O1-Image | 975 | ±10 | 965–985 | 10,662 | 2026-05 | 40.0 |
| 64 | P-Image-Ideogram (Medium) | 972 | ±14 | 958–986 | 3,761 | 2026-07 | 10.0 |
| 65 | Cosmos3-Super-Text2Image-4Step | 970 | ±10 | 960–980 | 9,044 | 2026-07 | 暂无 API |
| 66 | HunyuanImage 3.5 (Preview) | 970 | ±12 | 958–982 | 5,586 | 2026-09 | 24.0 |
| 67 | Ideogram 4.0 Fast | 963 | ±12 | 951–975 | 5,438 | 2026-07 | 10.5 |
| 68 | P-Image-Ideogram (Low) | 959 | ±14 | 945–973 | 3,832 | 2026-07 | 7.5 |
| 69 | Qwen Image Plus 2601 | 952 | ±16 | 936–968 | 2,584 | 2026-01 | 30.0 |
| 70 | Recraft V4.1 Flash | 949 | ±12 | 937–961 | 5,770 | 2026-09 | 7.0 |
| 71 | GPT Image 1 Mini (medium) | 936 | ±17 | 919–953 | 2,333 | 2025-10 | 11.0 |
| 72 | Z-Image Turbo | 936 | ±11 | 925–947 | 8,170 | 2025-12 | 5.0 |
| 73 | HunyuanImage 3.0 | 932 | ±16 | 916–948 | 2,443 | 2025-09 | 暂无 API |
| 74 | Dreamina 3.1 | 931 | ±17 | 914–948 | 2,372 | 2025-08 | 30.0 |
| 75 | FLUX.2 [klein] 9B | 929 | ±10 | 919–939 | 10,730 | 2026-01 | 15.0 |
| 76 | FLUX.1 Kontext [max] | 929 | ±12 | 917–941 | 5,276 | 2025-05 | 80.0 |
| 77 | Kling Image 3.0 Omni | 918 | ±17 | 901–935 | 2,672 | 2026-02 | 28.0 |
| 78 | FLUX.1 Kontext [pro] | 917 | ±18 | 899–935 | 2,229 | 2025-05 | 40.0 |
| 79 | ERNIE Image Turbo | 914 | ±10 | 904–924 | 10,560 | 2026-04 | 10.5 |
| 80 | Ideogram 3.0 | 911 | ±18 | 893–929 | 2,333 | 2025-03 | 60.0 |
| 81 | P-Image-Ideogram (Very Low) | 909 | ±15 | 894–924 | 3,309 | 2026-07 | 3.0 |
| 82 | ERNIE Image | 907 | ±10 | 897–917 | 10,760 | 2026-04 | 31.5 |
| 83 | Qwen Image | 903 | ±18 | 885–921 | 2,339 | 2025-08 | 20.0 |
| 84 | P-Image | 900 | ±18 | 882–918 | 2,351 | 2025-12 | 5.0 |
| 85 | FLUX.1 Krea [dev] | 896 | ±39 | 857–935 | 515 | 2025-07 | 25.0 |
| 86 | Vidu Q2 | 896 | ±17 | 879–913 | 2,524 | 2025-11 | 30.0 |
| 87 | Z-Image Base | 891 | ±17 | 874–908 | 2,500 | 2026-01 | 10.0 |
| 88 | GLM-Image | 885 | ±18 | 867–903 | 2,457 | 2026-01 | 50.0 |
| 89 | Lucid Origin Ultra | 877 | ±17 | 860–894 | 2,689 | 2025-08 | 86.7 |
| 90 | FLUX1.1 [pro] | 870 | ±19 | 851–889 | 2,247 | 2024-10 | 40.0 |
| 91 | Lucid Origin Fast | 860 | ±18 | 842–878 | 2,611 | 2025-08 | 17.9 |
