Modelos de IA mais rápidos: velocidade de saída
Modelos ordenados por tokens de saída por segundo. Essa velocidade difere do tempo até a primeira resposta e do tempo total, principalmente em modelos de raciocínio. Medições ausentes são excluídas.
Qual escolher?
Nenhum benchmark cobre todos os usos. Teste suas próprias perguntas e compare precisão, tempo de resposta, entradas e saídas aceitas e a conta real.
Modelos de IA
| Modelo | Pontuação de inteligência | Custo do benchmark (USD) | Preço de entrada / 1 milhão de tokens (USD) | Preço de saída / 1 milhão de tokens (USD) | Velocidade de saída |
|---|---|---|---|---|---|
| Celeris-1Celeris | 12.40 | $157 | 0.20 | 0.70 | Tok/s 1596 |
| Mercury 2Inception | 21.90 | $97.5 | 0.25 | 0.75 | Tok/s 835 |
| LFM2.5-8B-A1BLiquid AI | 8.10 | - | 0.00 | 0.00 | Tok/s 339 |
| Gemini 3.5 Flash-LiteGoogle | 37.40 | $153 | 0.30 | 2.50 | Tok/s 332 |
| Gemini 3.7 Flash (high)GoogleMelhor custo-benefício | 56.00 | $485 | 0.75 | 3.75 | Tok/s 323 |
| Nemotron 3 Nano Omni 30B A3B ReasoningNVIDIA | 15.00 | - | 0.07 | 0.30 | Tok/s 323 |
| Gemini 3.7 Flash (medium)GoogleMelhor custo-benefício | 53.40 | $278 | 0.75 | 3.75 | Tok/s 317 |
| Gemini 2.5 Flash-Lite (Reasoning)Google | 11.40 | - | 0.10 | 0.40 | Tok/s 312 |
| Gemini 3.7 Flash (low)GoogleMelhor custo-benefício | 50.90 | $165 | 0.75 | 3.75 | Tok/s 303 |
| Gemini 3.1 Flash-LiteGoogle | 25.60 | $93.8 | 0.25 | 1.50 | Tok/s 303 |
| Nemotron 3.5 LightningNVIDIA | 23.60 | $71.5 | 0.07 | 0.22 | Tok/s 302 |
| Nova MicroAmazon | 4.40 | - | 0.04 | 0.14 | Tok/s 282 |
| LFM2.5-VL-1.6BLiquid AI | 1.00 | - | 0.00 | 0.00 | Tok/s 265 |
| Qwen3.5 Omni FlashAlibaba | 19.20 | - | 0.10 | 0.80 | Tok/s 238 |
| Gemini 2.5 Flash-Lite (Non-reasoning)Google | 6.70 | - | 0.10 | 0.40 | Tok/s 232 |
| NVIDIA Nemotron 3 Nano 30B A3B (Reasoning)NVIDIA | 14.50 | $39.4 | 0.05 | 0.20 | Tok/s 230 |
| NVIDIA Nemotron 3 Nano 30B A3B (Non-reasoning)NVIDIA | 7.20 | - | 0.05 | 0.20 | Tok/s 225 |
| Gemini 2.5 Flash (Reasoning)Google | 20.30 | - | 0.30 | 2.50 | Tok/s 225 |
| Step 3.5 FlashStepFun | 26.00 | - | 0.10 | 0.30 | Tok/s 224 |
| o3-miniOpenAI | 19.20 | - | 1.10 | 4.40 | Tok/s 221 |
| o3-mini (high)OpenAI | 15.70 | - | 1.10 | 4.40 | Tok/s 220 |
| Step 3.5 Flash 2603StepFun | 26.50 | - | 0.10 | 0.30 | Tok/s 214 |
| Ministral 3 3BMistral | 7.10 | $91.6 | 0.10 | 0.10 | Tok/s 203 |
| Qwen3.7 MaxAlibaba | 46.70 | $1064 | 2.50 | 7.50 | Tok/s 201 |
| Gemini 2.5 Flash (Non-reasoning)Google | 14.20 | - | 0.30 | 2.50 | Tok/s 200 |
| Muse Spark 1.1 (xhigh)Meta | 53.20 | $547 | 1.25 | 4.25 | Tok/s 200 |
| Gemini 3.6 Flash (high)Google | 51.60 | $412 | 0.75 | 3.75 | Tok/s 199 |
| Qwen3 Next 80B A3B (Reasoning)Alibaba | 16.90 | $84.0 | 0.15 | 1.20 | Tok/s 198 |
| Trinity Large ThinkingArcee AI | 18.70 | $255 | 0.23 | 0.88 | Tok/s 197 |
| NVIDIA Nemotron Nano 9B V2 (Reasoning)NVIDIA | 8.70 | - | 0.04 | 0.16 | Tok/s 186 |
| GPT-5.4 nano (xhigh)OpenAI | 39.70 | $277 | 0.20 | 1.25 | Tok/s 185 |
