LLM 排行榜——比较 OpenAI、Anthropic、Google、SpaceXAI 等公司的 AI 模型
比较并排名 250 多个 AI 模型（LLM）在智能、价格、性能与速度（以分钟计的每任务耗时，以及以首个答案 Token 时间衡量的延迟）、上下文窗口等关键指标上的表现。
如需了解包括方法论在内的更多详情，请参阅我们的常见问题。
智能
Gemini 4 Carbon (high) 和 Claude Opus 5.5 (max) 是智能得分最高的模型，之后是 Gemini 4 Carbon (medium) 和 Claude Opus 5.5 (xhigh)。
每任务耗时
GPT-6 Luna (low) 和 GPT-6.1 Terra (low) 的每任务耗时最短，之后是 GPT-6.1 Sol (low) 和 Mistral Small 4。
首个答案 Token 延迟
Tiny Aya Global 和 Phi-4 Multimodal 的首个答案 Token 延迟最低，之后是 Celeris-1 和 NVIDIA Nemotron Nano 9B V2。
每任务成本
K2 Horizon 375B A23B 和 Command A+ 的每任务成本最低，之后是 GPT-6 Luna (low) 和 Mistral Small 4。
上下文窗口
Llama 4 Scout 和 GPT-6 Sol (Daybreak Blue, max) 支持最大的上下文窗口，之后是 Muse Wonder (max) 和 Muse Wonder (high)。
进一步分析 | |||||||
|---|---|---|---|---|---|---|---|
Gemini 4 Carbon (high) | 1M | Google | 60 | $7.76 | 30.3 | 67.30 | |
Claude Opus 5.5 (max) | 1M | Anthropic | 59 | $10.04 | 26.2 | 345.77 | |
Gemini 4 Carbon (medium) | 1M | Google | 58 | $6.59 | 22.8 | 22.99 | |
Claude Opus 5.5 (xhigh) | 1M | Anthropic | 57 | $6.05 | 16.0 | 68.14 | |
GPT-6 Astra (max) | 1M | OpenAI | 57 | $6.24 | 19.1 | 208.35 | |
GPT-6 Astra (xhigh) | 1M | OpenAI | 55 | $3.66 | 9.5 | 103.63 | |
GPT-6 Astra (high) | 1M | OpenAI | 55 | $3.24 | 8.2 | 41.73 | |
GPT-6.1 Sol (max) | 1M | OpenAI | 55 | $1.27 | 21.2 | 168.90 | |
Claude Sonnet 5.5 (max) | 1M | Anthropic | 55 | $10.02 | 30.7 | 258.46 | |
Gemini 4 Carbon (low) | 1M | Google | 54 | $7.19 | 43.2 | 7.92 | |
GPT-6.1 Sol (xhigh) | 1M | OpenAI | 52 | $0.76 | 9.1 | 69.80 | |
Claude Opus 5.5 (high) | 1M | Anthropic | 52 | $3.46 | 9.1 | 32.75 | |
GPT-6 Astra (medium) | 1M | OpenAI | 52 | $2.73 | 6.0 | 6.49 | |
GPT-6.1 Terra (max) | 1M | OpenAI | 51 | $0.96 | 13.1 | 159.59 | |
Gemini 4 Argon (high) | 1M | Google | 50 | $9.95 | 26.1 | 40.61 | |
GPT-6.1 Sol (high) | 1M | OpenAI | 50 | $0.59 | 6.6 | 38.27 | |
Muse Wonder Full (max) | 1M | Meta | 50 | -- | -- | -- | |
Claude Fable 5.1 (xhigh) | 1M | Anthropic | 50 | $10.85 | 18.6 | 58.20 | |
Claude Sonnet 5.5 (xhigh) | 1M | Anthropic | 50 | $4.09 | 14.7 | 31.01 | |
Claude Fable 5.1 (max) | 1M | Anthropic | 49 | $12.99 | 22.2 | 162.11 | |
Claude Opus 5.5 (medium) | 1M | Anthropic | 48 | $2.45 | 6.6 | 18.70 | |
Claude Fable 5.1 (high) | 1M | Anthropic | 47 | $7.54 | 12.6 | 15.28 | |
Kimi K3.1 Pro | 1M |  Kimi | 46 | $3.68 | 36.0 | 153.54 | |
GPT-6.1 Sol (medium) | 1M | OpenAI | 46 | $0.36 | 3.9 | 6.31 | |
GPT-6.1 Terra (xhigh) | 1M | OpenAI | 45 | $0.40 | 5.9 | 33.02 | |
Muse Wonder (max) | 1M | Meta | 44 | $6.95 | 12.1 | 20.56 | |