| 92 | LongCat Image | 856 | ±19 | 837–875 | 2,146 | 2025-12 | 130.0 |
| 93 | FLUX1.1 [pro] Ultra | 855 | ±19 | 836–874 | 2,284 | 2024-11 | 60.0 |
| 94 | FIBO | 849 | ±19 | 830–868 | 2,258 | 2025-10 | 40.0 |
| 95 | Luma Photon | 843 | ±19 | 824–862 | 2,251 | 2024-12 | 19.0 |
| 96 | FLUX.2 [klein] 4B | 837 | ±12 | 825–849 | 6,418 | 2026-01 | 14.0 |
| 97 | Phoenix 0.9 Ultra | 833 | ±20 | 813–853 | 2,214 | 2024-06 | 76.2 |
| 98 | Phoenix 1.0 Fast | 827 | ±20 | 807–847 | 2,151 | 2024-12 | 19.4 |
| 99 | SRPO | 826 | ±20 | 806–846 | 2,154 | 2025-09 | 26.0 |
| 100 | FLUX.1 [dev] | 823 | ±45 | 778–868 | 511 | 2024-08 | 25.0 |
| 101 | FIBO Lite | 813 | ±16 | 797–829 | 3,300 | 2025-11 | 20.0 |
| 102 | HiDream-O1-Image-Dev | 810 | ±20 | 790–830 | 2,323 | 2026-05 | 5.0 |
| 103 | Luma Photon Flash | 769 | ±23 | 746–792 | 2,052 | 2024-12 | 5.0 |
| 104 | FLUX.1 [schnell] | 740 | ±20 | 720–760 | 3,045 | 2024-08 | 3.0 |
| 105 | OmniGen V2 | 717 | ±26 | 691–743 | 2,043 | 2025-06 | 150.0 |
| 106 | Bagel | 594 | ±35 | 559–629 | 2,081 | 2025-05 | 100.0 |
说明： API 价格为原始榜单提供的每 1,000 张图片价格，单位为美元；不同模型的输出分辨率、生成质量和计费规则可能不同，不宜只按价格直接判断性价比。
Artificial Analysis 文本转语音（TTS）排行榜
来源：Artificial Analysis — Provider Voice Leaderboard
根据用户提供的榜单文本整理；模型名称、排名、Elo、样本量、语音数量、发布日期及 API 定价均保留来源中的表述。未提供的定价标为“暂无”。
| 排名 | 排名范围 | 模型 | 标记 | Elo | 误差范围 | 95% 区间 | 样本数 | Arena 语音数 | 发布日期 | API 定价 |
|---|---|---|---|---|---|---|---|---|---|---|
| 1 | 1-2 | Eleven v4 Turbo | — | 1336 | ±17 | 95% 区间 1319 至 1353 | 2,085 | 8 个语音 | Sept 2026 | US$40.0 /100 万字符 |
| 2 | 1-2 | Eleven v4 | — | 1332 | ±16 | 95% 区间 1316 至 1348 | 2,384 | 8 个语音 | Sept 2026 | US$80.0 /100 万字符 |
| 3 | 3-4 | Qwen-Audio-3.1-TTS-Plus | — | 1307 | ±16 | 95% 区间 1291 至 1323 | 2,030 | 8 个语音 | Sept 2026 | US$19.3 /100 万字符 |
| 4 | 4-5 | Sonic 3.6 | — | 1292 | ±14 | 95% 区间 1278 至 1306 | 2,842 | 8 个语音 | Aug 2026 | US$49.0 /100 万字符 |
| 5 | 4-6 | Gemini 3.8 Flash TTS | — | 1280 | ±14 | 95% 区间 1266 至 1294 | 3,091 | 8 个语音 | Sept 2026 | US$16.5 /100 万字符 |
| 6 | 5-7 | Realtime TTS-2 | — | 1270 | ±14 | 95% 区间 1256 至 1284 | 2,254 | 8 个语音 | Aug 2026 | US$20.8 /100 万字符 |
