llmspeed llms should run on premises

y intelligence indexx output tokens/s, log● api● open weights ○ no index yet— same model, more or less reasoning◌ new this fortnight
0204060 101001,000 output tokens/s → ↑ intelligence index Gemini 4 Carbon (High) · Google 91 tokens/s · intelligence 60 · $4 in / $20 out per 1MGemini 4 Carbon (High) Claude Opus 5.5 (Max) · Anthropic 95 tokens/s · intelligence 59.1 · $4 in / $20 out per 1MClaude Opus 5.5 (Max) Gemini 4 Carbon (Medium) · Google 93 tokens/s · intelligence 58.3 · $4 in / $20 out per 1M Claude Opus 5.5 (Xhigh) · Anthropic 81 tokens/s · intelligence 57.4 · $4 in / $20 out per 1M GPT-6 Astra (Max) · OpenAI 52 tokens/s · intelligence 56.9 · $10 in / $50 out per 1M GPT-6 Astra (Xhigh) · OpenAI 49 tokens/s · intelligence 55.3 · $10 in / $50 out per 1M GPT-6 Astra (High) · OpenAI 46 tokens/s · intelligence 54.8 · $10 in / $50 out per 1M GPT-6.1 Sol (Max) · OpenAI 62 tokens/s · intelligence 54.6 · $2 in / $10 out per 1M Claude Sonnet 5.5 (Max) · Anthropic 136 tokens/s · intelligence 54.5 · $2 in / $10 out per 1MClaude Sonnet 5.5 (Max) Gemini 4 Carbon (Low) · Google 86 tokens/s · intelligence 54.2 · $4 in / $20 out per 1M GPT-6.1 Sol (Xhigh) · OpenAI 67 tokens/s · intelligence 52.5 · $2 in / $10 out per 1M Claude Opus 5.5 (High) · Anthropic 76 tokens/s · intelligence 51.8 · $4 in / $20 out per 1M GPT-6 Astra (Medium) · OpenAI 42 tokens/s · intelligence 51.6 · $10 in / $50 out per 1M GPT-6.1 Terra (Max) · OpenAI 112 tokens/s · intelligence 51.3 · $1 in / $5 out per 1M Gemini 4 Argon (High) · Google 91 tokens/s · intelligence 50.4 · $4 in / $20 out per 1M GPT-6.1 Sol (High) · OpenAI 59 tokens/s · intelligence 50.3 · $2 in / $10 out per 1M Claude Fable 5.1 (Xhigh) · Anthropic 62 tokens/s · intelligence 50 · $10 in / $50 out per 1M Claude Sonnet 5.5 (Xhigh) · Anthropic 106 tokens/s · intelligence 49.9 · $2 in / $10 out per 1M Claude Fable 5.1 (Max) · Anthropic 68 tokens/s · intelligence 49.2 · $10 in / $50 out per 1M Claude Opus 5.5 (Medium) · Anthropic 74 tokens/s · intelligence 48.2 · $4 in / $20 out per 1M Claude Fable 5.1 (High) · Anthropic 55 tokens/s · intelligence 47.2 · $10 in / $50 out per 1M Kimi K3.1 Pro · Kimi · open weights 27 tokens/s · intelligence 46.5 · $2 in / $10 out per 1M GPT-6.1 Sol (Medium) · OpenAI 53 tokens/s · intelligence 46.3 · $2 in / $10 out per 1M GPT-6.1 Terra (Xhigh) · OpenAI 89 tokens/s · intelligence 45.3 · $1 in / $5 out per 1M Claude Opus 5 (Max) · Anthropic 50 tokens/s · intelligence 44.2 · $5 in / $25 out per 1M GPT-6 Sol (Max) · OpenAI 127 tokens/s · intelligence 44.2 · $2 in / $10 out per 1M Muse Wonder (Max) · Meta 140 tokens/s · intelligence 