GPT-6 Astra (low) | 1M | OpenAI | 43 | $1.42 | 3.2 | 4.02 | |
Claude Fable 5.1 (medium) | 1M | Anthropic | 43 | $5.56 | 9.4 | 8.77 | |
GPT-6.1 Terra (high) | 1M | OpenAI | 42 | $0.26 | 3.8 | 10.91 | |
Claude Sonnet 5.5 (high) | 1M | Anthropic | 42 | $1.64 | 6.8 | 12.82 | |
Muse Wonder (high) | 1M | Meta | 41 | -- | 8.8 | 10.24 | |
Claude Fable 5.1 (low) | 1M | Anthropic | 39 | $4.47 | 7.3 | 8.10 | |
GPT-6.1 Terra (medium) | 1M | OpenAI | 38 | $0.17 | 2.5 | 4.43 | |
Claude Haiku 5.5 (max) | 1M | Anthropic | 36 | $0.41 | 14.6 | 178.94 | |
Grok 4.7 (xhigh) | 500k |  SpaceXAI | 35 | $7.77 | 24.2 | 52.94 | |
Grok 4.7 (high) | 500k | SpaceXAI | 35 | $6.18 | 20.7 | 32.67 | |
GPT-6.1 Sol (low) | 1M | OpenAI | 35 | $0.16 | 1.7 | 2.68 | |
Claude Opus 5.5 (low) | 1M | Anthropic | 35 | $0.98 | 3.0 | 5.40 | |
Claude Haiku 5.5 (xhigh) | 1M | Anthropic | 35 | $0.25 | 9.5 | 49.32 | |
Tencent Hy4 | 1M |  Tencent | 35 | -- | -- | -- | |
Qwen3.8 Max (0902) | 984k |  Alibaba | 34 | $10.02 | 60.6 | 131.50 | |
DeepSeek V4.1 Flash 1017 (max) | 1M |  DeepSeek | 33 | $0.60 | 13.3 | 50.85 | |
GLM-5.3 (max) | 1M |  Z AI | 32 | $5.01 | 30.7 | 96.04 | |
Grok 4.7 (low) | 500k | SpaceXAI | 31 | $2.92 | 9.8 | 9.84 | |
Gemini 3.8 Flash (high) | 1M | Google | 31 | $7.77 | 14.1 | 22.67 | |
Claude Sonnet 5.5 (medium) | 1M | Anthropic | 31 | $0.78 | 3.7 | 5.35 | |
GPT-6.1 Terra (non-reasoning) | 1M | OpenAI | 31 | $0.23 | 2.2 | 1.71 | |
Gemini 3.8 Flash (medium) | 1M | Google | 30 | $6.26 | -- | -- | |
MiMo-V2.6-Pro | 1M |  Xiaomi | 30 | $0.26 | 45.7 | 51.73 | |
Claude Haiku 5.5 (high) | 1M | Anthropic | 30 | $0.16 | 6.7 | 22.97 | |
Ling 3.1 Flash 1019 | 1M |  InclusionAI | 30 | $3.19 | -- | -- | |
GPT-6.1 Terra (low) | 1M | OpenAI | 29 | $0.07 | 1.0 | 2.34 | |
MiniMax-M3.1-Flash | 1M |  MiniMax | 29 | $0.49 | 19.4 | 66.23 | |
DeepSeek V4.1 Flash 1017 (high) | 1M | DeepSeek | 29 | $0.46 | 10.3 | 24.32 | |
Step 5 Preview | 1M |  StepFun | 29 | $2.35 | 14.6 | 24.90 | |
GPT-6 Luna (max) | 1M | OpenAI | 28 | $0.15 | 12.0 | 84.91 | |
GLM-5.3-Flash | 1M | Z AI | 28 | $0.58 | 37.8 | 84.22 | |
MiMo-V2.6-Flash | 1M | Xiaomi | 27 | $0.13 | 61.5 | 55.02 | |
Mistral Large 4 (high) | 1M | Mistral | 26 | $2.06 | -- | -- | |
Claude Sonnet 5.5 (low) | 1M | Anthropic | 26 | $0.51 | 2.5 | 4.04 | |
Muse Spark Open (xhigh) | 1M | Meta | 25 | $2.02 | 7.3 | 13.81 | |
Claude Haiku 5.5 (medium) | 1M | Anthropic | 25 | $0.09 | 4.0 | 11.39 | |
GPT-6 Luna (xhigh) | 1M | OpenAI | 23 | $0.10 | 7.4 | 22.24 | |
