Modelos de IA de pesos abertos: benchmarks e preços

Os pesos são os arquivos necessários para executar o modelo por conta própria com software e hardware adequados. Compare modelos com pesos disponíveis para download. Isso não implica licença de código aberto, uso comercial irrestrito ou hospedagem gratuita. Os preços API são do serviço hospedado; confira a licença antes de implantar.

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

ModeloPontuação de inteligênciaCusto do benchmark (USD)Preço de entrada / 1 milhão de tokens (USD)Preço de saída / 1 milhão de tokens (USD)Velocidade de saída
A.X-K2SK Telecom22.70----
Agnes 2.5 Pro AlphaSapiens AI26.80-0.450.90Tok/s 178
Apertus 70B InstructSwiss AI Initiative5.10-0.822.92-
Apertus 8B InstructSwiss AI Initiative4.80-0.100.20-
Apriel-v1.5-15B-ThinkerServiceNow13.80-0.000.00-
Apriel-v1.6-15B-ThinkerServiceNow13.40-0.000.00-
Arctic InstructSnowflake5.40----
Cogito v2.1 (Reasoning)Deep Cogito--1.251.25-
Command ACohere7.00-2.5010.00Tok/s 67
Command-R (Mar '24)Cohere5.00----
Command-R+ (Apr '24)Cohere5.30----
DBRX InstructDatabricks5.30----
DeepHermes 3 - Llama-3.1 8B Preview (Non-reasoning)Nous Research5.10----
DeepHermes 3 - Mistral 24B Preview (Non-reasoning)Nous Research6.10----
DeepSeek Coder V2 Lite InstructDeepSeek5.30----
DeepSeek LLM 67B Chat (V1)DeepSeek5.30----
DeepSeek R1 (Jan '25)DeepSeek11.40$4612.004.00-
DeepSeek R1 0528 (May '25)DeepSeek13.10-1.353.00-
DeepSeek R1 0528 Qwen3 8BDeepSeek8.10----
DeepSeek R1 Distill Llama 70BDeepSeek7.90-0.701.10Tok/s 21
DeepSeek R1 Distill Llama 8BDeepSeek6.50----
DeepSeek R1 Distill Qwen 1.5BDeepSeek5.50----
DeepSeek R1 Distill Qwen 14BDeepSeek7.80----
DeepSeek R1 Distill Qwen 32BDeepSeek8.40----
DeepSeek V3 (Dec '24)DeepSeek8.50$25.40.320.89-
DeepSeek V3 0324DeepSeek9.70$87.20.841.18-
DeepSeek V3.1 (Non-reasoning)DeepSeek13.70-0.571.68-
DeepSeek V3.1 (Reasoning)DeepSeek13.50-0.591.69-
DeepSeek V3.1 Terminus (Non-reasoning)DeepSeek13.90-0.271.00-
DeepSeek V3.1 Terminus (Reasoning)DeepSeek15.40-1.642.75-
DeepSeek V3.2 (Non-reasoning)DeepSeek16.00-0.280.42-
DeepSeek V3.2 (Reasoning)DeepSeek21.50-0.280.42-
DeepSeek V3.2 Exp (Non-reasoning)DeepSeek13.90-0.280.42-
DeepSeek V3.2 Exp (Reasoning)DeepSeek16.60-0.280.42-
DeepSeek V3.2 SpecialeDeepSeek14.50----
DeepSeek V4 Flash (Non-reasoning)DeepSeek18.90-0.090.20-
DeepSeek V4 Flash (Reasoning, High Effort)DeepSeek24.80$2100.130.28-
DeepSeek V4 Flash (Reasoning, Max Effort)DeepSeek24.60$1910.130.28-