| 7 | 6-7 | Qwen-Audio-3.0-TTS-Plus | — | 1266 | ±14 | 95% 区间 1252 至 1280 | 2,507 | 8 个语音 | Sept 2026 | US$19.3 /100 万字符 |
| 8 | 8-10 | Gemini 3.8 Flash-Lite TTS | — | 1252 | ±13 | 95% 区间 1239 至 1265 | 3,078 | 8 个语音 | Sept 2026 | US$11.0 /100 万字符 |
| 9 | 8-10 | Fish Audio Drama 3 Preview | Hugging Face | 1248 | ±15 | 95% 区间 1233 至 1263 | 2,237 | 8 个语音 | Oct 2026 | US$15.0 /100 万字符 |
| 10 | 8-10 | Simba 3.2 | — | 1246 | ±12 | 95% 区间 1234 至 1258 | 3,436 | 8 个语音 | Jul 2026 | US$6.6 /100 万字符 |
| 11 | 11-13 | Breeze TTS 2 | Hugging Face | 1230 | ±14 | 95% 区间 1216 至 1244 | 2,378 | 8 个语音 | Aug 2026 | US$34.0 /100 万字符 |
| 12 | 11-17 | Voxtral 3 TTS | — | 1219 | ±15 | 95% 区间 1204 至 1234 | 1,710 | 8 个语音 | Oct 2026 | US$13.0 /100 万字符 |
| 13 | 11-17 | Realtime TTS-2 Flash | — | 1219 | ±13 | 95% 区间 1206 至 1232 | 2,667 | 8 个语音 | Aug 2026 | US$10.4 /100 万字符 |
| 14 | 12-17 | v3 Conversational | — | 1215 | ±13 | 95% 区间 1202 至 1228 | 2,837 | 8 个语音 | Aug 2026 | US$50.0 /100 万字符 |
| 15 | 12-17 | StepAudio 2.5 TTS (Aug 2026) | — | 1214 | ±14 | 95% 区间 1200 至 1228 | 2,254 | 8 个语音 | Aug 2026 | US$85.0 /100 万字符 |
| 16 | 12-17 | Gemini 3.1 Flash TTS | — | 1213 | ±11 | 95% 区间 1202 至 1224 | 4,327 | 7 个语音 | Apr 2026 | US$18.3 /100 万字符 |
| 17 | 12-17 | Luna TTS | — | 1210 | ±12 | 95% 区间 1198 至 1222 | 3,598 | 3 个语音 | Jun 2026 | US$15.0 /100 万字符 |
| 18 | 18-22 | Sonic 3.5 | — | 1192 | ±11 | 95% 区间 1181 至 1203 | 3,881 | 8 个语音 | May 2026 | US$49.0 /100 万字符 |
| 19 | 18-22 | Ming-3.0-tts | Hugging Face | 1191 | ±14 | 95% 区间 1177 至 1205 | 2,169 | 3 个语音 | Oct 2026 | 暂无 |
| 20 | 18-24 | Fish Audio S2.1 Pro (Oct 2026) | — | 1186 | ±14 | 95% 区间 1172 至 1200 | 2,123 | 8 个语音 | Oct 2026 | US$15.0 /100 万字符 |
| 21 | 18-24 | Eleven v3 | — | 1184 | ±10 | 95% 区间 1174 至 1194 | 5,140 | 8 个语音 | Feb 2026 | US$100.0 /100 万字符 |
| 22 | 18-25 | Soniox TTS Real-Time v2 | — | 1182 | ±11 | 95% 区间 1171 至 1193 | 5,436 | 8 个语音 | Aug 2026 | US$14.2 /100 万字符 |
| 23 | 20-25 | Gradium TTS (Sep 2026) | — | 1177 | ±13 | 95% 区间 1164 至 1190 | 2,245 | 8 个语音 | Sept 2026 | US$47.8 /100 万字符 |
| 24 | 21-25 | Speech 2.8 HD | — | 1176 | ±10 | 95% 区间 1166 至 1186 | 5,246 | 8 个语音 | Feb 2026 | US$100.0 /100 万字符 |