| Qwen3 Next 80B A3B InstructAlibaba | 13.80 | - | 0.15 | 1.20 | Tok/s 183 |
| Gemini 3.5 Flash (medium)Google | 46.70 | - | 1.50 | 9.00 | Tok/s 180 |
| GPT-5.4 nano (medium)OpenAI | 30.80 | - | 0.20 | 1.25 | Tok/s 177 |
| Gemini 3 Flash Preview (Non-reasoning)Google | 27.90 | - | 0.50 | 3.00 | Tok/s 171 |
| Gemini 3.5 Flash (high)Google | 52.00 | $1042 | 1.50 | 9.00 | Tok/s 171 |
| gpt-oss-20b (low)OpenAI | 14.40 | - | 0.07 | 0.20 | Tok/s 170 |
| Nova 2.0 Lite (medium)Amazon | 19.20 | - | 0.30 | 2.50 | Tok/s 170 |
| Qwen3.5 35B A3B (Non-reasoning)Alibaba | 24.30 | $228 | 0.25 | 2.00 | Tok/s 169 |
| GPT-5.4 nano (Non-Reasoning)OpenAI | 17.80 | - | 0.20 | 1.25 | Tok/s 168 |
| NVIDIA Nemotron Nano 9B V2 (Non-reasoning)NVIDIA | 7.20 | - | 0.05 | 0.20 | Tok/s 168 |
| GPT-5 nano (high)OpenAI | 20.10 | - | 0.05 | 0.40 | Tok/s 167 |
| Nova 2.0 Lite (low)Amazon | 18.00 | - | 0.30 | 2.50 | Tok/s 164 |
| GPT-5.4 mini (xhigh)OpenAI | 40.90 | $1096 | 0.75 | 4.50 | Tok/s 163 |
| NVIDIA Nemotron Nano 12B v2 VL (Non-reasoning)NVIDIA | 4.20 | - | 0.20 | 0.60 | Tok/s 163 |
| GPT-5.4 mini (Non-Reasoning)OpenAI | 16.80 | - | 0.75 | 4.50 | Tok/s 162 |
| GPT-5 nano (minimal)OpenAI | 7.80 | - | 0.05 | 0.40 | Tok/s 161 |
| Gemini 3 Flash Preview (Reasoning)Google | 38.70 | - | 0.50 | 3.00 | Tok/s 161 |
| gpt-oss-120b (high)OpenAI | 24.10 | $95.0 | 0.15 | 0.60 | Tok/s 161 |
| gpt-oss-120b (low)OpenAI | 14.90 | $24.2 | 0.15 | 0.59 | Tok/s 160 |
| Mistral Small 4 (Reasoning)Mistral | 19.70 | $87.7 | 0.15 | 0.60 | Tok/s 160 |
| GPT-5.4 mini (medium)OpenAI | 30.50 | - | 0.75 | 4.50 | Tok/s 160 |
| GPT-5 nano (medium)OpenAI | 19.20 | - | 0.05 | 0.40 | Tok/s 159 |
| GPT-5.6 Luna (max)OpenAIMelhor custo-benefício | 52.30 | $172 | 0.20 | 1.20 | Tok/s 159 |
| GPT-5.6 Luna (xhigh)OpenAIMelhor custo-benefício | 50.10 | $95.1 | 0.20 | 1.20 | Tok/s 158 |
| Mistral Small (Sep '24)Mistral | 4.30 | - | 0.20 | 0.60 | Tok/s 156 |
| Nova LiteAmazon | 6.70 | - | 0.06 | 0.24 | Tok/s 155 |
| Gemini 3.5 Flash (minimal)Google | 35.80 | - | 1.50 | 9.00 | Tok/s 154 |
| Nova 2.0 Lite (Non-reasoning)Amazon | 11.80 | - | 0.30 | 2.50 | Tok/s 153 |
| Nova 2.0 Lite (high)Amazon | 20.80 | - | 0.30 | 2.50 | Tok/s 152 |
| Solar Pro 3Upstage | 14.50 | $158 | 0.15 | 0.60 | Tok/s 152 |
| Qwen3.5 35B A3B (Reasoning)Alibaba | 29.90 | - | 0.25 | 2.00 | Tok/s 151 |
| Mistral Small 3.2Mistral | 10.70 | $129 | 0.10 | 0.30 | Tok/s 151 |
| Mistral Small (Feb '24)Mistral | 3.20 | - | 0.15 | 0.60 | Tok/s 150 |
| Mistral Small 3.1Mistral | 14.90 | $29.8 | 0.10 | 0.30 | Tok/s 149 |
| GPT-5.6 Luna (Non-reasoning)OpenAIMelhor custo-benefício | 26.80 | $10.2 | 0.20 | 1.20 | Tok/s 145 |
| GPT-5.6 Luna (high)OpenAIMelhor custo-benefício | 47.00 | $55.0 | 0.20 | 1.20 | Tok/s 144 |
| GPT-5.6 Luna (medium)OpenAIMelhor custo-benefício | 38.90 | $21.1 | 0.20 | 1.20 | Tok/s 144 |
| Magistral Small 1.2Mistral | 11.50 | $301 | 0.50 | 1.50 | Tok/s 143 |
| Qwen3.5 122B A10B (Non-reasoning)Alibaba | 28.20 | $221 | 0.40 | 3.20 | Tok/s 143 |