DeepSeek V4 Flash 0731 (Reasoning, Max Effort)DeepSeek34.50$4740.441.32Tok/s 128
DeepSeek V4 Pro (Non-reasoning)DeepSeek20.80-0.430.87Tok/s 73
DeepSeek V4 Pro (Reasoning, High Effort)DeepSeek30.10-0.430.87Tok/s 70
DeepSeek V4 Pro (Reasoning, Max Effort)DeepSeek30.90$2770.430.87Tok/s 77
DeepSeek V4 Pro 0813 (Reasoning, Max Effort)DeepSeek36.30$11221.323.96Tok/s 80
DeepSeek-Coder-V2DeepSeek6.00----
DeepSeek-V2-ChatDeepSeek5.50----
DeepSeek-V2.5DeepSeek6.60----
DeepSeek-V2.5 (Dec '24)DeepSeek6.60----
Devstral Small (Jul '25)Mistral7.60----
Devstral Small (May '25)Mistral8.70----
DiffusionGemma 26B A4BGoogle9.50----
ERNIE 4.5 300B A47BBaidu7.50-0.281.10-
Exaone 4.0 1.2B (Non-reasoning)LG AI Research5.20----
Exaone 4.0 1.2B (Reasoning)LG AI Research5.30----
EXAONE 4.0 32B (Non-reasoning)LG AI Research6.30----
EXAONE 4.0 32B (Reasoning)LG AI Research8.20----
EXAONE 4.5 33BLG AI Research13.20----
EXAONE 4.5 33B (Non-reasoning)LG AI Research-----
Falcon-H1R-7BTII UAE7.80----
G9v3-39A5BAI9Stars21.80-0.000.00-
G9v3-3BAI9Stars10.80-0.000.00-
Gemma 3 1B InstructGoogle4.80-0.000.00-
Gemma 3 270MGoogle5.10----
Gemma 3 4B InstructGoogle4.80-0.000.00-
Gemma 3n E2B InstructGoogle4.80-0.000.00-
Gemma 3n E4B InstructGoogle4.80----
Gemma 3n E4B Instruct Preview (May '25)Google5.80----
Gemma 4 12B (Non-reasoning)Google9.40-0.100.30Tok/s 107
Gemma 4 12B (Reasoning)Google14.20-0.100.30Tok/s 108
Gemma 4 26B A4B (Non-reasoning)Google13.10-0.130.40Tok/s 82
Gemma 4 26B A4B (Reasoning)Google16.70-0.120.37-
Gemma 4 31B (Non-reasoning)Google13.90-0.140.40Tok/s 60
Gemma 4 E2B (Non-reasoning)Google6.50----
Gemma 4 E2B (Reasoning)Google7.80----
Gemma 4 E4B (Non-reasoning)Google7.50-0.020.10Tok/s 40
Gemma 4 E4B (Reasoning)Google8.90-0.020.10Tok/s 41
GLM-4.5 (Reasoning)Z AI12.80----
GLM-4.5-AirZ AI11.10-0.170.98Tok/s 64
GLM-4.5V (Non-reasoning)Z AI6.70-0.601.80Tok/s 51
GLM-4.5V (Reasoning)Z AI7.60-0.601.80Tok/s 46
GLM-4.6 (Non-reasoning)Z AI14.90-0.572.20Tok/s 69
GLM-4.6 (Reasoning)Z AI18.50-0.552.20Tok/s 53
GLM-4.6V (Non-reasoning)Z AI8.40-0.300.90Tok/s 44
GLM-4.6V (Reasoning)Z AI11.20-0.300.90Tok/s 63
GLM-4.7 (Non-reasoning)Z AI17.40-0.602.20Tok/s 106
GLM-4.7 (Reasoning)Z AI22.20-0.602.20Tok/s 98
GLM-4.7-Flash (Non-reasoning)Z AI10.60-0.070.40Tok/s 94
GLM-4.7-Flash (Reasoning)Z AI14.90-0.070.40Tok/s 105
GLM-5 (Non-reasoning)Z AI21.80-1.003.20Tok/s 59
GLM-5 (Reasoning)Z AI27.90-1.003.20Tok/s 71
GLM-5.1 (Non-reasoning)Z AI24.20-1.384.40Tok/s 41
GLM-5.1 (Reasoning)Z AI27.40-1.204.40Tok/s 57
GLM-5.2 (max)Z AI38.60-1.404.40Tok/s 64