| 25 | 22-25 | Lightning V3.1 Pro (Jul 2026) | — | 1171 | ±12 | 95% 区间 1159 至 1183 | 2,859 | 8 个语音 | Jul 2026 | US$19.5 /100 万字符 |
| 26 | 26-29 | Falcon 2 | — | 1158 | ±13 | 95% 区间 1145 至 1171 | 2,501 | 8 个语音 | Jul 2026 | US$10.0 /100 万字符 |
| 27 | 26-29 | Speech 2.8 Turbo | — | 1156 | ±10 | 95% 区间 1146 至 1166 | 5,016 | 8 个语音 | Feb 2026 | US$60.0 /100 万字符 |
| 28 | 26-31 | Gradium TTS (Aug 2026) | — | 1154 | ±13 | 95% 区间 1141 至 1167 | 2,493 | 8 个语音 | Aug 2026 | US$47.2 /100 万字符 |
| 29 | 26-32 | SpaceXAI TTS | — | 1152 | ±13 | 95% 区间 1139 至 1165 | 1,935 | 6 个语音 | Mar 2026 | US$15.0 /100 万字符 |
| 30 | 28-32 | Azure HD 2.5 | — | 1143 | ±11 | 95% 区间 1132 至 1154 | 3,024 | 8 个语音 | Mar 2026 | US$22.0 /100 万字符 |
| 31 | 29-33 | Async Flash v1.5 | — | 1141 | ±12 | 95% 区间 1129 至 1153 | 2,725 | 8 个语音 | May 2026 | US$10.1 /100 万字符 |
| 32 | 30-33 | Step TTS 2 (Mar 2026) | — | 1140 | ±12 | 95% 区间 1128 至 1152 | 2,835 | 2 个语音 | Feb 2026 | US$40.0 /100 万字符 |
| 33 | 31-36 | Speech 2.6 HD | — | 1130 | ±11 | 95% 区间 1119 至 1141 | 4,316 | 8 个语音 | Oct 2025 | US$100.0 /100 万字符 |
| 34 | 33-36 | Async Pro v1.0 | — | 1128 | ±12 | 95% 区间 1116 至 1140 | 2,602 | 8 个语音 | May 2026 | US$20.2 /100 万字符 |
| 35 | 33-36 | Speech 2.6 Turbo | — | 1122 | ±11 | 95% 区间 1111 至 1133 | 4,590 | 8 个语音 | Oct 2025 | US$60.0 /100 万字符 |
| 36 | 33-39 | Simba 3.0 | — | 1119 | ±11 | 95% 区间 1108 至 1130 | 3,053 | 8 个语音 | Feb 2026 | US$6.6 /100 万字符 |
| 37 | 36-41 | Speech-02-HD | — | 1111 | ±11 | 95% 区间 1100 至 1122 | 4,668 | 4 个语音 | Mar 2025 | US$100.0 /100 万字符 |
| 38 | 35-42 | Fish Audio S2 Pro | Hugging Face | 1110 | ±12 | 95% 区间 1098 至 1122 | 2,750 | 4 个语音 | Mar 2026 | US$15.0 /100 万字符 |
| 39 | 35-42 | Chatterbox HD | — | 1110 | ±12 | 95% 区间 1098 至 1122 | 2,692 | — | May 2025 | US$40.0 /100 万字符 |
| 40 | 37-44 | TTS-1 HD | — | 1103 | ±11 | 95% 区间 1092 至 1114 | 3,787 | 6 个语音 | Nov 2023 | US$30.0 /100 万字符 |
| 41 | 37-47 | Step Audio EditX (Mar 2026) | Hugging Face | 1100 | ±12 | 95% 区间 1088 至 1112 | 2,359 | 2 个语音 | Feb 2026 | 暂无 |
| 42 | 40-47 | Turbo v2.5 | — | 1099 | ±10 | 95% 区间 1089 至 1109 | 7,058 | 8 个语音 | Jul 2024 | US$50.0 /100 万字符 |