| o4-mini (high)OpenAI | 26.10 | - | 1.10 | 4.40 | Tok/s 143 |
| Mistral Small 3Mistral | 6.70 | - | 0.10 | 0.30 | Tok/s 142 |
| GPT-4.1OpenAI | 19.60 | - | 2.00 | 8.00 | Tok/s 142 |
| Qwen3 30B A3B 2507 InstructAlibaba | 8.90 | - | 0.20 | 0.80 | Tok/s 140 |
| GLM-5.2 (Non-reasoning)Z AI | 34.80 | - | 1.40 | 4.40 | Tok/s 140 |
| Nex-N2-ProNex AGI | 41.70 | - | 0.50 | 2.50 | Tok/s 138 |
| Qwen3.6 35B A3B (Non-reasoning)Alibaba | 24.60 | $530 | 0.38 | 2.25 | Tok/s 138 |
| Qwen3 30B A3B 2507 (Reasoning)Alibaba | 14.60 | $152 | 0.20 | 2.40 | Tok/s 138 |
| GPT-5.6 Luna (low)OpenAIMelhor custo-benefício | 33.90 | $13.7 | 0.20 | 1.20 | Tok/s 138 |
| Mistral Small 4 (Non-reasoning)Mistral | 12.30 | - | 0.15 | 0.60 | Tok/s 137 |
| Llama 3.1 Instruct 8BMeta | 7.40 | - | 0.02 | 0.05 | Tok/s 137 |
| gpt-oss-20b (high)OpenAI | 15.20 | $32.7 | 0.06 | 0.19 | Tok/s 136 |
| Mistral MediumMistral | 3.20 | - | 1.50 | 7.50 | Tok/s 135 |
| Llama 4 ScoutMeta | 10.30 | $11.6 | 0.18 | 0.66 | Tok/s 135 |
| Gemini 2.5 ProGoogle | 25.90 | $611 | 1.25 | 10.00 | Tok/s 133 |
| Agnes 2.5 Pro AlphaSapiens AI | 39.70 | - | 0.45 | 0.90 | Tok/s 132 |
| Mistral Medium 3.5Mistral | 30.40 | $883 | 1.50 | 7.50 | Tok/s 132 |
| Nemotron 3 Ultra 550B A55B (Reasoning)NVIDIA | 38.30 | $534 | 0.60 | 2.75 | Tok/s 132 |
| GPT-4o (Nov '24)OpenAI | 11.10 | - | 2.50 | 10.00 | Tok/s 131 |
| GPT-5.4 (xhigh)OpenAI | 53.10 | $2183 | 2.50 | 15.00 | Tok/s 131 |
| Qwen3.5 122B A10B (Reasoning)Alibaba | 32.80 | $444 | 0.40 | 3.20 | Tok/s 130 |
| Nemotron 3 Super 120B A12B (Reasoning)NVIDIA | 25.70 | $254 | 0.20 | 0.80 | Tok/s 129 |
| Qwen3 Coder NextAlibaba | 21.30 | $250 | 0.35 | 1.20 | Tok/s 128 |
| Magistral Medium 1.2Mistral | 18.00 | $920 | 2.00 | 5.00 | Tok/s 127 |
| GPT-5.3 Codex (xhigh)OpenAI | 45.50 | - | 1.75 | 14.00 | Tok/s 125 |
| GPT-4.1 nanoOpenAI | 9.60 | $36.7 | 0.10 | 0.40 | Tok/s 125 |
| Mistral 7B InstructMistral | 1.70 | - | 0.25 | 0.25 | Tok/s 122 |
| Kimi K2 ThinkingKimi | 33.50 | - | 0.60 | 2.50 | Tok/s 121 |
| Ministral 3 8BMistral | 9.00 | $114 | 0.15 | 0.15 | Tok/s 120 |
| GPT-5.6 Terra (max)OpenAI | 56.60 | $1390 | 2.00 | 12.00 | Tok/s 119 |
| Qwen3.6 35B A3B (Reasoning)Alibaba | 32.10 | $505 | 0.38 | 2.25 | Tok/s 118 |
| Grok 4.3 (high)SpaceXAI | 37.90 | $288 | 1.25 | 2.50 | Tok/s 118 |
| Step 3.7 FlashStepFun | 30.90 | $320 | 0.20 | 1.15 | Tok/s 117 |
| GPT-5.5 Instant (June 2026)OpenAI | 29.20 | $423 | 5.00 | 30.00 | Tok/s 117 |
| Granite 4.1 8BIBM | 6.40 | - | 0.05 | 0.10 | Tok/s 115 |
| Nova 2.0 Pro Preview (medium)Amazon | 22.10 | - | 1.25 | 10.00 | Tok/s 115 |
| Nova 2.0 Pro Preview (low)Amazon | 19.80 | - | 1.25 | 10.00 | Tok/s 115 |
| Gemma 4 12B (Non-reasoning)Google | 13.20 | - | 0.10 | 0.30 | Tok/s 114 |
| Qwen3 VL 8B (Reasoning)Alibaba | 10.50 | - | 0.18 | 2.10 | Tok/s 114 |
| Llama 4 MaverickMeta | 14.50 | $35.9 | 0.27 | 0.85 | Tok/s 113 |
| Grok 4.3 (medium)SpaceXAI | 36.90 | - | 1.25 | 2.50 | Tok/s 113 |
| GLM-5.2 (max)Z AI | 52.60 | $843 | 1.40 | 4.40 | Tok/s 113 |
