Desempenho versus preço
Canto superior esquerdo = barato e forte. Clique em um ponto para ver detalhes.
| Posição | Modelo | Pontuação geral | Artificial Analysis | LiveBench | Epoch AI | ARC-AGI | Scale SEAL | Vals AI | Terminal-Bench | EQ-Bench | SimpleBench | BenchLM |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 1 | Google | 93.4 | — | — | — | — | — | — | #8 | — | #2 | #2 |
| 2 | Anthropic | 92.2 | — | — | #1 | #5 | — | #1 | #2 | — | #1 | #1 |
| 3 | Google | 91.2 | — | — | #3 | #3 | — | #4 | #4 | — | #3 | #4 |
| 4 | Anthropic | 90.1 | #1 | #2 | #5 | #6 | — | #5 | #2 | #2 | #6 | #6 |
| 5 | OpenAI | 90.1 | — | — | #2 | #1 | — | #8 | #6 | — | #17 | #3 |
| 6 | Google | 88.4 | — | — | #9 | #7 | — | #7 | #13 | — | #4 | #11 |
| 7 | OpenAI | 88.2 | #5 | #4 | #4 | #2 | #1 | #11 | #10 | — | #9 | #5 |
| 8 | Google | 87.9 | #6 | — | #11 | #9 | — | #6 | #14 | #12 | #4 | #10 |
| 9 | Meta | 87.4 | #3 | — | — | — | — | #9 | — | — | — | #14 |
| 10 | Anthropic | 87.1 | #4 | #1 | #8 | #11 | #2 | — | #12 | — | #7 | #12 |
| 11 | OpenAI | 86.9 | #8 | #5 | #6 | #3 | — | #13 | #9 | #8 | #10 | #8 |
| 12 | OpenAI | 86.9 | — | — | — | #15 | — | #10 | #7 | — | #30 | #7 |
| 13 | Anthropic | 86.7 | #2 | #21 | #10 | #20 | — | #2 | #5 | — | #21 | #13 |
| 14 | Anthropic | 85.2 | #10 | #3 | #13 | #13 | — | — | #19 | #5 | — | #16 |
| 15 | Anthropic | 85.2 | #9 | #11 | #12 | #10 | — | — | #15 | #1 | #14 | #15 |
| 16 | Meta | 84.3 | #7 | — | #17 | — | — | #12 | #18 | #3 | — | #20 |
| 17 | OpenAI | 84.0 | — | — | #7 | — | — | #17 | #15 | — | #26 | #9 |
| 18 | Alibaba (Qwen) | 82.9 | — | — | — | — | — | #14 | — | — | — | #16 |
| 19 | Moonshot AI | 82.7 | — | — | #18 | — | — | #15 | #11 | — | — | #22 |
| 20 | OpenAI | 82.1 | #12 | #13 | #14 | #12 | — | — | #17 | — | #24 | #18 |
| 21 | OpenAI | 81.3 | #13 | #8 | #15 | #7 | — | — | #23 | #13 | #35 | #19 |
| 22 | OpenAI | 80.5 | — | — | #21 | #21 | — | — | — | — | — | — |
| 23 | Meta | 80.3 | #11 | #5 | #28 | — | — | — | #40 | — | #13 | — |
| 24 | Tencent | 79.2 | — | — | — | — | — | #20 | #25 | — | — | — |
| 25 | OpenAI | 79.1 | — | — | #16 | #17 | — | — | — | — | #18 | #21 |
| 26 | DeepSeek | 78.9 | #29 | #8 | #24 | #24 | — | #16 | #23 | — | #33 | #22 |
| 27 | Meta | 78.6 | — | — | — | — | — | #17 | #27 | — | #25 | — |
| 28 | OpenAI | 77.6 | #33 | #10 | #22 | #16 | — | — | — | #6 | #31 | #24 |
| 29 | Thinking Machines | 77.1 | — | — | — | #23 | — | #26 | — | — | — | — |
| 30 | Xiaomi | 76.8 | #15 | — | — | — | — | #24 | #29 | — | — | — |
| 31 | Z.ai (Zhipu) | 76.7 | #17 | #31 | #34 | — | — | #19 | #20 | — | #34 | — |
| 31 | xAI | 76.7 | #14 | #22 | #38 | #34 | — | #23 | #22 | — | — | — |
| 33 | OpenAI | 76.5 | — | #17 | #29 | #27 | — | — | — | #10 | — | — |
| 34 | Alibaba (Qwen) | 76.2 | #16 | #16 | #41 | — | — | — | #29 | — | — | — |
| 35 | Google | 75.7 | #26 | #33 | #25 | #13 | #5 | #27 | #36 | — | #11 | — |
| 36 | xAI | 74.5 | #18 | #17 | #30 | #31 | — | — | #35 | — | #21 | — |
| 37 | Google | 74.2 | #30 | #15 | #27 | #17 | — | — | #42 | — | — | — |
| 38 | Meta | 74.1 | #28 | #17 | #46 | — | — | — | — | — | #23 | — |
| 39 | StepFun | 74.0 | #19 | — | — | — | — | #28 | — | — | — | — |
| 40 | Moonshot AI | 73.3 | #20 | #13 | #26 | #38 | — | #31 | #32 | #4 | #45 | — |
| 41 | OpenAI | 72.5 | #22 | #20 | #19 | #22 | — | — | #34 | #15 | #66 | — |
| 42 | Anthropic | 72.5 | — | #28 | #32 | #26 | #10 | — | — | #7 | #38 | — |
| 43 | Anthropic | 72.4 | #23 | #29 | #23 | #28 | — | — | #31 | #9 | #35 | — |
| 44 | Meta | 72.0 | — | — | #54 | — | #7 | — | — | — | — | — |
| 45 | Anthropic | 71.6 | #21 | #48 | #20 | #36 | — | #22 | #21 | — | #50 | — |
| 46 | Alibaba (Qwen) | 70.2 | #27 | #29 | — | — | — | — | — | — | — | — |
| 47 | Google | 69.8 | #51 | #25 | #40 | #25 | #3 | — | #49 | #25 | #15 | — |
| 48 | Xiaomi | 68.9 | #37 | — | — | — | — | #28 | #38 | — | — | — |
| 49 | xAI | 68.7 | — | — | #60 | #33 | — | — | — | — | — | — |
| 50 | Anthropic | 67.7 | — | #39 | #39 | #30 | #11 | — | — | #17 | #32 | — |
Um ranking geral de modelos e empresas de IA que combina estatisticamente 11 leaderboards de IA de referência.
A cada hora. Novos modelos de qualquer leaderboard são adicionados automaticamente.
A métrica principal de cada site é padronizada e um modelo fatorial corrige as diferenças entre sites para estimar a capacidade comum de cada modelo.
Não. Avaliações ausentes nunca penalizam. Em vez disso, a incerteza é mostrada como um intervalo de 90%.
O nº 1 do ranking geral. A melhor escolha depende do seu uso, então confira também os rankings por área.