GLM-5.2 (Non-reasoning)Z AI22.40-1.404.40Tok/s 117
GLM-5.3 (max)Z AI44.90$25031.404.40Tok/s 70
GLM-5.3-FlashZ AIMelhor custo-benefício41.90$2800.150.50Tok/s 66
gpt-oss-120b (high)OpenAI12.30$1220.150.59Tok/s 215
gpt-oss-120b (low)OpenAI10.20-0.150.54Tok/s 247
gpt-oss-20b (high)OpenAI9.00$29.40.060.19Tok/s 198
gpt-oss-20b (low)OpenAI10.00-0.070.20Tok/s 249
Granite 3.3 8B (Non-reasoning)IBM4.90-0.030.25Tok/s 16
Granite 4.0 1BIBM5.00----
Granite 4.0 350MIBM4.80----
Granite 4.0 H 1BIBM5.20----
Granite 4.0 H 350MIBM4.80----
Granite 4.0 H SmallIBM6.00-0.060.25Tok/s 22
Granite 4.0 MicroIBM5.10----
Granite 4.1 30BIBM7.40----
Granite 4.1 3BIBM5.90----
Granite 4.1 8BIBM6.60-0.050.10Tok/s 70
Granite 4.2 30BIBM14.80-0.160.65Tok/s 76
Granite 4.2 3BIBMMelhor custo-benefício9.10$14.90.030.12Tok/s 216
Granite 4.2 8BIBM12.40-0.060.25Tok/s 120
Grok 2 (Dec '24)SpaceXAI7.10----
Grok-1SpaceXAI6.30----
Hermes 3 - Llama-3.1 70BNous Research6.00-0.700.70Tok/s 29
Hermes 4 - Llama-3.1 405B (Non-reasoning)Nous Research7.40-1.003.00Tok/s 43
Hermes 4 - Llama-3.1 405B (Reasoning)Nous Research7.50-1.003.00Tok/s 42
Hermes 4 - Llama-3.1 70B (Non-reasoning)Nous Research6.70----
Hermes 4 - Llama-3.1 70B (Reasoning)Nous Research7.90----
Hy3Tencent25.80$1490.140.55Tok/s 94
Hy3-preview (Non-reasoning)Tencent17.00-0.060.21-
Hy3-preview (Reasoning)Tencent22.70-0.060.21-
HyperCLOVA X SEED Think (32B)Naver11.40----
HyperNova 60B 2605 (high, based on gpt-oss-120b)Multiverse Computing11.70-0.040.14Tok/s 353
Inkling (xhigh)Thinking Machines25.50$10261.004.05Tok/s 72
Inkling SmallThinking Machines26.10$2410.301.20Tok/s 126
INTELLECT-3Prime Intellect10.60----
Jamba 1.5 LargeAI21 Labs6.00-2.008.00-
Jamba 1.5 MiniAI21 Labs5.20-0.200.40-
Jamba 1.6 LargeAI21 Labs6.00----
Jamba 1.6 MiniAI21 Labs5.20----
Jamba 1.7 LargeAI21 Labs6.10----
Jamba 1.7 MiniAI21 Labs5.20----
Jamba Reasoning 3BAI21 Labs5.70----
K-EXAONE (Non-reasoning)LG AI Research11.20----
K-EXAONE (Reasoning)LG AI Research14.40----
K-EXAONE 2.0LG AI Research19.70----
K2 Horizon 375B A23BMBZUAI Institute de Foundation Models33.90----
K2 Think V2MBZUAI Institute de Foundation Models11.50----
K2-V2 (high)MBZUAI Institute de Foundation Models9.90----
K2-V2 (low)MBZUAI Institute de Foundation Models7.30----
K2-V2 (medium)MBZUAI Institute de Foundation Models9.00----
Kimi K2Kimi12.70-0.572.30Tok/s 40
Kimi K2 0905Kimi15.30-0.602.50Tok/s 40
Kimi K2 ThinkingKimi22.00-0.602.50Tok/s 121
Kimi K2.5 (Non-reasoning)Kimi19.40-0.603.00-
Kimi K2.5 (Reasoning)Kimi23.50-0.602.75-