Mistral Large 4 (medium) | 1M | Mistral | 23 | $1.41 | -- | -- | |
GLM-5.3 (low) | 1M | Z AI | 23 | $1.57 | 15.5 | 13.04 | |
DeepSeek V4 Pro 0813 (max) | 1M | DeepSeek | 23 | $1.77 | 21.1 | 66.22 | |
Gemini 3.8 Flash (low) | 1M | Google | 23 | -- | -- | -- | |
GPT-6 Luna (high) | 1M | OpenAI | 21 | $0.07 | 5.6 | 12.31 | |
Nanbeige4.3-33B | 524k |  Nanbeige | 20 | -- | -- | -- | |
Qwen3.8 27B (xhigh) | 256k | Alibaba | 20 | $3.53 | 39.8 | 122.66 | |
MiniMax-M3 | 1M | MiniMax | 19 | $1.29 | 13.2 | 73.60 | |
Claude Haiku 5.5 (low) | 1M | Anthropic | 19 | $0.04 | 2.2 | 9.14 | |
DeepSeek V4.1 Flash 1017 (non-reasoning) | 1M | DeepSeek | 19 | $0.43 | 7.0 | 1.29 | |
GPT-6 Luna (medium) | 1M | OpenAI | 18 | $0.04 | -- | -- | |
K2 Horizon 375B A23B | 524k | Institute of Foundation Models | 18 | $0.00 | 5.3 | 25.19 | |
Beam 1.0 | 1M |  Reflection | 18 | $0.28 | 2.3 | 7.17 | |
Qwen3.8 27B (medium) | 256k | Alibaba | 17 | $2.16 | 29.5 | 35.61 | |
Qwen3.8 27B (low) | 256k | Alibaba | 16 | $1.91 | 26.1 | 18.27 | |
Nemotron 3.5 Super VL | 1M | NVIDIA | 14 | $0.54 | 4.6 | 11.98 | |
Inkling Small | 1M |  Thinking Machines | 13 | $0.20 | 4.1 | 15.50 | |
DeepSeek V4 Pro 0813 (non-reasoning) | 1M | DeepSeek | 13 | $0.98 | 13.9 | 2.53 | |
Inkling (xhigh) | 1M | Thinking Machines | 12 | $1.35 | 8.1 | 20.94 | |
Nemotron 3 Ultra | 262k | NVIDIA | 12 | $2.23 | 6.3 | 7.73 | |
Gemini 3.5 Flash-Lite | 1M | Google | 12 | $0.69 | 1.7 | 5.45 | |
Qwen3.8 27B (non-reasoning) | 256k | Alibaba | 10 | $3.77 | 17.3 | 5.48 | |
GPT-6 Luna (low) | 1M | OpenAI | 10 | $0.01 | 0.6 | 2.66 | |
GPT-6 Luna (non-reasoning) | 1M | OpenAI | 9 | $0.07 | 1.8 | 1.13 | |
Muse Glimmer (high) | 131k | Meta | 6 | $0.16 | 4.4 | 21.37 | |
Command A+ | 192k | Cohere | 6 | $0.00 | 3.5 | 23.46 | |
Nemotron 3.5 Lightning | 1M | NVIDIA | 6 | $0.29 | 4.3 | 17.11 | |
Granite 4.2 8B | 131k |  IBM | 5 | $0.04 | 17.0 | 99.33 | |
Granite 4.2 3B | 131k | IBM | 5 | -- | -- | -- | |
Mistral Medium 3.5 | 256k | Mistral | 5 | $1.04 | 4.5 | 21.92 | |
Gemma 4 31B | 256k | Google | 4 | -- | -- | 64.71 | |
Mistral Small 4 | 256k | Mistral | 4 | $0.04 | 1.7 | 14.09 | |
Nemotron 3 Nano 4B | 262k | NVIDIA | 2 | -- | -- | -- | |
JT-35B-Flash | 256k |  China Mobile | -- | -- | -- | -- | |
Qwen3.5 Omni Plus | 256k | Alibaba | -- | -- | -- | 2.78 | |
Nova 2.0 Omni (medium) | 1M |  Amazon | -- | -- | -- | -- | |
Qwen3.5 122B A10B (non-reasoning) | 262k | Alibaba | -- | -- | -- | 4.47 | |