43.6 · $4 in / $20 out per 1MMuse Wonder (Max) GPT-6 Astra (Low) · OpenAI 41 tokens/s · intelligence 43.3 · $10 in / $50 out per 1M Claude Fable 5.1 (Medium) · Anthropic 52 tokens/s · intelligence 43.2 · $10 in / $50 out per 1M GPT-6.1 Terra (High) · OpenAI 87 tokens/s · intelligence 42.4 · $1 in / $5 out per 1M Claude Sonnet 5.5 (High) · Anthropic 104 tokens/s · intelligence 41.5 · $2 in / $10 out per 1M Muse Wonder (High) · Meta 164 tokens/s · intelligence 40.5 · $4 in / $20 out per 1MMuse Wonder (High) Claude Fable 5.1 (Low) · Anthropic 53 tokens/s · intelligence 38.7 · $10 in / $50 out per 1M GPT-6.1 Terra (Medium) · OpenAI 89 tokens/s · intelligence 38.2 · $1 in / $5 out per 1M Muse Spark 1.3 (Max) · Meta 146 tokens/s · intelligence 36.6 · $1.25 in / $4.25 out per 1M Claude Haiku 5.5 (Max) · Anthropic 233 tokens/s · intelligence 36.5 · $0.1 in / $0.5 out per 1MClaude Haiku 5.5 (Max) Grok 4.7 (Xhigh) · SpaceXAI 75 tokens/s · intelligence 35.5 · $2 in / $6 out per 1M Grok 4.7 (High) · SpaceXAI 65 tokens/s · intelligence 35.2 · $2 in / $6 out per 1M Claude Opus 5.5 (Low) · Anthropic 76 tokens/s · intelligence 34.8 · $4 in / $20 out per 1M GPT-6.1 Sol (Low) · OpenAI 51 tokens/s · intelligence 34.8 · $2 in / $10 out per 1M Claude Haiku 5.5 (Xhigh) · Anthropic 178 tokens/s · intelligence 34.6 · $0.1 in / $0.5 out per 1M Tencent Hy4 · Tencent · open weights 94 tokens/s · intelligence 34.6 · $0.834 in / $2.5 out per 1M Qwen3.8 Max (0902) · Alibaba 39 tokens/s · intelligence 33.7 · $2 in / $6 out per 1M DeepSeek V4.1 Flash 1017 (Max) · DeepSeek · open weights 162 tokens/s · intelligence 32.6 · $0.3 in / $1.2 out per 1M GPT-5.6 Terra (Max) · OpenAI 116 tokens/s · intelligence 32.3 · $2 in / $12 out per 1M GLM-5.3 (Max) · Z AI · open weights 75 tokens/s · intelligence 31.7 · $1.4 in / $4.4 out per 1M Grok 4.7 (Low) · SpaceXAI 55 tokens/s · intelligence 31.1 · $2 in / $6 out per 1M Grok 4.6 (Xhigh) · SpaceXAI 64 tokens/s · intelligence 31 · $2 in / $6 out per 1M Claude Sonnet 5.5 (Medium) · Anthropic 97 tokens/s · intelligence 30.8 · $2 in / $10 out per 1M Gemini 3.8 Flash (High) · Google 121 tokens/s · intelligence 30.8 · $0.75 in / $3.75 out per 1M Inkling 2 Small · Thinking Machines 139 tokens/s · intelligence 30.6 · $0.3 in / $1 out per 1M GPT-6.1 Terra (Non-reasoning) · OpenAI 78 tokens/s · intelligence 30.5 · $1 in / $5 out per 1M Gemini 3.8 Flash (Medium) · Google 136 tokens/s · intelligence 30.1 · $0.75 in / $3.75 out per 1M Claude Haiku 5.5 (High) · Anthropic 165 tokens/s · intelligence 29.9 · $0.1 in / $0.5 out per 1M MiMo-V2.6-Pro · Xiaomi · open weights 41 tokens/s · intelligence 29.9 · $0.435 in / $0.87 out per 1M