| o3OpenAI | 31.10 | - | 2.00 | 8.00 | Tok/s 112 |
| GPT-5.1 (high)OpenAI | 37.50 | $782 | 1.25 | 10.00 | Tok/s 112 |
| Qwen3 VL 30B A3B (Reasoning)Alibaba | 13.40 | - | 0.20 | 2.40 | Tok/s 111 |
| Muse Glimmer (high)Meta | 35.10 | $105 | 0.32 | 1.35 | Tok/s 111 |
| Gemma 4 12B (Reasoning)Google | 22.20 | - | 0.10 | 0.30 | Tok/s 109 |
| Qwen3 VL 30B A3B InstructAlibaba | 9.90 | - | 0.20 | 0.80 | Tok/s 109 |
| Llama 3.1 Nemotron Instruct 70BNVIDIA | 7.40 | - | 1.20 | 1.20 | Tok/s 109 |
| GLM-4.7 (Non-reasoning)Z AI | 27.10 | - | 0.60 | 2.20 | Tok/s 108 |
| Gemini 3.1 Pro PreviewGoogle | 47.70 | $812 | 2.00 | 12.00 | Tok/s 108 |
| Qwen2.5 TurboAlibaba | 6.00 | - | 0.05 | 0.20 | Tok/s 108 |
| Qwen3 VL 8B InstructAlibaba | 8.20 | - | 0.18 | 0.70 | Tok/s 107 |
| KAT Coder Pro V2KwaiKAT | 33.70 | - | 0.30 | 1.20 | Tok/s 107 |
| Qwen3 30B A3B (Non-reasoning)Alibaba | 6.60 | - | 0.20 | 0.80 | Tok/s 107 |
| GLM-4.7 (Reasoning)Z AI | 34.50 | $607 | 0.60 | 2.20 | Tok/s 107 |
| Mistral Medium 3.1Mistral | 14.70 | $134 | 0.40 | 2.00 | Tok/s 106 |
| Llama Nemotron Super 49B v1.5 (Reasoning)NVIDIA | 12.40 | - | 0.40 | 0.40 | Tok/s 106 |
| Qwen3 32B (Non-reasoning)Alibaba | 8.50 | - | 0.16 | 0.64 | Tok/s 104 |
| GPT-5.6 Terra (high)OpenAI | 50.10 | $395 | 2.00 | 12.00 | Tok/s 104 |
| GLM-4.7-Flash (Non-reasoning)Z AI | 15.60 | - | 0.07 | 0.40 | Tok/s 104 |
| DeepSeek V4 Flash 0731 (Reasoning, Max Effort)DeepSeek | 51.80 | $323 | 0.44 | 1.32 | Tok/s 103 |
| Qwen3 30B A3B (Reasoning)Alibaba | 9.20 | - | 0.20 | 2.40 | Tok/s 103 |
| GPT-5.6 Terra (xhigh)OpenAI | 52.80 | $590 | 2.00 | 12.00 | Tok/s 102 |
| GPT-4.1 miniOpenAI | 14.80 | $64.2 | 0.40 | 1.60 | Tok/s 102 |
| Qwen3 32B (Reasoning)Alibaba | 11.40 | - | 0.16 | 0.64 | Tok/s 101 |
| GPT-4o miniOpenAI | 6.70 | - | 0.15 | 0.60 | Tok/s 101 |
| NVIDIA Nemotron Nano 12B v2 VL (Reasoning)NVIDIA | 8.80 | - | 0.20 | 0.60 | Tok/s 101 |
| Grok 4.3 (low)SpaceXAI | 36.30 | - | 1.25 | 2.50 | Tok/s 100 |
| Qwen3 Omni 30B A3B (Reasoning)Alibaba | 9.50 | - | 0.25 | 0.97 | Tok/s 99 |
| Grok 4.20 0309 v2 (Reasoning)SpaceXAI | 38.00 | - | 1.25 | 2.50 | Tok/s 99 |
| GPT-5.6 Terra (Non-reasoning)OpenAI | 34.60 | $98.5 | 2.00 | 12.00 | Tok/s 98 |
| Grok 4.3 (Non-reasoning)SpaceXAI | 25.00 | $201 | 1.25 | 2.50 | Tok/s 98 |
| GPT-5.6 Terra (medium)OpenAI | 46.80 | $192 | 2.00 | 12.00 | Tok/s 98 |
| Nova 2.0 Pro Preview (Non-reasoning)Amazon | 14.40 | - | 1.25 | 10.00 | Tok/s 98 |
| GPT-5.1 (Non-reasoning)OpenAI | 20.70 | - | 1.25 | 10.00 | Tok/s 96 |
| MiniMax-M2.5MiniMax | 34.50 | - | 0.30 | 1.20 | Tok/s 94 |
| GPT-5.4 (low)OpenAI | 40.20 | - | 2.50 | 15.00 | Tok/s 94 |
| GPT-5 mini (minimal)OpenAI | 14.30 | - | 0.25 | 2.00 | Tok/s 94 |
| GPT-5.6 Terra (low)OpenAI | 41.30 | $130 | 2.00 | 12.00 | Tok/s 94 |
| Ministral 3 14BMistral | 11.20 | $119 | 0.20 | 0.20 | Tok/s 94 |
| GLM-4.7-Flash (Reasoning)Z AI | 23.30 | - | 0.07 | 0.40 | Tok/s 93 |
| Qwen3 Omni 30B A3B InstructAlibaba | 4.80 | - | 0.25 | 0.97 | Tok/s 93 |