Gemma 4 12B (non-reasoning) | 262k | Google | -- | -- | -- | 3.97 | |
Gemma 3 270M | 32k | Google | -- | -- | -- | -- | |
Apodex 1.1 | 256k |  Apodex | -- | -- | -- | -- | |
Qwen3.5 2B (non-reasoning) | 262k | Alibaba | -- | -- | -- | -- | |
Qwen3 Omni 30B A3B | 66k | Alibaba | -- | -- | -- | 1.27 | |
Nova 2.0 Pro Preview (medium) | 256k | Amazon | -- | -- | -- | 20.66 | |
Qwen3.5 397B A17B | 262k | Alibaba | -- | -- | -- | 66.38 | |
Granite 4.2 30B | 131k | IBM | -- | -- | -- | -- | |
K2 Horizon 7B | 524k | Institute of Foundation Models | -- | -- | -- | -- | |
Mistral Small 4 (non-reasoning) | 256k | Mistral | -- | -- | -- | 1.77 | |
Solar Pro 4 | 512k |  Upstage | -- | -- | -- | 52.61 | |
gpt-oss-20b (low) | 131k | OpenAI | -- | -- | -- | 1.68 | |
o3 | 200k | OpenAI | -- | -- | -- | 7.80 | |
K-EXAONE 2.0 | 262k |  LG AI Research | -- | -- | -- | -- | |
North Mini Code | 256k | Cohere | -- | -- | -- | 7.87 | |
Nova 2.0 Omni (non-reasoning) | 1M | Amazon | -- | -- | -- | -- | |
Granite 4.0 Micro | 128k | IBM | -- | -- | -- | -- | |
Qwen3.5 9B (non-reasoning) | 262k | Alibaba | -- | -- | -- | 1.01 | |
Qwen3.5 122B A10B | 262k | Alibaba | -- | -- | -- | 19.39 | |
Hermes 4 70B (non-reasoning) | 128k |  Nous Research | -- | -- | -- | -- | |
Nemotron 3 Nano (non-reasoning) | 1M | NVIDIA | -- | -- | -- | 0.89 | |
Qwen3 Next 80B A3B | 262k | Alibaba | -- | -- | -- | 1.16 | |
Phi-4 Multimodal | 128k | Microsoft | -- | -- | -- | 0.48 | |
Qwen3.5 9B | 262k | Alibaba | -- | -- | -- | 39.93 | |
NVIDIA Nemotron Nano 9B V2 (non-reasoning) | 131k | NVIDIA | -- | -- | -- | 1.55 | |
EXAONE 4.0 32B (non-reasoning) | 131k | LG AI Research | -- | -- | -- | -- | |
G9v3-3B | 131k |  AI9Stars | -- | -- | -- | -- | |
Qwen3.5 4B (non-reasoning) | 262k | Alibaba | -- | -- | -- | 2.07 | |
Qwen3.5 397B A17B (non-reasoning) | 262k | Alibaba | -- | -- | -- | 2.57 | |
K2 Horizon 3.7B | 524k | Institute of Foundation Models | -- | -- | -- | -- | |
Olmo 3 7B Think | 66k |  Allen Institute for AI | -- | -- | -- | -- | |
HyperCLOVA X SEED Think (32B) | 128k |  Naver | -- | -- | -- | -- | |
HyperNova 60B 2605 (high) | 131k |  Multiverse Computing | -- | -- | -- | -- | |
Nova Micro | 130k | Amazon | -- | -- | -- | 1.32 | |
EXAONE 4.5 33B (non-reasoning) | 262k | LG AI Research | -- | -- | -- | -- | |
INTELLECT-3 | 131k |  Prime Intellect | -- | -- | -- | -- | |
gpt-oss-20b (high) | 131k | OpenAI | -- | -- | -- | 42.24 | |
Olmo 3.1 32B Think | 66k | Allen Institute for AI | -- | -- | -- | -- | |