| 43 | 40-49 | Speech 2.6 | — | 1097 | ±12 | 95% 区间 1085 至 1109 | 2,493 | 8 个语音 | Jul 2026 | 暂无 |
| 44 | 40-49 | Multilingual v2 | — | 1096 | ±10 | 95% 区间 1086 至 1106 | 7,520 | 8 个语音 | Aug 2023 | US$100.0 /100 万字符 |
| 45 | 41-49 | Speech-02-Turbo | — | 1092 | ±11 | 95% 区间 1081 至 1103 | 4,002 | 4 个语音 | Mar 2025 | US$60.0 /100 万字符 |
| 46 | 41-49 | Voxtral TTS | Hugging Face | 1091 | ±12 | 95% 区间 1079 至 1103 | 2,460 | 3 个语音 | Mar 2026 | US$16.0 /100 万字符 |
| 47 | 42-49 | TTS-1 | — | 1089 | ±10 | 95% 区间 1079 至 1099 | 6,307 | 6 个语音 | Nov 2023 | US$15.0 /100 万字符 |
| 48 | 43-51 | Gemini 2.5 Flash Lite TTS | — | 1087 | ±11 | 95% 区间 1076 至 1098 | 3,322 | 5 个语音 | Dec 2025 | US$9.2 /100 万字符 |
| 49 | 43-52 | OpenAudio S1 | — | 1087 | ±11 | 95% 区间 1076 至 1098 | 4,658 | 8 个语音 | Jun 2025 | US$15.0 /100 万字符 |
| 50 | 48-56 | Flash v2.5 | — | 1078 | ±11 | 95% 区间 1067 至 1089 | 5,849 | 7 个语音 | Dec 2024 | US$50.0 /100 万字符 |
| 51 | 48-58 | GPT-Realtime-2 | — | 1077 | ±12 | 95% 区间 1065 至 1089 | 2,324 | 5 个语音 | May 2026 | US$191.6 /100 万字符 |
| 52 | 49-58 | Studio | — | 1076 | ±11 | 95% 区间 1065 至 1087 | 4,601 | 4 个语音 | Feb 2023 | US$160.0 /100 万字符 |
| 53 | 48-59 | Journey | — | 1075 | ±12 | 95% 区间 1063 至 1087 | 2,368 | 1 个语音 | Dec 2023 | US$160.0 /100 万字符 |
| 54 | 50-60 | T2A-01-HD | — | 1069 | ±11 | 95% 区间 1058 至 1080 | 5,088 | 8 个语音 | Jan 2025 | US$50.0 /100 万字符 |
| 55 | 50-62 | Maya 2 Flash | — | 1068 | ±13 | 95% 区间 1055 至 1081 | 2,167 | 4 个语音 | Jul 2026 | 暂无 |
| 56 | 50-62 | Polly Generative | — | 1067 | ±11 | 95% 区间 1056 至 1078 | 3,851 | 4 个语音 | May 2024 | US$30.0 /100 万字符 |
| 57 | 52-65 | Kokoro 82M v1.0 | Hugging Face | 1066 | ±11 | 95% 区间 1055 至 1077 | 5,450 | 8 个语音 | Jan 2025 | US$0.7 /100 万字符 |
| 58 | 51-65 | Magpie-Multilingual 357M (Feb 2026) | Hugging Face | 1065 | ±12 | 95% 区间 1053 至 1077 | 2,315 | 5 个语音 | Feb 2026 | 暂无 |
| 59 | 51-65 | Gemini 2.5 Flash TTS (Dec 2025) | — | 1065 | ±12 | 95% 区间 1053 至 1077 | 2,967 | 8 个语音 | Dec 2025 | US$9.2 /100 万字符 |
| 60 | 54-65 | Chirp 3: HD | — | 1062 | ±12 | 95% 区间 1050 至 1074 | 3,459 | 8 个语音 | Mar 2025 | US$30.0 /100 万字符 |
| 61 | 54-68 | Octave 2 | — | 1058 | ±12 | 95% 区间 1046 至 1070 | 3,213 | 8 个语音 | Oct 2025 | US$87.5 /100 万字符 |