Kimi K2.6Kimi31.30-0.954.00Tok/s 56
Kimi K2.6 (Non-reasoning)Kimi23.60-0.954.00Tok/s 51
Kimi K2.7 CodeKimi26.30$9930.954.00Tok/s 71
Kimi K3 (low)Kimi34.50-3.0015.00Tok/s 38
Kimi K3 (max)Kimi43.80$36583.0015.00Tok/s 42
Kimi Linear 48B A3B InstructKimi7.30----
LFM2 1.2BLiquid AI4.80----
LFM2 2.6BLiquid AI5.20----
LFM2 24B A2BLiquid AI5.90----
LFM2 8B A1BLiquid AI4.90----
LFM2.5-1.2B-InstructLiquid AI5.20----
LFM2.5-1.2B-ThinkingLiquid AI5.20----
LFM2.5-2.6BLiquid AI8.40-0.000.00Tok/s 200
LFM2.5-8B-A1BLiquid AI7.20-0.000.00-
LFM2.5-VL-1.6BLiquid AI4.80-0.000.00-
Ling 2.6 FlashInclusionAI9.70----
Ling 3.0 FlashInclusionAI24.90-0.070.22Tok/s 310
Ling-1TInclusionAI9.20----
Ling-2.6-1TInclusionAI17.00-0.302.50-
Ling-flash-2.0InclusionAI7.80-0.140.57Tok/s 7
Ling-mini-2.0InclusionAI5.50----
Llama 2 Chat 13BMeta5.30----
Llama 2 Chat 70BMeta5.30----
Llama 2 Chat 7BMeta5.70-0.050.25-
Llama 3 Instruct 70BMeta5.50-0.652.75-
Llama 3 Instruct 8BMeta4.80-0.040.14-
Llama 3.1 Instruct 405BMeta7.30----
Llama 3.1 Instruct 70BMeta6.60-0.560.56Tok/s 60
Llama 3.1 Instruct 8BMeta6.90-0.020.05Tok/s 141
Llama 3.1 Nemotron Instruct 70BNVIDIA6.90-1.201.20Tok/s 31
Llama 3.1 Nemotron Nano 4B v1.1 (Reasoning)NVIDIA7.30----
Llama 3.1 Nemotron Ultra 253B v1 (Reasoning)NVIDIA7.50----
Llama 3.1 Tulu3 405BAllen Institute for AI7.20----
Llama 3.2 Instruct 11B (Vision)Meta5.40-0.340.34Tok/s 12
Llama 3.2 Instruct 1BMeta4.80----
Llama 3.2 Instruct 3BMeta5.70----
Llama 3.2 Instruct 90B (Vision)Meta6.40----
Llama 3.3 Instruct 70BMeta7.70-0.660.72Tok/s 88
Llama 3.3 Nemotron Super 49B v1 (Non-reasoning)NVIDIA7.30----
Llama 3.3 Nemotron Super 49B v1 (Reasoning)NVIDIA8.90----
Llama 4 MaverickMeta9.30$45.50.260.91Tok/s 93
Llama 4 ScoutMeta6.50$77.60.190.68Tok/s 93
Llama 65BMeta5.00----
Llama Nemotron Super 49B v1.5 (Non-reasoning)NVIDIA7.40-0.400.40Tok/s 35
Llama Nemotron Super 49B v1.5 (Reasoning)NVIDIA9.00-0.400.40Tok/s 41
LongCat 2.0LongCat19.70$5450.752.95Tok/s 50
LongCat Flash LiteLongCat11.50----
Magistral Small 1Mistral8.20----
Magistral Small 1.2Mistral8.60-0.501.50-
MiMo-V2-Flash (Feb 2026)Xiaomi22.40----
MiMo-V2-Flash (Non-reasoning)Xiaomi16.00----
MiMo-V2-Flash (Reasoning)Xiaomi20.80-0.100.30-
MiMo-V2.5XiaomiMelhor custo-benefício22.30$42.60.140.28Tok/s 65
MiMo-V2.5-ProXiaomiMelhor custo-benefício26.40$1380.430.87Tok/s 36
MiMo-V2.5-Pro (Non-reasoning)Xiaomi18.30-0.430.87Tok/s 36
MiniCPM-V 4.6 1.3BOpenBMB5.70----
MiniCPM5-1B (Non-reasoning)OpenBMB8.70----
MiniCPM5-1B (Reasoning)OpenBMB8.80----