Nex-N2-Pro | 262k |  Nex AGI | -- | -- | -- | -- | |
Command A | 256k | Cohere | -- | -- | -- | 7.51 | |
LFM2.5-VL-1.6B | 32k |  Liquid AI | -- | -- | -- | -- | |
JT-MINI | 128k | China Mobile | -- | -- | -- | -- | |
Molmo2-8B | 37k | Allen Institute for AI | -- | -- | -- | -- | |
Qwen3.5 0.8B (non-reasoning) | 262k | Alibaba | -- | -- | -- | -- | |
Ling-3.0-flash-VL | 262k | InclusionAI | -- | -- | -- | -- | |
MiniCPM-V 4.6 1.3B | 262k |  OpenBMB | -- | -- | -- | -- | |
EXAONE 4.0 32B | 131k | LG AI Research | -- | -- | -- | -- | |
Sarvam 105B (high) | 128k |  Sarvam | -- | -- | -- | 8.87 | |
Llama 3.1 405B | 128k | Meta | -- | -- | -- | -- | |
JT-4.1 Flash 236B A21B | 262k | China Mobile | -- | -- | -- | -- | |
Ring-flash-2.0 | 128k | InclusionAI | -- | -- | -- | -- | |
LFM2.5-1.2B-Instruct | 32k | Liquid AI | -- | -- | -- | -- | |
Apertus 70B Instruct | 66k |  Swiss AI Initiative | -- | -- | -- | -- | |
Doubao Seed Code | 256k |  ByteDance Seed | -- | -- | -- | -- | |
EXAONE 4.5 33B | 262k | LG AI Research | -- | -- | -- | -- | |
Gemma 4 E4B (non-reasoning) | 128k | Google | -- | -- | -- | 1.79 | |
K2 Horizon MoVA 36B A4B | 524k | Institute of Foundation Models | -- | -- | -- | -- | |
Granite 4.0 350M | 33k | IBM | -- | -- | -- | -- | |
Kimi Linear 48B A3B Instruct | 1M | Kimi | -- | -- | -- | -- | |
Solar Open 100B | 128k | Upstage | -- | -- | -- | -- | |
GPT-5.3 Codex (xhigh) | 400k | OpenAI | -- | -- | -- | 50.84 | |
LFM2.5-1.2B-Thinking | 32k | Liquid AI | -- | -- | -- | -- | |
K2 Think V2 | 262k | Institute of Foundation Models | -- | -- | -- | -- | |
LongCat 2.0 | 1M |  LongCat | -- | -- | -- | 33.58 | |
Qwen3.5 2B | 262k | Alibaba | -- | -- | -- | -- | |
Tri-21B-Think | 32k |  Trillion Labs | -- | -- | -- | -- | |
Granite 4.0 H Small | 128k | IBM | -- | -- | -- | -- | |
Llama 3.2 11B (Vision) | 128k | Meta | -- | -- | -- | 3.85 | |
LongCat Flash Lite | 256k | LongCat | -- | -- | -- | -- | |
Nemotron 3.5 Bolt E4B 4GB | 131k | NVIDIA | -- | -- | -- | -- | |
Nova 2.0 Lite (high) | 1M | Amazon | -- | -- | -- | 21.27 | |
Solar Mini 4 | 1M | Upstage | -- | -- | -- | 52.10 | |
MiniCPM5-2B | 131k | OpenBMB | -- | -- | -- | -- | |
Qwen3.5 4B | 262k | Alibaba | -- | -- | -- | 81.46 | |
Motif-2-12.7B | 128k |  Motif Technologies | -- | -- | -- | -- | |
Qwen3.6 35B A3B (non-reasoning) | 262k | Alibaba | -- | -- | -- | 2.81 | |
Ministral 3 3B | 256k | Mistral | -- | -- | -- | 1.54 | |
Hermes 4 70B | 128k | Nous Research | -- | -- | -- | -- | |