| GPT-4o (Aug '24)OpenAI | 9.40 | - | 2.50 | 10.00 | Tok/s 93 |
| GLM-4.5V (Non-reasoning)Z AI | 6.80 | - | 0.60 | 1.80 | Tok/s 92 |
| Ling 2.6 FlashInclusionAI | 14.20 | - | 0.10 | 0.30 | Tok/s 91 |
| GPT-4o (May '24)OpenAI | 8.40 | - | 5.00 | 15.00 | Tok/s 91 |
| Qwen3 Coder 30B A3B InstructAlibaba | 13.60 | - | 0.45 | 2.25 | Tok/s 91 |
| GPT-5.4 (Non-reasoning)OpenAI | 28.30 | - | 2.50 | 15.00 | Tok/s 90 |
| MiniMax-M3MiniMax | 45.40 | $205 | 0.30 | 1.20 | Tok/s 90 |
| Claude 4.5 Haiku (Non-reasoning)Anthropic | 24.10 | - | 1.00 | 5.00 | Tok/s 90 |
| Llama Nemotron Super 49B v1.5 (Non-reasoning)NVIDIA | 8.50 | - | 0.40 | 0.40 | Tok/s 90 |
| Qwen3 VL 32B (Reasoning)Alibaba | 18.10 | - | 0.16 | 0.64 | Tok/s 90 |
| Claude 4.5 Haiku (Reasoning)Anthropic | 29.90 | $471 | 1.00 | 5.00 | Tok/s 89 |
| MiniMax-M2MiniMax | 28.90 | - | 0.30 | 1.20 | Tok/s 89 |
| MiniMax-M2.1MiniMax | 32.10 | - | 0.30 | 1.20 | Tok/s 88 |
| Qwen3.5 397B A17B (Reasoning)Alibaba | 34.30 | $527 | 0.60 | 3.60 | Tok/s 88 |
| Llama 3.3 Instruct 70BMeta | 9.30 | - | 0.66 | 0.72 | Tok/s 88 |
| Qwen3.5 9B (Non-reasoning)Alibaba | 20.60 | - | 0.17 | 0.25 | Tok/s 87 |
| GPT-5 mini (medium)OpenAI | 31.60 | - | 0.25 | 2.00 | Tok/s 85 |
| GPT-5.5 (xhigh)OpenAI | 56.30 | $2778 | 5.00 | 30.00 | Tok/s 85 |
| GLM-5.3 (max)Z AI | 59.50 | $1239 | 1.40 | 4.40 | Tok/s 85 |
| Grok 4.20 0309 v2 (Non-reasoning)SpaceXAI | 22.20 | - | 1.25 | 2.50 | Tok/s 85 |
| Ling-flash-2.0InclusionAI | 9.60 | - | 0.14 | 0.57 | Tok/s 84 |
| GLM-4.5V (Reasoning)Z AI | 9.00 | - | 0.60 | 1.80 | Tok/s 83 |
| GPT-5 mini (high)OpenAI | 25.80 | $130 | 0.25 | 2.00 | Tok/s 83 |
| Qwen3.5 397B A17B (Non-reasoning)Alibaba | 32.70 | - | 0.60 | 3.60 | Tok/s 83 |
| GLM-5.1 (Reasoning)Z AI | 41.00 | $700 | 1.39 | 4.40 | Tok/s 83 |
| Gemma 4 31B (Non-reasoning)Google | 22.30 | $33.8 | 0.15 | 0.40 | Tok/s 82 |
| GPT-5 (medium)OpenAI | 34.60 | - | 1.25 | 10.00 | Tok/s 82 |
| Qwen3.5 27B (Non-reasoning)Alibaba | 30.00 | - | 0.30 | 2.40 | Tok/s 82 |
| Qwen3.5 9B (Reasoning)Alibaba | 21.80 | $240 | 0.14 | 0.20 | Tok/s 81 |
| GPT-5 (high)OpenAI | 35.30 | $789 | 1.25 | 10.00 | Tok/s 81 |
| GPT-5.5 (medium)OpenAI | 51.40 | $939 | 5.00 | 30.00 | Tok/s 79 |
| GPT-5.2 (xhigh)OpenAI | 43.30 | - | 1.75 | 14.00 | Tok/s 78 |
| GPT-5 (minimal)OpenAI | 17.30 | - | 1.25 | 10.00 | Tok/s 78 |
| GLM-4.5-AirZ AI | 16.70 | - | 0.17 | 0.98 | Tok/s 77 |
| Qwen3.5 27B (Reasoning)Alibaba | 34.60 | - | 0.30 | 2.40 | Tok/s 77 |
| DeepSeek V4 Pro (Non-reasoning)DeepSeek | 31.90 | - | 0.43 | 0.87 | Tok/s 76 |
| Claude Sonnet 5 (Adaptive Reasoning, Max Effort)Anthropic | 55.30 | $4011 | 2.00 | 10.00 | Tok/s 76 |
| DeepSeek V4 Pro (Reasoning, High Effort)DeepSeek | 43.70 | $109 | 0.43 | 0.87 | Tok/s 75 |
| Gemma 4 26B A4B (Non-reasoning)Google | 20.40 | - | 0.13 | 0.40 | Tok/s 75 |
| DeepSeek V4 Pro 0813 (Reasoning, Max Effort)DeepSeek | 53.20 | $605 | 1.32 | 3.96 | Tok/s 74 |
| GPT-5.5 (high)OpenAI | 54.70 | $1721 | 5.00 | 30.00 | Tok/s 74 |