MiniCPM5-2BOpenBMB14.30----
MiniMax M1 40kMiniMax10.00----
MiniMax M1 80kMiniMax11.70-0.552.20-
MiniMax-M2MiniMax18.60-0.301.20Tok/s 94
MiniMax-M2.1MiniMax20.90-0.301.20Tok/s 91
MiniMax-M2.5MiniMax22.80-0.301.20Tok/s 96
MiniMax-M2.7MiniMax23.20$2360.301.20Tok/s 73
MiniMax-M3MiniMax29.60$5380.301.20Tok/s 156
Ministral 3 14BMistral6.00$2500.200.20Tok/s 92
Ministral 3 3BMistral4.80$61.30.100.10Tok/s 186
Ministral 3 8BMistral5.50$1690.150.15Tok/s 89
Mistral 7B InstructMistral5.00-0.250.25Tok/s 76
Mistral Large 2 (Jul '24)Mistral6.80-2.006.00-
Mistral Large 2 (Nov '24)Mistral7.60----
Mistral Large 3Mistral9.70$1300.501.50Tok/s 79
Mistral Medium 3.5Mistral14.90$11601.507.50Tok/s 148
Mistral Small (Sep '24)Mistral5.80-0.200.60Tok/s 160
Mistral Small 3Mistral6.70-0.100.30Tok/s 158
Mistral Small 3.1Mistral7.40$35.80.100.30Tok/s 150
Mistral Small 3.2Mistral7.00$1630.100.30Tok/s 157
Mistral Small 4 (Non-reasoning)Mistral9.00-0.150.60Tok/s 158
Mistral Small 4 (Reasoning)Mistral11.50$78.70.150.60Tok/s 168
Mixtral 8x22B InstructMistral5.70----
Mixtral 8x7B InstructMistral5.10-0.450.70-
Molmo 7B-DAllen Institute for AI5.60----
Molmo2-8BAllen Institute for AI5.00----
Motif 3Motif Technologies33.60----
Muse Glimmer (high)Meta18.10$1520.351.50Tok/s 104
Nanbeige4.1-3BNanbeige8.40----
Nemotron 3 Nano Omni 30B A3B ReasoningNVIDIA10.30-0.090.36-
Nemotron 3 Super 120B A12B (Reasoning)NVIDIA13.60$8590.190.65Tok/s 99
Nemotron 3 Ultra 550B A55B (Reasoning)NVIDIA23.40$8010.602.60Tok/s 168
Nemotron 3.5 LightningNVIDIA13.60$1050.060.20Tok/s 289
Nemotron Cascade 2 30B A3BNVIDIA11.70----
Nex-N2-ProNex AGI28.20-0.502.50Tok/s 115
North Mini CodeCohere12.80-0.000.00Tok/s 45
NVIDIA Nemotron 3 Nano 30B A3B (Non-reasoning)NVIDIA6.80-0.050.20Tok/s 227
NVIDIA Nemotron 3 Nano 30B A3B (Reasoning)NVIDIA8.90$63.70.050.20Tok/s 205
NVIDIA Nemotron 3 Nano 4BNVIDIA7.40----
NVIDIA Nemotron Nano 12B v2 VL (Non-reasoning)NVIDIA5.80-0.200.60Tok/s 70
NVIDIA Nemotron Nano 12B v2 VL (Reasoning)NVIDIA7.50-0.200.60Tok/s 32
NVIDIA Nemotron Nano 9B V2 (Non-reasoning)NVIDIA6.80-0.050.20Tok/s 146
NVIDIA Nemotron Nano 9B V2 (Reasoning)NVIDIA7.40-0.040.16Tok/s 110
OLMo 2 32BAllen Institute for AI6.00----
OLMo 2 7BAllen Institute for AI5.60----
Olmo 3 32B ThinkAllen Institute for AI6.50----
Olmo 3 7B InstructAllen Institute for AI5.20-0.100.20-
Olmo 3 7B ThinkAllen Institute for AI5.60----
Olmo 3.1 32B InstructAllen Institute for AI6.50----
Olmo 3.1 32B ThinkAllen Institute for AI7.10-0.000.00-
OpenChat 3.5 (1210)OpenChat5.30----