Ling-3.0-flash-Fin | 262k | InclusionAI | -- | -- | -- | -- | |
Gemma 4 26B A4B | 256k | Google | -- | -- | -- | 25.87 | |
NVIDIA Nemotron Nano 12B v2 VL | 128k | NVIDIA | -- | -- | -- | -- | |
Qwen3.5 Omni Flash | 256k | Alibaba | -- | -- | -- | 2.93 | |
Nova 2.0 Lite (low) | 1M | Amazon | -- | -- | -- | 9.59 | |
Magistral Small 1.2 | 128k | Mistral | -- | -- | -- | 3.00 | |
Gemma 4 26B A4B (non-reasoning) | 256k | Google | -- | -- | -- | 1.54 | |
Ministral 3 14B | 256k | Mistral | -- | -- | -- | 2.83 | |
Gemma 4 E2B (non-reasoning) | 128k | Google | -- | -- | -- | -- | |
Sarvam 30B (high) | 66k | Sarvam | -- | -- | -- | -- | |
Mercury 2.5 | 260k |  Inception | -- | -- | -- | 2.39 | |
Nova 2.0 Lite (non-reasoning) | 1M | Amazon | -- | -- | -- | 1.67 | |
Nemotron 3 Nano | 1M | NVIDIA | -- | -- | -- | 9.55 | |
DeepSeek V4 Flash Vision (max) | 1M | DeepSeek | -- | -- | -- | 56.29 | |
MiniCPM5-1B | 128k | OpenBMB | -- | -- | -- | -- | |
Jamba 1.7 Large | 256k |  AI21 Labs | -- | -- | -- | -- | |
Nemotron 3.5 Bolt 74B-A9B | 262k | NVIDIA | -- | -- | -- | -- | |
Nanbeige4.1-3B | 256k | Nanbeige | -- | -- | -- | -- | |
Solar Open2 250B | 1M | Upstage | -- | -- | -- | -- | |
Ling 3.0 Tiny | 262k | InclusionAI | -- | -- | -- | 45.28 | |
Celeris-1 | 131k | Celeris | -- | -- | -- | 0.69 | |
Tiny Aya Global | 8k | Cohere | -- | -- | -- | 0.27 | |
A.X-K2 | 262k |  SK Telecom | -- | -- | -- | -- | |
Exaone 4.0 1.2B (non-reasoning) | 64k | LG AI Research | -- | -- | -- | -- | |
Hermes 4 405B | 128k | Nous Research | -- | -- | -- | 22.33 | |
Magistral Medium 1.2 | 128k | Mistral | -- | -- | -- | -- | |
Qwen3.6 35B A3B | 262k | Alibaba | -- | -- | -- | 34.81 | |
Apertus 8B Instruct | 66k | Swiss AI Initiative | -- | -- | -- | -- | |
Nemotron 3 Nano Omni 30B A3B | 256k | NVIDIA | -- | -- | -- | 8.09 | |
Kimi K2.7 Code | 256k | Kimi | -- | -- | -- | 52.92 | |
Qwen3.8 2.4T A95B | 262k | Alibaba | -- | -- | -- | 132.46 | |
Olmo 3 7B | 66k | Allen Institute for AI | -- | -- | -- | -- | |
JT-4.1 Flash 236B A21B (non-reasoning) | 262k | China Mobile | -- | -- | -- | -- | |
Llama 4 Maverick | 1M | Meta | -- | -- | -- | 1.63 | |
GPT-6 Sol (Daybreak Blue, max) | 1M | OpenAI | -- | -- | -- | -- | |
Olmo 3.1 32B Instruct | 66k | Allen Institute for AI | -- | -- | -- | -- | |
Ministral 3 8B | 256k | Mistral | -- | -- | -- | 2.16 | |
Llama 3.3 70B | 128k | Meta | -- | -- | -- | 1.83 | |
Nemotron Cascade 2 30B A3B | 1M | NVIDIA | -- | -- | -- | -- | |
R1 1776 | 128k |  Perplexity | -- | -- | -- | -- | |