GPT-6.1 Terra (Low) · OpenAI 84 tokens/s · intelligence 29.4 · $1 in / $5 out per 1M MiniMax-M3.1-Flash · MiniMax · open weights 152 tokens/s · intelligence 29.1 · $0.15 in / $0.6 out per 1M DeepSeek V4.1 Flash 1017 (High) · DeepSeek · open weights 163 tokens/s · intelligence 28.6 · $0.3 in / $1.2 out per 1M Step 5 Preview · StepFun · open weights 150 tokens/s · intelligence 28.5 · $1 in / $2.7 out per 1M GPT-6 Luna (Max) · OpenAI 123 tokens/s · intelligence 27.9 · $0.1 in / $0.5 out per 1M GLM-5.3-Flash · Z AI · open weights 55 tokens/s · intelligence 27.8 · $0.15 in / $0.5 out per 1M MiMo-V2.6-Flash · Xiaomi · open weights 37 tokens/s · intelligence 26.7 · $0.14 in / $0.28 out per 1M Mistral Large 4 (High) · Mistral · open weights 117 tokens/s · intelligence 26.1 · $0.7 in / $2.1 out per 1M Claude Sonnet 5.5 (Low) · Anthropic 100 tokens/s · intelligence 25.6 · $2 in / $10 out per 1M Mistral Large 4 Preview · Mistral 124 tokens/s · intelligence 25.3 · $0.7 in / $2.1 out per 1M Muse Spark Open (Xhigh) · Meta · open weights 214 tokens/s · intelligence 24.9 · $1.25 in / $4.25 out per 1M Claude Haiku 5.5 (Medium) · Anthropic 166 tokens/s · intelligence 24.6 · $0.1 in / $0.5 out per 1M GPT-5.6 Luna (Max) · OpenAI 125 tokens/s · intelligence 23.8 · $0.2 in / $1.2 out per 1M GPT-6 Luna (Xhigh) · OpenAI 118 tokens/s · intelligence 23.5 · $0.1 in / $0.5 out per 1M Mistral Large 4 (Medium) · Mistral · open weights 128 tokens/s · intelligence 23.1 · $0.7 in / $2.1 out per 1M DeepSeek V4 Pro 0813 (Max) · DeepSeek · open weights 79 tokens/s · intelligence 23.1 · $1.32 in / $3.96 out per 1M GLM-5.3 (Low) · Z AI · open weights 72 tokens/s · intelligence 23.1 · $1.4 in / $4.4 out per 1M Gemini 3.8 Flash (Low) · Google 128 tokens/s · intelligence 22.5 · $0.75 in / $3.75 out per 1M GPT-6 Luna (High) · OpenAI 118 tokens/s · intelligence 21.4 · $0.1 in / $0.5 out per 1M Qwen3.8 27B (Xhigh) · Alibaba · open weights 37 tokens/s · intelligence 20.1 · $0.5 in / $3 out per 1M MiniMax-M3 · MiniMax · open weights 121 tokens/s · intelligence 18.9 · $0.3 in / $1.2 out per 1M Claude Haiku 5.5 (Low) · Anthropic 161 tokens/s · intelligence 18.6 · $0.1 in / $0.5 out per 1M GPT-6 Luna (Medium) · OpenAI 116 tokens/s · intelligence 18.5 · $0.1 in / $0.5 out per 1M DeepSeek V4.1 Flash 1017 (Non-reasoning) · DeepSeek · open weights 155 tokens/s · intelligence 18.5 · $0.3 in / $1.2 out per 1M K2 Horizon 375B A23B · Institute of Foundation Models 277 tokens/s · intelligence 17.8 · $0 in / $0 out per 1MK2 Horizon 375B A23B Beam 1.0 · Reflection · open weights 348 tokens/s · intelligence 17.7 · $0.4 in / $1.2 out per 1MBeam 1.0 Qwen3.8 27B (Medium) · Alibaba · open weights 39 tokens/s · intelligence 17.5 · $0.5 in / $3 out per 1M Kimi K3 (Low) · Kimi · open weights 50 tokens/s · intelligence 17.4 · $3 in / $15 out per 1M Qwen3.8 27B (Low) · Alibaba · open weights 38 tokens/s · intelligence 15.7 · $0.5 in / $3 out per 1M Nemotron 3.5 Super VL · NVIDIA · open weights 266 tokens/s · intelligence 13.6 · $0.175 in / $0.5 out per 1M Inkling Small · Thinking Machines · open weights 81 tokens/s · intelligence 12.9 · $0.3 in / $1.2 out per 1M DeepSeek V4 Pro 0813 (Non-reasoning) · DeepSeek · open weights 78 tokens/s · intelligence 12.6 · $1.32 in / $3.96 out per 1M Inkling (Xhigh) · Thinking Machines · open weights 139 tokens/s · intelligence 12.5 · $1 in / $4.05 out per 1M Nemotron 3 Ultra 550B A55B (Reasoning) · NVIDIA · open weights 230 tokens/s · intelligence 11.8 · $0.6 in / $2.7 out per 1M Gemini 3.5 Flash-Lite · Google 392 tokens/s · intelligence 11.7 · $0.3 in / $2.5 out per 1MGemini 3.5 Flash-Lite Qwen3.8 27B (Non-reasoning) · Alibaba · open weights 36 tokens/s · intelligence 10 · $0.5 in / $3 out per 1M GPT-6 Luna (Low) · OpenAI 117 tokens/s · intelligence 9.7 · $0.1 in / $0.5 out per 1M GPT-6 Luna (Non-reasoning) · OpenAI 110 tokens/s · intelligence 9.4 · $0.1 in / $0.5 out per 1M Claude 4.5 Haiku (Reasoning) · Anthropic 88 tokens/s · intelligence 7.4 · $1 in / $5 out per 1M Muse Glimmer (High) · Meta · open weights 108 tokens/s · intelligence 6.3 · $0.3 in / $1.2 out per 1M Command A+ · Cohere 247 tokens/s · intelligence 6.2 · $0 in / $0 out per 1M Gemini 3.1 Flash-Lite · Google 319 tokens/s · intelligence 6.2 · $0.25 in / $1.5 out per 1M Nemotron 3.5 Lightning · NVIDIA · open weights 293 tokens/s · intelligence 5.5 · $0.06 in / $0.2 out per 1M Granite 4.2 8B · IBM · open weights 124 tokens/s · intelligence 5.1 · $0.06 in / $0.25 out per 1M Mistral Medium 3.5 · Mistral 172 tokens/s · intelligence 4.6 · $1.5 in / $7.5 out per 1M Gemma 4 31B (Reasoning) · Google · open weights 35 tokens/s · intelligence 4.4 · $0 in / $0 out per 1M Mistral Small 4 (Reasoning) · Mistral · open weights 163 tokens/s · intelligence 4.3 · $0.15 in / $0.6 out per 1M North Mini Code · Cohere · open weights 99 tokens/s · intelligence 3.4 · $0 in / $0 out per 1M Celeris-1 · Celeris 1518 tokens/s · no intelligence index yet · $0.2 in / $0.7 out per 1MCeleris-1

try it — click a dot: the model opens on arena.ai (or OpenRouter chat when arena does not list it) and this prompt lands in your clipboard, paste it there:

Explain in three sentences why the sky is blue, then give me one surprising fact about it.
draw a dotclick the chart to fill in the axes. dots live in a cookie - only you see them.
new model?
