Ranking de Modelos de IA · Blog do Bufano

OpenRouter

Uso real — quais modelos os desenvolvedores mais utilizam, medido por tokens processados.

Ver fonte → 19 de jun. de 26, 20:04

1 Google: Nano Banana 2 (Gemini 3.1 Flash Image) Google visionfastest —

2 Google: Nano Banana Pro (Gemini 3 Pro Image) Google visionagentic —

3 Cohere: North Mini Code (free) Cohere codingfastest —

4 Z.ai: GLM 5.2 Z.ai —

5 OpenRouter: Fusion OpenRouter —

6 MoonshotAI: Kimi K2.7 Code MoonshotAI visioncoding —

7 Anthropic: Claude Fable Latest Anthropic vision —

8 Anthropic: Claude Fable 5 Anthropic vision —

9 Nex AGI: Nex-N2-Pro (free) Nex AGI vision —

10 NVIDIA: Nemotron 3.5 Content Safety (free) NVIDIA visionreasoning —

11 NVIDIA: Nemotron 3 Ultra (free) NVIDIA reasoning —

12 NVIDIA: Nemotron 3 Ultra NVIDIA reasoning —

13 Qwen: Qwen3.7 Plus Qwen vision —

14 MiniMax: MiniMax M3 MiniMax vision —

15 StepFun: Step 3.7 Flash StepFun visionfastest —

16 Anthropic: Claude Opus 4.8 (Fast) Anthropic visionfastest —

17 Anthropic: Claude Opus 4.8 Anthropic vision —

18 Qwen: Qwen3.7 Max Qwen —

19 xAI: Grok Build 0.1 xAI vision —

20 Google: Gemini 3.5 Flash Google visionfastest —

21 Anthropic: Claude Opus 4.7 (Fast) Anthropic visionfastest —

22 Perceptron: Perceptron Mk1 Perceptron vision —

23 inclusionAI: Ring-2.6-1T inclusionAI —

24 Google: Gemini 3.1 Flash Lite Google visionfastest —

25 OpenAI: GPT Chat Latest OpenAI vision —

26 xAI: Grok 4.3 xAI vision —

27 IBM: Granite 4.1 8B IBM —

28 Mistral: Mistral Medium 3.5 Mistral vision —

29 Owl Alpha —

30 NVIDIA: Nemotron 3 Nano Omni (free) NVIDIA visionreasoning —

31 Poolside: Laguna XS.2 (free) Poolside —

32 Poolside: Laguna XS.2 Poolside —

33 Poolside: Laguna M.1 (free) Poolside —

34 Poolside: Laguna M.1 Poolside —

35 Anthropic Claude Haiku Latest visionfastest —

36 OpenAI GPT Mini Latest visionfastest —

37 Google Gemini Pro Latest vision —

38 MoonshotAI Kimi Latest vision —

39 Google Gemini Flash Latest visionfastest —

40 Anthropic Claude Sonnet Latest vision —

41 OpenAI GPT Latest vision —

42 Qwen: Qwen3.5 Plus 2026-04-20 Qwen vision —

43 Qwen: Qwen3.6 Flash Qwen visionfastest —

44 Qwen: Qwen3.6 35B A3B Qwen vision —

45 Qwen: Qwen3.6 Max Preview Qwen —

46 Qwen: Qwen3.6 27B Qwen vision —

47 OpenAI: GPT-5.5 Pro OpenAI vision —

48 OpenAI: GPT-5.5 OpenAI vision —

49 DeepSeek: DeepSeek V4 Pro DeepSeek —

50 DeepSeek: DeepSeek V4 Flash DeepSeek fastest —

51 inclusionAI: Ling-2.6-1T inclusionAI —

52 Tencent: Hy3 preview Tencent —

53 Xiaomi: MiMo-V2.5-Pro Xiaomi —

54 Xiaomi: MiMo-V2.5 Xiaomi vision —

55 OpenAI: GPT-5.4 Image 2 OpenAI vision —

56 inclusionAI: Ling-2.6-flash inclusionAI fastest —

57 Anthropic: Claude Opus Latest Anthropic vision —

58 Pareto Code Router coding —

59 MoonshotAI: Kimi K2.6 MoonshotAI vision —

60 Anthropic: Claude Opus 4.7 Anthropic vision —

61 Anthropic: Claude Opus 4.6 (Fast) Anthropic visionfastest —

62 Z.ai: GLM 5.1 Z.ai —

63 Google: Gemma 4 26B A4B (free) Google vision —

64 Google: Gemma 4 26B A4B Google vision —