Qwen3.8-Flash-Next | 256k | Alibaba | -- | -- | -- | 98.46 | |
Llama 3.2 90B (Vision) | 128k | Meta | -- | -- | -- | -- | |
K2 Horizon 0.9B | 131k | Institute of Foundation Models | -- | -- | -- | -- | |
Phi-4 Mini | 128k | Microsoft | -- | -- | -- | 4.63 | |
Nova 2.0 Pro Preview (non-reasoning) | 256k | Amazon | -- | -- | -- | 1.47 | |
K2-V2 (high) | 512k | Institute of Foundation Models | -- | -- | -- | -- | |
Gemini 3 Deep Think | 128k | Google | -- | -- | -- | -- | |
ERNIE 5.0 Thinking Preview | 128k |  Baidu | -- | -- | -- | -- | |
Ling 3.1 Flash | 1M | InclusionAI | -- | -- | -- | 14.66 | |
Granite 4.0 H 350M | 33k | IBM | -- | -- | -- | -- | |
Qwen3.5 0.8B | 262k | Alibaba | -- | -- | -- | -- | |
Apriel-v1.6-15B-Thinker | 128k |  ServiceNow | -- | -- | -- | -- | |
Qwen3.7 Plus | 1M | Alibaba | -- | -- | -- | 43.45 | |
DeepHermes 3 - Mistral 24B (non-reasoning) | 32k | Nous Research | -- | -- | -- | -- | |
NVIDIA Nemotron Nano 12B v2 VL (non-reasoning) | 128k | NVIDIA | -- | -- | -- | 1.51 | |
Step3 VL 10B | 66k | StepFun | -- | -- | -- | -- | |
Llama 4 Scout | 10M | Meta | -- | -- | -- | 1.53 | |
ERNIE 4.5 300B A47B | 131k | Baidu | -- | -- | -- | -- | |
Molmo 7B-D | 4k | Allen Institute for AI | -- | -- | -- | -- | |
Motif 3 | 262k | Motif Technologies | -- | -- | -- | -- | |
Gemma 4 E2B | 128k | Google | -- | -- | -- | -- | |
Hy3 | 256k | Tencent | -- | -- | -- | 48.18 | |
Nova 2.0 Omni (low) | 1M | Amazon | -- | -- | -- | -- | |
Jamba 1.7 Mini | 258k | AI21 Labs | -- | -- | -- | -- | |
Trinity Large Thinking | 512k |  Arcee AI | -- | -- | -- | -- | |
DeepHermes 3 - Llama-3.1 8B (non-reasoning) | 128k | Nous Research | -- | -- | -- | -- | |
Qwen3 Omni 30B A3B | 66k | Alibaba | -- | -- | -- | 23.92 | |
Gemini 3.1 Pro Preview | 1M | Google | -- | -- | -- | 19.00 | |
G9v3-39A5B | 131k | AI9Stars | -- | -- | -- | -- | |
DiffusionGemma 26B A4B | 256k | Google | -- | -- | -- | -- | |
GPT-5.5 Instant (June 2026) | 400k | OpenAI | -- | -- | -- | 1.34 | |
Falcon-H1R-7B | 256k |  TII UAE | -- | -- | -- | -- | |
Qwen3 Next 80B A3B | 262k | Alibaba | -- | -- | -- | -- | |
gpt-oss-120b (low) | 131k | OpenAI | -- | -- | -- | 2.23 | |
Llama Nemotron Ultra | 128k | NVIDIA | -- | -- | -- | -- | |
Gemma 4 31B (non-reasoning) | 256k | Google | -- | -- | -- | 4.12 | |
MiniCPM5-1B (non-reasoning) | 128k | OpenBMB | -- | -- | -- | -- | |
Hermes 4 405B (non-reasoning) | 128k | Nous Research | -- | -- | -- | 5.35 | |
Exaone 4.0 1.2B | 64k | LG AI Research | -- | -- | -- | -- | |