all 104 models
modelintelligencetok/s$ in$ outsince
Gemini 4 Carbon (High) Google●60914202026-10try
Claude Opus 5.5 (Max) Anthropic●59.1954202026-09try
Gemini 4 Carbon (Medium) Google●58.3934202026-10try
Claude Opus 5.5 (Xhigh) Anthropic●57.4814202026-09try
GPT-6 Astra (Max) OpenAI●56.95210502026-09try
GPT-6 Astra (Xhigh) OpenAI●55.34910502026-09try
GPT-6 Astra (High) OpenAI●54.84610502026-09try
GPT-6.1 Sol (Max) OpenAI●54.6622102026-09try
Claude Sonnet 5.5 (Max) Anthropic●54.51362102026-09try
Gemini 4 Carbon (Low) Google●54.2864202026-10try
GPT-6.1 Sol (Xhigh) OpenAI●52.5672102026-09try
Claude Opus 5.5 (High) Anthropic●51.8764202026-09try
GPT-6 Astra (Medium) OpenAI●51.64210502026-09try
GPT-6.1 Terra (Max) OpenAI●51.3112152026-10try
Gemini 4 Argon (High) Google●50.4914202026-09try
GPT-6.1 Sol (High) OpenAI●50.3592102026-09try
Claude Fable 5.1 (Xhigh) Anthropic●506210502026-09try
Claude Sonnet 5.5 (Xhigh) Anthropic●49.91062102026-09try
Claude Fable 5.1 (Max) Anthropic●49.26810502026-09try
Claude Opus 5.5 (Medium) Anthropic●48.2744202026-09try
Claude Fable 5.1 (High) Anthropic●47.25510502026-09try
Kimi K3.1 Pro Kimi●46.5272102026-10try
GPT-6.1 Sol (Medium) OpenAI●46.3532102026-09try
GPT-6.1 Terra (Xhigh) OpenAI●45.389152026-10try
Claude Opus 5 (Max) Anthropic●44.2505252026-07try
GPT-6 Sol (Max) OpenAI●44.21272102026-09try
Muse Wonder (Max) Meta●43.61404202026-10try
GPT-6 Astra (Low) OpenAI●43.34110502026-09try
Claude Fable 5.1 (Medium) Anthropic●43.25210502026-09try
GPT-6.1 Terra (High) OpenAI●42.487152026-10try
Claude Sonnet 5.5 (High) Anthropic●41.51042102026-09try
Muse Wonder (High) Meta●40.51644202026-10try
Claude Fable 5.1 (Low) Anthropic●38.75310502026-09try
GPT-6.1 Terra (Medium) OpenAI●38.289152026-10try
Muse Spark 1.3 (Max) Meta●36.61461.254.252026-09try
Claude Haiku 5.5 (Max) Anthropic●36.52330.10.52026-10try
Grok 4.7 (Xhigh) SpaceXAI●35.575262026-09try
Grok 4.7 (High) SpaceXAI●35.265262026-09try
Claude Opus 5.5 (Low) Anthropic●34.8764202026-09try
GPT-6.1 Sol (Low) OpenAI●34.8512102026-09try
Claude Haiku 5.5 (Xhigh) Anthropic●34.61780.10.52026-10try
Tencent Hy4 Tencent●34.6940.8342.52026-11try
Qwen3.8 Max (0902) Alibaba●33.739262026-09try
DeepSeek V4.1 Flash 1017 (Max) DeepSeek●32.61620.31.22026-10try
GPT-5.6 Terra (Max) OpenAI●32.31162122026-07try
GLM-5.3 (Max) Z AI●31.7751.44.42026-08try
Grok 4.7 (Low) SpaceXAI●31.155262026-09try
Grok 4.6 (Xhigh) SpaceXAI●3164262026-08try
Claude Sonnet 5.5 (Medium) Anthropic●30.8972102026-09try
Gemini 3.8 Flash (High) Google●30.81210.753.752026-09try
Inkling 2 Small Thinking Machines●30.61390.312026-11try
GPT-6.1 Terra (Non-reasoning) OpenAI●30.578152026-10try
Gemini 3.8 Flash (Medium) Google●30.11360.753.752026-09try