Phi-3 Mini Instruct 3.8BMicrosoft5.80----
Phi-4Microsoft5.90-0.130.50Tok/s 41
Phi-4 Mini InstructMicrosoft6.30-0.000.00Tok/s 45
Phi-4 Multimodal InstructMicrosoft5.80-0.000.00Tok/s 17
Pixtral LargeMistral7.10----
Qwen Chat 14BAlibaba5.00----
Qwen Chat 72BAlibaba5.40----
Qwen1.5 Chat 110BAlibaba5.70----
Qwen2 Instruct 72BAlibaba6.30----
Qwen2.5 Coder Instruct 32BAlibaba6.70----
Qwen2.5 Coder Instruct 7B Alibaba5.80----
Qwen2.5 Instruct 32BAlibaba6.90----
Qwen2.5 Instruct 72BAlibaba7.70-0.470.49-
Qwen3 0.6B (Non-reasoning)Alibaba4.80----
Qwen3 0.6B (Reasoning)Alibaba4.80----
Qwen3 1.7B (Non-reasoning)Alibaba4.90----
Qwen3 1.7B (Reasoning)Alibaba5.20----
Qwen3 14B (Non-reasoning)Alibaba6.70-0.351.40Tok/s 60
Qwen3 14B (Reasoning)Alibaba6.40-0.354.20Tok/s 58
Qwen3 235B A22B (Non-reasoning)Alibaba8.30-0.702.80Tok/s 61
Qwen3 235B A22B (Reasoning)Alibaba9.50-0.708.40Tok/s 58
Qwen3 235B A22B 2507 (Reasoning)Alibaba12.70$2350.232.30Tok/s 57
Qwen3 235B A22B 2507 InstructAlibaba12.00-0.230.92Tok/s 59
Qwen3 30B A3B (Non-reasoning)Alibaba6.60-0.200.80Tok/s 107
Qwen3 30B A3B (Reasoning)Alibaba7.60-0.202.40Tok/s 104
Qwen3 30B A3B 2507 (Reasoning)Alibaba9.80$1880.202.40Tok/s 141
Qwen3 30B A3B 2507 InstructAlibaba7.50-0.200.80Tok/s 148
Qwen3 32B (Non-reasoning)Alibaba7.30-0.160.64Tok/s 105
Qwen3 32B (Reasoning)Alibaba7.20-0.160.64Tok/s 104
Qwen3 4B (Non-reasoning)Alibaba6.60----
Qwen3 4B (Reasoning)Alibaba7.20----
Qwen3 4B 2507 (Reasoning)Alibaba8.80----
Qwen3 4B 2507 InstructAlibaba6.70----
Qwen3 8B (Non-reasoning)Alibaba6.00-0.180.70Tok/s 39
Qwen3 8B (Reasoning)Alibaba5.20-0.182.10Tok/s 39
Qwen3 Coder 30B A3B InstructAlibaba9.60-0.452.25Tok/s 90
Qwen3 Coder 480B A35B InstructAlibaba11.90-1.507.50Tok/s 57
Qwen3 Coder NextAlibaba10.10$4760.351.20Tok/s 117
Qwen3 Next 80B A3B (Reasoning)Alibaba11.20-0.151.20Tok/s 193
Qwen3 Next 80B A3B InstructAlibaba9.60-0.151.20Tok/s 178
Qwen3 Omni 30B A3B (Reasoning)Alibaba7.80-0.250.97Tok/s 100
Qwen3 Omni 30B A3B InstructAlibaba6.00-0.250.97Tok/s 95
Qwen3 VL 235B A22B (Reasoning)Alibaba13.40-0.404.00Tok/s 56
Qwen3 VL 235B A22B InstructAlibaba9.90-0.401.60Tok/s 51
Qwen3 VL 30B A3B (Reasoning)Alibaba9.50-0.202.40Tok/s 110
Qwen3 VL 30B A3B InstructAlibaba7.90-0.200.80Tok/s 110
Qwen3 VL 32B (Reasoning)Alibaba11.90-0.160.64Tok/s 90
Qwen3 VL 32B InstructAlibaba8.40-0.160.64Tok/s 68
Qwen3 VL 4B (Reasoning)Alibaba7.00----
Qwen3 VL 4B InstructAlibaba5.70----
Qwen3 VL 8B (Reasoning)Alibaba8.20-0.182.10Tok/s 113
Qwen3 VL 8B InstructAlibaba7.30-0.180.70Tok/s 119