Gemma 4 E4B | 128k | Google | -- | -- | -- | 59.61 | |
K2-V2 (low) | 512k | Institute of Foundation Models | -- | -- | -- | -- | |
Granite 4.0 H 1B | 128k | IBM | -- | -- | -- | -- | |
Reka Flash 3 | 128k |  Reka AI | -- | -- | -- | 8.72 | |
K2-V2 (medium) | 512k | Institute of Foundation Models | -- | -- | -- | -- | |
Mi:dm K 2.5 Pro | 128k |  Korea Telecom | -- | -- | -- | -- | |
LFM2.5-2.6B | 128k | Liquid AI | -- | -- | -- | -- | |
LFM2.5-8B-A1B | 33k | Liquid AI | -- | -- | -- | -- | |
GPT-5.5 Pro (xhigh) | 922k | OpenAI | -- | -- | -- | -- | |
Nova Premier | 1M | Amazon | -- | -- | -- | -- | |
Cogito v2.1 | 128k |  Deep Cogito | -- | -- | -- | -- | |
Ring-2.6-1T | 262k | InclusionAI | -- | -- | -- | 9.66 | |
Phi-4 | 16k | Microsoft | -- | -- | -- | 1.85 | |
gpt-oss-120b (high) | 131k | OpenAI | -- | -- | -- | 13.68 | |
NVIDIA Nemotron Nano 9B V2 | 131k | NVIDIA | -- | -- | -- | 0.86 | |
Agnes 2.5 Pro Beta | 1M |  Sapiens AI | -- | -- | -- | -- | |
Nova 2.0 Pro Preview (low) | 256k | Amazon | -- | -- | -- | 6.76 | |
Jamba Reasoning 3B | 262k | AI21 Labs | -- | -- | -- | -- | |
Quasar 438B (max) | 1M | Multiverse Computing | -- | -- | -- | 11.76 | |
LFM2 24B A2B | 33k | Liquid AI | -- | -- | -- | -- | |
Nova 2.0 Lite (medium) | 1M | Amazon | -- | -- | -- | 19.10 | |
KAT-Coder-Pro V2 | 256k |  KwaiKAT | -- | -- | -- | -- | |
Qwen3 Coder Next | 256k | Alibaba | -- | -- | -- | 1.39 | |
Gemma 4 12B | 262k | Google | -- | -- | -- | 33.26 | |
关键定义
常见问题
在参与排名的 96 个模型中，Gemini 4 Carbon (High) 目前以 60 的 Intelligence Index 得分位居 Artificial Analysis LLM 排行榜第一。
按 Intelligence Index 排名，顶尖模型是：1. Gemini 4 Carbon (High)（60）, 2. Claude Opus 5.5 (Max, Default Fallback)（59）, 3. Gemini 4 Carbon (Medium)（58）, 4. Claude Opus 5.5 (Xhigh, Default Fallback)（57）, 5. GPT-6 Astra (Max)（57）。
GPT-6 Luna (Low) 最快，每任务耗时最短，完成一项 Intelligence Index 任务平均需 0.6 分钟，其次是 GPT-6.1 Terra (Low)（1.0 分钟）和 GPT-6.1 Sol (Low)（1.7 分钟）。越低越好。
K2 Horizon 375B A23B 的每项 Intelligence Index 任务成本最低，为 $0.00，其次是 Command A+（$0.00）和 GPT-6 Luna (Low)（$0.01）。
DeepSeek V4.1 Flash 1017 (Max) 是排名最高的开放权重模型，Intelligence Index 得分为 33。排行榜上的 96 个模型中，有 37 个开放权重模型。
按 Intelligence Index 排名，顶尖开放权重模型是：1. DeepSeek V4.1 Flash 1017 (Max)（33）, 2. GLM-5.3 (Max)（32）, 3. MiMo-V2.6-Pro（30）。
在 91 个推理模型中，Gemini 4 Carbon (High) 以 60 的 Intelligence Index 得分领先。推理模型会先通过扩展思考来解决复杂问题，再给出回答。
排行榜提供筛选条件，可按模型类型（推理与非推理）、开放性（开放权重与专有）及其他标准缩小结果范围。
点击排行榜中的任意模型名称，即可访问其专属比较页面，查看智能、价格、速度、延迟等详细图表。您还可以比较每个模型的 API 服务商。 查看所有模型