Claude Haiku 5.5 (High) Anthropic●29.91650.10.52026-10try
MiMo-V2.6-Pro Xiaomi●29.9410.4350.872026-09try
GPT-6.1 Terra (Low) OpenAI●29.484152026-10try
MiniMax-M3.1-Flash MiniMax●29.11520.150.62026-10try
DeepSeek V4.1 Flash 1017 (High) DeepSeek●28.61630.31.22026-10try
Step 5 Preview StepFun●28.515012.72026-09try
GPT-6 Luna (Max) OpenAI●27.91230.10.52026-09try
GLM-5.3-Flash Z AI●27.8550.150.52026-08try
MiMo-V2.6-Flash Xiaomi●26.7370.140.282026-09try
Mistral Large 4 (High) Mistral●26.11170.72.12026-10try
Claude Sonnet 5.5 (Low) Anthropic●25.61002102026-09try
Mistral Large 4 Preview Mistral●25.31240.72.12026-10try
Muse Spark Open (Xhigh) Meta●24.92141.254.252026-10try
Claude Haiku 5.5 (Medium) Anthropic●24.61660.10.52026-10try
GPT-5.6 Luna (Max) OpenAI●23.81250.21.22026-07try
GPT-6 Luna (Xhigh) OpenAI●23.51180.10.52026-09try
Mistral Large 4 (Medium) Mistral●23.11280.72.12026-10try
DeepSeek V4 Pro 0813 (Max) DeepSeek●23.1791.323.962026-08try
GLM-5.3 (Low) Z AI●23.1721.44.42026-08try
Gemini 3.8 Flash (Low) Google●22.51280.753.752026-09try
GPT-6 Luna (High) OpenAI●21.41180.10.52026-09try
Qwen3.8 27B (Xhigh) Alibaba●20.1370.532026-08try
MiniMax-M3 MiniMax●18.91210.31.22026-06try
Claude Haiku 5.5 (Low) Anthropic●18.61610.10.52026-10try
GPT-6 Luna (Medium) OpenAI●18.51160.10.52026-09try
DeepSeek V4.1 Flash 1017 (Non-reasoning) DeepSeek●18.51550.31.22026-10try
K2 Horizon 375B A23B Institute of Foundation Models●17.8277002026-09try
Beam 1.0 Reflection●17.73480.41.22026-10try
Qwen3.8 27B (Medium) Alibaba●17.5390.532026-08try
Kimi K3 (Low) Kimi●17.4503152026-07try
Qwen3.8 27B (Low) Alibaba●15.7380.532026-08try
Nemotron 3.5 Super VL NVIDIA●13.62660.1750.52026-10try
Inkling Small Thinking Machines●12.9810.31.22026-07try
DeepSeek V4 Pro 0813 (Non-reasoning) DeepSeek●12.6781.323.962026-08try
Inkling (Xhigh) Thinking Machines●12.513914.052026-07try
Nemotron 3 Ultra 550B A55B (Reasoning) NVIDIA●11.82300.62.72026-06try
Gemini 3.5 Flash-Lite Google●11.73920.32.52026-07try
Qwen3.8 27B (Non-reasoning) Alibaba●10360.532026-08try
GPT-6 Luna (Low) OpenAI●9.71170.10.52026-09try
GPT-6 Luna (Non-reasoning) OpenAI●9.41100.10.52026-09try
Claude 4.5 Haiku (Reasoning) Anthropic●7.488152025-10try
Muse Glimmer (High) Meta●6.31080.31.22026-08try
Command A+ Cohere●6.2247002026-05try
Gemini 3.1 Flash-Lite Google●6.23190.251.52026-03try
Nemotron 3.5 Lightning NVIDIA●5.52930.060.22026-08try
Granite 4.2 8B IBM●5.11240.060.252026-08try
Mistral Medium 3.5 Mistral●4.61721.57.52026-04try
Gemma 4 31B (Reasoning) Google●4.435002026-04try
Mistral Small 4 (Reasoning) Mistral●4.31630.150.62026-03try
North Mini Code Cohere●3.499002026-06try
Celeris-1 Celeris●15180.20.72026-07try