| Hermes 4 - Llama-3.1 70B (Reasoning)Nous Research | 9.90 | - | 0.13 | 0.40 | Tok/s 74 |
| Inkling (xhigh)Thinking Machines | 42.30 | $698 | 1.00 | 4.05 | Tok/s 74 |
| DeepSeek V4 Pro (Reasoning, Max Effort)DeepSeek | 45.30 | $176 | 0.43 | 0.87 | Tok/s 74 |
| GPT-5.5 (low)OpenAI | 44.50 | $374 | 5.00 | 30.00 | Tok/s 74 |
| GLM-4.6V (Non-reasoning)Z AI | 10.90 | - | 0.30 | 0.90 | Tok/s 73 |
| Grok 3 mini Reasoning (high)SpaceXAI | 22.90 | - | 0.30 | 0.50 | Tok/s 72 |
| Claude Sonnet 5 (Adaptive Reasoning, Xhigh Effort)Anthropic | - | - | 2.00 | 10.00 | Tok/s 71 |
| GLM-4.6V (Reasoning)Z AI | 16.90 | - | 0.30 | 0.90 | Tok/s 70 |
| GPT-5.5 (Non-reasoning)OpenAI | 35.80 | $212 | 5.00 | 30.00 | Tok/s 68 |
| GPT-5.2 (Non-reasoning)OpenAI | 26.50 | - | 1.75 | 14.00 | Tok/s 68 |
| GPT-5 (low)OpenAI | 31.90 | - | 1.25 | 10.00 | Tok/s 68 |
| Qwen3 VL 32B InstructAlibaba | 11.00 | - | 0.16 | 0.64 | Tok/s 67 |
| Hy3Tencent | 42.20 | $93.2 | 0.14 | 0.55 | Tok/s 67 |
| Claude Fable 5 (Adaptive Reasoning, Max Effort, Opus 4.8 Fallback)Anthropic | 62.10 | $5455 | 10.00 | 50.00 | Tok/s 67 |
| Inkling SmallThinking Machines | 41.20 | $193 | 0.30 | 1.20 | Tok/s 67 |
| Qwen3 Coder 480B A35B InstructAlibaba | 18.20 | - | 1.50 | 7.50 | Tok/s 66 |
| GPT-5.6 Sol (max)OpenAI | 60.90 | $2823 | 5.00 | 30.00 | Tok/s 65 |
| Qwen3 235B A22B 2507 (Reasoning)Alibaba | 19.90 | $170 | 0.23 | 2.30 | Tok/s 64 |
| GPT-5.6 Sol (high)OpenAIMelhor custo-benefício | 57.30 | $955 | 5.00 | 30.00 | Tok/s 63 |
| GPT-5.6 Sol (medium)OpenAI | 55.60 | $580 | 5.00 | 30.00 | Tok/s 63 |
| GPT-5.6 Sol (xhigh)OpenAI | 59.00 | $1525 | 5.00 | 30.00 | Tok/s 63 |
| Claude Sonnet 5 (Adaptive Reasoning, High Effort)Anthropic | - | - | 2.00 | 10.00 | Tok/s 62 |
| GLM-5.1 (Non-reasoning)Z AI | 36.30 | - | 1.39 | 4.40 | Tok/s 62 |
| Qwen3.6 Max PreviewAlibaba | 41.10 | - | 1.30 | 7.80 | Tok/s 62 |
| Qwen3 235B A22B (Reasoning)Alibaba | 13.50 | - | 0.70 | 8.40 | Tok/s 61 |
| MiniMax-M2.7MiniMax | 38.90 | $143 | 0.30 | 1.20 | Tok/s 61 |
| Command ACohere | 7.50 | - | 2.50 | 10.00 | Tok/s 60 |
| Qwen3 14B (Non-reasoning)Alibaba | 6.80 | - | 0.35 | 1.40 | Tok/s 60 |
| GPT-5.6 Sol (Non-reasoning)OpenAI | 41.90 | $240 | 5.00 | 30.00 | Tok/s 59 |
| Qwen3 235B A22B (Non-reasoning)Alibaba | 10.80 | - | 0.70 | 2.80 | Tok/s 59 |
| Qwen3 14B (Reasoning)Alibaba | 10.40 | - | 0.35 | 4.20 | Tok/s 59 |
| MiMo-V2.5-ProXiaomi | 42.90 | $98.9 | 0.43 | 0.87 | Tok/s 59 |
| Qwen3 Max (Preview)Alibaba | 19.40 | - | 1.20 | 6.00 | Tok/s 59 |
| Claude Sonnet 5 (Non-reasoning, High Effort)Anthropic | 42.60 | $416 | 2.00 | 10.00 | Tok/s 58 |
| Claude Sonnet 5 (Adaptive Reasoning, Medium Effort)Anthropic | - | - | 2.00 | 10.00 | Tok/s 58 |
| Claude Sonnet 5 (Adaptive Reasoning, Low Effort)Anthropic | - | - | 2.00 | 10.00 | Tok/s 58 |
| Qwen3 MaxAlibaba | 24.50 | - | 1.20 | 6.00 | Tok/s 58 |
| Qwen3 VL 235B A22B (Reasoning)Alibaba | 20.90 | - | 0.40 | 4.00 | Tok/s 57 |
| GPT-5.6 Sol (low)OpenAI | 50.70 | $344 | 5.00 | 30.00 | Tok/s 57 |