Qwen3.5 0.8B (Non-reasoning)Alibaba5.40----
Qwen3.5 0.8B (Reasoning)Alibaba6.10----
Qwen3.5 122B A10B (Non-reasoning)Alibaba17.70-0.403.20Tok/s 144
Qwen3.5 122B A10B (Reasoning)Alibaba16.20$6610.403.20Tok/s 129
Qwen3.5 27B (Non-reasoning)Alibaba19.40-0.302.40Tok/s 84
Qwen3.5 27B (Reasoning)Alibaba22.90-0.302.40Tok/s 77
Qwen3.5 2B (Non-reasoning)Alibaba6.20----
Qwen3.5 2B (Reasoning)Alibaba6.90----
Qwen3.5 35B A3B (Non-reasoning)Alibaba15.10-0.252.00Tok/s 164
Qwen3.5 35B A3B (Reasoning)Alibaba19.30-0.252.00Tok/s 146
Qwen3.5 397B A17B (Non-reasoning)Alibaba21.40-0.603.60Tok/s 83
Qwen3.5 397B A17B (Reasoning)Alibaba19.10$9410.603.60Tok/s 86
Qwen3.5 4B (Non-reasoning)Alibaba10.80-0.030.15Tok/s 22
Qwen3.5 4B (Reasoning)Alibaba13.10-0.030.15Tok/s 27
Qwen3.5 9B (Non-reasoning)Alibaba13.30-0.170.25Tok/s 85
Qwen3.5 9B (Reasoning)Alibaba13.70-0.140.20Tok/s 83
Qwen3.6 27B (Non-reasoning)Alibaba19.80-0.603.60Tok/s 58
Qwen3.6 27B (Reasoning)Alibaba21.90$11450.603.60Tok/s 57
Qwen3.6 35B A3B (Non-reasoning)Alibaba15.20-0.382.25Tok/s 144
Qwen3.6 35B A3B (Reasoning)Alibaba22.30-0.382.25Tok/s 120
Qwen3.8 2.4T A95BAlibaba40.00$26322.006.00Tok/s 40
Qwen3.8 27B (low)Alibaba29.20-0.503.00Tok/s 54
Qwen3.8 27B (medium)Alibaba30.70-0.503.00Tok/s 54
Qwen3.8 27B (Non-reasoning)Alibaba22.40-0.503.00Tok/s 56
Qwen3.8 27B (xhigh)Alibaba33.90$11700.503.00Tok/s 47
Qwen3.8-Flash-NextAlibaba42.20-0.150.47Tok/s 72
QwQ 32BAlibaba9.50-0.661.00-
QwQ 32B-PreviewAlibaba7.60----
R1 1776Perplexity6.40----
Reka Flash 3Reka AI5.60-0.200.80-
Ring-1TInclusionAI10.90----
Ring-2.6-1TInclusionAI17.30$5760.302.50Tok/s 121
Ring-flash-2.0InclusionAI7.20-0.140.57-
Sarvam 105B (high)Sarvam8.80-0.040.17-
Sarvam 30B (high)Sarvam6.60-0.030.11-
Sarvam M (Reasoning)Sarvam5.30-0.000.00-
Seed-OSS-36B-InstructByteDance Seed12.10-0.210.57Tok/s 30
Solar MiniUpstage6.40-0.150.15-
Solar Open 100B (Reasoning)Upstage10.40----
Solar Open2 250BUpstage24.70----
Step 3.5 FlashStepFun16.60-0.100.30Tok/s 226
Step 3.7 FlashStepFun19.50-0.201.15Tok/s 81
Step3 VL 10BStepFun7.70----
Tiny Aya GlobalCohere4.80-0.000.00-
Tri-21B-ThinkTrillion Labs9.00----
Tri-21B-think PreviewTrillion Labs9.60----
Trinity Large ThinkingArcee AI10.90$5180.250.90Tok/s 320

Guias

Modelos de IA com melhor custo-benefícioAPIs de IA mais baratas: preços por tokenModelos de IA mais rápidos: velocidade de saídaModelos de IA de pesos abertos: benchmarks e preçosComo funciona o preço por token das APIs de IA

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.