| 62 | 54-68 | Sonic 3 | — | 1058 | ±12 | 95% 区间 1046 至 1070 | 3,580 | 8 个语音 | Oct 2025 | US$49.0 /100 万字符 |
| 63 | 56-70 | MiMo-V2.5-TTS | — | 1055 | ±12 | 95% 区间 1043 至 1067 | 2,141 | 4 个语音 | - | 暂无 |
| 64 | 56-70 | Flux TTS | — | 1055 | ±13 | 95% 区间 1042 至 1068 | 2,226 | 8 个语音 | Aug 2026 | US$45.0 /100 万字符 |
| 65 | 57-70 | Coda | — | 1055 | ±12 | 95% 区间 1043 至 1067 | 2,485 | 7 个语音 | May 2026 | US$50.0 /100 万字符 |
| 66 | 60-72 | Maya 2 Global | — | 1050 | ±13 | 95% 区间 1037 至 1063 | 2,021 | 4 个语音 | Jul 2026 | 暂无 |
| 67 | 61-72 | Async Flash v1.0 | — | 1050 | ±11 | 95% 区间 1039 至 1061 | 4,106 | 8 个语音 | Jul 2025 | US$10.1 /100 万字符 |
| 68 | 61-72 | Maya1 | Hugging Face | 1048 | ±12 | 95% 区间 1036 至 1060 | 3,336 | 4 个语音 | Jun 2025 | 暂无 |
| 69 | 61-74 | Higgs Audio V3 TTS | Hugging Face | 1045 | ±13 | 95% 区间 1032 至 1058 | 2,234 | 8 个语音 | Jun 2026 | US$15.0 /100 万字符 |
| 70 | 60-78 | OpenAudio S1 Mini | Hugging Face | 1045 | ±20 | 95% 区间 1025 至 1065 | 1,682 | 8 个语音 | Jun 2025 | US$15.0 /100 万字符 |
| 71 | 66-77 | Gemini 2.5 Pro (Dec 2025) | — | 1040 | ±12 | 95% 区间 1028 至 1052 | 2,490 | 4 个语音 | Dec 2025 | US$18.3 /100 万字符 |
| 72 | 67-77 | Octave TTS | — | 1039 | ±11 | 95% 区间 1028 至 1050 | 4,741 | 8 个语音 | Feb 2025 | US$87.5 /100 万字符 |
| 73 | 69-78 | Bland Speech v3 | — | 1035 | ±12 | 95% 区间 1023 至 1047 | 2,542 | 8 个语音 | Aug 2026 | US$40.0 /100 万字符 |
| 74 | 66-79 | Azure Neural | — | 1034 | ±20 | 95% 区间 1014 至 1054 | 1,156 | 8 个语音 | Sept 2018 | US$15.0 /100 万字符 |
| 75 | 71-79 | T2A-01-Turbo | — | 1031 | ±11 | 95% 区间 1020 至 1042 | 5,000 | 8 个语音 | Jan 2025 | US$30.0 /100 万字符 |
| 76 | 71-79 | Polly Long-Form | — | 1031 | ±13 | 95% 区间 1018 至 1044 | 1,986 | 3 个语音 | Nov 2023 | US$100.0 /100 万字符 |
| 77 | 71-79 | MAI-Voice-1 | — | 1030 | ±12 | 95% 区间 1018 至 1042 | 2,363 | 4 个语音 | Apr 2026 | 暂无 |
| 78 | 73-79 | Chatterbox | GitHub | 1025 | ±11 | 95% 区间 1014 至 1036 | 4,724 | 8 个语音 | May 2025 | US$25.0 /100 万字符 |
| 79 | 74-79 | Lightning v3.1 | — | 1023 | ±12 | 95% 区间 1011 至 1035 | 2,356 | 7 个语音 | Mar 2026 | US$25.0 /100 万字符 |
| 80 | 80-81 | Raon SpeechLM | — | 1006 | ±13 | 95% 区间 993 至 1019 | 2,216 | 8 个语音 | Apr 2026 | 暂无 |