| Mistral Medium 3Mistral | 12.50 | - | 0.40 | 2.00 | Tok/s 57 |
| Qwen3.7 PlusAlibaba | 39.40 | $322 | 0.40 | 1.60 | Tok/s 57 |
| Grok 4.6 (high)SpaceXAIMelhor custo-benefício | 60.90 | $1068 | 2.00 | 6.00 | Tok/s 56 |
| Qwen3 235B A22B 2507 InstructAlibaba | 18.40 | - | 0.23 | 0.92 | Tok/s 56 |
| Qwen3.6 PlusAlibaba | 40.50 | $526 | 0.50 | 3.00 | Tok/s 56 |
| Qwen3 Max Thinking (Preview)Alibaba | 25.50 | - | 1.20 | 6.00 | Tok/s 56 |
| Hermes 4 - Llama-3.1 70B (Non-reasoning)Nous Research | 6.70 | - | 0.13 | 0.40 | Tok/s 56 |
| Reka Flash (Sep '24)Reka AI | 6.00 | - | 0.20 | 0.80 | Tok/s 56 |
| MiMo-V2.5-Pro (Non-reasoning)Xiaomi | 28.40 | - | 0.43 | 0.87 | Tok/s 56 |
| MiMo-V2.5Xiaomi | 38.00 | $25.2 | 0.14 | 0.28 | Tok/s 56 |
| Gemma 3n E4B InstructGoogle | 1.00 | - | 0.06 | 0.12 | Tok/s 55 |
| Qwen3.6 27B (Non-reasoning)Alibaba | 31.30 | $359 | 0.60 | 3.60 | Tok/s 55 |
| GLM-4.6 (Non-reasoning)Z AI | 23.40 | - | 0.57 | 2.20 | Tok/s 55 |
| Qwen3.6 27B (Reasoning)Alibaba | 37.70 | $667 | 0.60 | 3.60 | Tok/s 55 |
| Grok Build 0.1 0616SpaceXAI | 40.70 | $377 | 1.00 | 2.00 | Tok/s 55 |
| Llama 3.1 Instruct 70BMeta | 6.50 | - | 0.56 | 0.56 | Tok/s 55 |
| Claude Sonnet 4.6 (Adaptive Reasoning, Max Effort)Anthropic | 48.40 | $3355 | 3.00 | 15.00 | Tok/s 55 |
| Claude Opus 4.8 (Adaptive Reasoning, Max Effort)Anthropic | 57.30 | $3752 | 5.00 | 25.00 | Tok/s 54 |
| Gemma 4 E4B (Reasoning)Google | 12.20 | - | 0.02 | 0.10 | Tok/s 54 |
| Claude Opus 5 (Adaptive Reasoning, Xhigh Effort)AnthropicMelhor custo-benefício | 62.50 | $2909 | 5.00 | 25.00 | Tok/s 54 |
| Claude Opus 5 (Adaptive Reasoning, Max Effort)AnthropicMelhor custo-benefício | 63.10 | $3836 | 5.00 | 25.00 | Tok/s 53 |
| Gemma 4 E4B (Non-reasoning)Google | 8.70 | - | 0.02 | 0.10 | Tok/s 52 |
| Claude Opus 5 (Adaptive Reasoning, High Effort)AnthropicMelhor custo-benefício | 61.50 | $1974 | 5.00 | 25.00 | Tok/s 51 |
| Grok 4.5 (high)SpaceXAI | 55.80 | $579 | 2.00 | 6.00 | Tok/s 50 |
| GLM-4.6 (Reasoning)Z AI | 29.30 | $314 | 0.55 | 2.20 | Tok/s 50 |
| Claude Opus 5 (Adaptive Reasoning, Medium Effort)Anthropic | 58.60 | $1116 | 5.00 | 25.00 | Tok/s 50 |
| Qwen3.5 Omni PlusAlibaba | 31.30 | - | 0.40 | 4.80 | Tok/s 49 |
| Claude Opus 5 (Adaptive Reasoning, Low Effort)Anthropic | 52.50 | $555 | 5.00 | 25.00 | Tok/s 49 |
| Mistral Large 3Mistral | 15.90 | $71.1 | 0.50 | 1.50 | Tok/s 48 |
| Llama 3.1 Nemotron Ultra 253B v1 (Reasoning)NVIDIA | 8.90 | - | 0.60 | 1.80 | Tok/s 48 |
| Claude Opus 4.7 (Adaptive Reasoning, Max Effort)Anthropic | 55.00 | $3743 | 5.00 | 25.00 | Tok/s 48 |
| Llama 3.2 Instruct 11B (Vision)Meta | 3.00 | - | 0.34 | 0.34 | Tok/s 47 |
| Qwen3 VL 235B A22B InstructAlibaba | 14.40 | - | 0.40 | 1.60 | Tok/s 47 |
| Solar Pro 4Upstage | 41.60 | $321 | 0.30 | 1.20 | Tok/s 47 |
| Qwen3.8 2.4T A95BAlibaba | 57.70 | $1672 | 2.00 | 6.00 | Tok/s 46 |
| GLM-5 (Reasoning)Z AI | 40.60 | - | 1.00 | 3.20 | Tok/s 46 |
| Kimi K2.6Kimi | 45.10 | $841 | 0.95 | 4.00 | Tok/s 46 |
| Claude Opus 4.5 (Reasoning)Anthropic | 41.90 | - | 5.00 | 25.00 | Tok/s 46 |
| Claude Opus 4.5 (Non-reasoning)Anthropic | 35.60 | - | 5.00 | 25.00 | Tok/s 45 |
| Kimi K2.5 (Non-reasoning)Kimi | 30.10 | - | 0.60 | 3.00 | Tok/s 45 |
| Qwen3.8 MaxAlibaba | 58.10 | $1741 | 2.00 | 6.00 | Tok/s 45 |
| Phi-4 Mini InstructMicrosoft | 5.70 | - | 0.00 | 0.00 | Tok/s 44 |
| Claude Opus 4.7 (Non-reasoning, High Effort)Anthropic | 43.90 | - | 5.00 | 25.00 | Tok/s 44 |
| Claude Sonnet 4.6 (Non-reasoning, Low Effort)Anthropic | 35.10 | - | 3.00 | 15.00 | Tok/s 44 |
| Claude Sonnet 4.6 (Non-reasoning, High Effort)Anthropic | 36.80 | - | 3.00 | 15.00 | Tok/s 43 |
| Kimi K2.5 (Reasoning)Kimi | 36.00 | $304 | 0.60 | 3.00 | Tok/s 43 |
| Kimi K2Kimi | 19.70 | - | 0.57 | 2.30 | Tok/s 41 |
| Kimi K2 0905Kimi | 24.00 | - | 0.60 | 2.50 | Tok/s 41 |
| Claude 4.5 Sonnet (Reasoning)Anthropic | 37.40 | $1037 | 3.00 | 15.00 | Tok/s 40 |
| Kimi K2.6 (Non-reasoning)Kimi | 35.40 | - | 0.95 | 4.00 | Tok/s 40 |
| GLM-5 (Non-reasoning)Z AI | 33.20 | - | 1.00 | 3.20 | Tok/s 40 |
| LongCat 2.0LongCat | 34.00 | $442 | 0.75 | 2.95 | Tok/s 39 |
| Qwen3 8B (Non-reasoning)Alibaba | 4.80 | - | 0.18 | 0.70 | Tok/s 39 |
| Claude Opus 4.6 (Adaptive Reasoning, Max Effort)Anthropic | 44.90 | - | 5.00 | 25.00 | Tok/s 39 |
| Phi-4Microsoft | 4.60 | - | 0.13 | 0.50 | Tok/s 39 |
| Claude 4.5 Sonnet (Non-reasoning)Anthropic | 29.90 | - | 3.00 | 15.00 | Tok/s 39 |
| Qwen3 8B (Reasoning)Alibaba | 8.30 | - | 0.18 | 2.10 | Tok/s 39 |
| Kimi K3 (max)Kimi | 59.70 | $2425 | 3.00 | 15.00 | Tok/s 38 |
| Kimi K2.7 CodeKimi | 43.00 | $545 | 0.95 | 4.00 | Tok/s 37 |
| Claude Opus 4.6 (Non-reasoning, High Effort)Anthropic | 38.80 | - | 5.00 | 25.00 | Tok/s 36 |
| Kimi K3 (low)Kimi | 48.30 | $283 | 3.00 | 15.00 | Tok/s 36 |
| Hermes 3 - Llama-3.1 70BNous Research | 4.80 | - | 0.70 | 0.70 | Tok/s 34 |
| Nova PremierAmazon | 12.70 | - | 2.50 | 12.50 | Tok/s 33 |
| GPT-4 TurboOpenAI | 7.70 | - | 10.00 | 30.00 | Tok/s 32 |
| Hermes 4 - Llama-3.1 405B (Reasoning)Nous Research | 8.80 | - | 1.00 | 3.00 | Tok/s 30 |
| Qwen3.5 4B (Reasoning)Alibaba | 20.40 | - | 0.03 | 0.15 | Tok/s 30 |
| Seed-OSS-36B-InstructByteDance Seed | 18.50 | - | 0.21 | 0.57 | Tok/s 30 |
| Qwen3.5 4B (Non-reasoning)Alibaba | 16.10 | - | 0.03 | 0.15 | Tok/s 28 |
| DeepSeek R1 Distill Llama 70BDeepSeek | 9.80 | - | 0.70 | 1.10 | Tok/s 25 |
| Granite 4.0 H SmallIBM | 4.90 | - | 0.06 | 0.25 | Tok/s 25 |
| Hermes 4 - Llama-3.1 405B (Non-reasoning)Nous Research | 8.60 | - | 1.00 | 3.00 | Tok/s 22 |
| Phi-4 Multimodal InstructMicrosoft | 4.20 | - | 0.00 | 0.00 | Tok/s 17 |
| Granite 3.3 8B (Non-reasoning)IBM | 1.30 | - | 0.03 | 0.25 | Tok/s 14 |
Guias
Como comparamos os modelos
A inteligência é medida pelo Artificial Analysis Intelligence Index. O custo do benchmark é o custo total em USD para executar o teste, incluindo tokens de raciocínio; não é o preço da API por token. Todas as versões usam os mesmos dados e cálculos. A pontuação geral não mede o desempenho em um idioma ou tarefa específicos.