| 81 | 81 | Zonos-v0.1 | GitHub | 1000 | 锚点 | 固定为 1000，作为本榜单 Elo 刻度的参照点。 | 4,875 | 8 个语音 | Feb 2025 | US$20.0 /100 万字符 |
| 82 | 82 | Murf Speech Gen 2 | — | 988 | ±12 | 95% 区间 976 至 1000 | 5,035 | 8 个语音 | Jul 2024 | US$100.0 /100 万字符 |
| 83 | 83-85 | VibeVoice 1.5B | Hugging Face | 956 | ±14 | 95% 区间 942 至 970 | 2,160 | 2 个语音 | Aug 2025 | 暂无 |
| 84 | 83-86 | OpenVoice v2 | Hugging Face | 955 | ±13 | 95% 区间 942 至 968 | 3,085 | — | Apr 2024 | US$8.3 /100 万字符 |
| 85 | 83-87 | Neuphonic TTS | — | 943 | ±14 | 95% 区间 929 至 957 | 2,487 | 7 个语音 | Oct 2024 | US$20.8 /100 万字符 |
| 86 | 84-87 | Qwen3 TTS Flash | — | 942 | ±14 | 95% 区间 928 至 956 | 2,220 | 4 个语音 | Sept 2025 | US$11.0 /100 万字符 |
| 87 | 86-89 | Qwen3 TTS | — | 930 | ±13 | 95% 区间 917 至 943 | 3,113 | 8 个语音 | Jan 2026 | US$13.8 /100 万字符 |
| 88 | 87-90 | Kugel 3 | — | 926 | ±13 | 95% 区间 913 至 939 | 4,394 | 8 个语音 | Jun 2026 | 暂无 |
| 89 | 87-90 | WaveNet | — | 919 | ±12 | 95% 区间 907 至 931 | 6,399 | 8 个语音 | Sept 2016 | US$4.0 /100 万字符 |
| 90 | 88-91 | XTTS v2 | Hugging Face | 915 | ±15 | 95% 区间 900 至 930 | 2,703 | — | Nov 2023 | US$40.4 /100 万字符 |
| 91 | 89-94 | Mist V2 | — | 905 | ±15 | 95% 区间 890 至 920 | 3,046 | 7 个语音 | Mar 2025 | US$31.1 /100 万字符 |
| 92 | 91-95 | Polly Neural | — | 897 | ±14 | 95% 区间 883 至 911 | 2,972 | 7 个语音 | Jul 2019 | US$16.0 /100 万字符 |
| 93 | 91-95 | Neural2 | — | 895 | ±12 | 95% 区间 883 至 907 | 6,683 | 8 个语音 | Jun 2022 | US$16.0 /100 万字符 |
| 94 | 91-95 | StyleTTS 2 | GitHub | 890 | ±15 | 95% 区间 875 至 905 | 2,675 | — | Jun 2023 | US$2.8 /100 万字符 |
| 95 | 92-95 | Standard | — | 888 | ±12 | 95% 区间 876 至 900 | 6,765 | 8 个语音 | Aug 2018 | US$4.0 /100 万字符 |
| 96 | 96 | Noiz TTS | — | 873 | ±15 | 95% 区间 858 至 888 | 2,426 | 8 个语音 | Nov 2025 | 暂无 |
| 97 | 97-98 | MetaVoice v1 | GitHub | 847 | ±18 | 95% 区间 829 至 865 | 1,654 | — | Feb 2024 | 暂无 |
| 98 | 97-98 | Polly Standard | — | 831 | ±16 | 95% 区间 815 至 847 | 3,581 | 4 个语音 | Nov 2016 | US$4.0 /100 万字符 |
说明
- Elo：榜单中的相对偏好评分；不同任务类型的排行榜不应直接混为同一尺度。
- 95% 区间：按来源原文保留，反映评分不确定性。
- Arena 语音数：该条目在榜单中展示的语音数量；原文为破折号时保留为破折号。
- API 定价：按原榜单的“每 100 万字符”价格记录，不代表所有供应商或所有语音配置下的实际最终费用。
整理条目数：98。