65 Google: Gemma 4 31B (free) Google vision —

66 Google: Gemma 4 31B Google vision —

67 Qwen: Qwen3.6 Plus Qwen vision —

68 Arcee AI: Trinity Large Thinking Arcee AI reasoning —

69 xAI: Grok 4.20 Multi-Agent xAI visionagentic —

70 xAI: Grok 4.20 xAI vision —

71 Google: Lyria 3 Pro Preview Google vision —

72 Google: Lyria 3 Clip Preview Google vision —

73 Kwaipilot: KAT-Coder-Pro V2 Kwaipilot coding —

74 Reka Edge vision —

75 MiniMax: MiniMax M2.7 MiniMax —

76 OpenAI: GPT-5.4 Nano OpenAI vision —

77 OpenAI: GPT-5.4 Mini OpenAI visionfastest —

78 Mistral: Mistral Small 4 Mistral vision —

79 Z.ai: GLM 5 Turbo Z.ai fastest —

80 NVIDIA: Nemotron 3 Super (free) NVIDIA reasoning —

81 NVIDIA: Nemotron 3 Super NVIDIA reasoning —

82 ByteDance Seed: Seed-2.0-Lite ByteDance Seed visionfastest —

83 Qwen: Qwen3.5-9B Qwen vision —

84 OpenAI: GPT-5.4 Pro OpenAI vision —

85 OpenAI: GPT-5.4 OpenAI vision —

86 Inception: Mercury 2 Inception —

87 OpenAI: GPT-5.3 Chat OpenAI vision —

88 Google: Gemini 3.1 Flash Lite Preview Google visionfastest —

89 ByteDance Seed: Seed-2.0-Mini ByteDance Seed visionfastest —

90 Google: Nano Banana 2 (Gemini 3.1 Flash Image Preview) Google visionfastest —

91 Qwen: Qwen3.5-35B-A3B Qwen vision —

92 Qwen: Qwen3.5-27B Qwen vision —

93 Qwen: Qwen3.5-122B-A10B Qwen vision —

94 Qwen: Qwen3.5-Flash Qwen visionfastest —

95 LiquidAI: LFM2-24B-A2B LiquidAI —

96 Google: Gemini 3.1 Pro Preview Custom Tools Google visionagentic —

97 OpenAI: GPT-5.3-Codex OpenAI vision —

98 AionLabs: Aion-2.0 AionLabs —

99 Google: Gemini 3.1 Pro Preview Google visionagentic —

100 Anthropic: Claude Sonnet 4.6 Anthropic vision —

Chatbot Arena

Elo crowd-sourced de milhões de comparações cegas head-to-head, votadas por humanos.

Ver fonte → 19 de jun. de 26, 20:04

1 Claude Fable 5 (High) Anthropic · Proprietary most battle-tested 14 29.86% ±5.65%

2 Claude Opus 4.8 (Thinking) Anthropic · Proprietary most battle-tested 9 15.23% ±4.20%

3 GPT 5.5 (xHigh) OpenAI · Proprietary most battle-tested 8 14.82% ±5.22%

4 Claude Opus 4.7 (Thinking) Anthropic · Proprietary most battle-tested 8 11.79% ±4.46%

5 GPT 5.5 (High) OpenAI · Proprietary most battle-tested 8 10.87% ±3.14%

6 Claude Opus 4.7 Anthropic · Proprietary most battle-tested 8 10.73% ±4.55%

7 Claude Opus 4.6 Anthropic · Proprietary most battle-tested 7 9.75% ±4.14%

8 GPT 5.5 OpenAI · Proprietary most battle-tested 7 8.43% ±3.07%

9 GPT 5.4 (High) OpenAI · Proprietary most battle-tested 7 5.28% ±3.11%

10 GLM 5.2 (Max) Z.ai · MIT · SiliconFlow most battle-tested 5 12.69% ±6.49%

11 Claude Opus 4.8 Anthropic · Proprietary most battle-tested 3 11.95% ±4.48%

12 Claude Sonnet 4.6 Anthropic · Proprietary most battle-tested 3 3.46% ±3.53%

13 GLM 5.1 Z.ai · MIT · SiliconFlow most battle-tested 2 0.53% ±3.31%

14 Gemini 3.5 Flash Google · Proprietary most battle-tested 0 1.71% ±2.74%

15 Gemini 3.1 Pro Preview Google · Proprietary most battle-tested 0 0.98% ±2.51%

16 DeepSeek V4 Pro DeepSeek · MIT · SiliconFlow most battle-tested 1 1.84% ±4.32%

17 Kimi K2.6 Moonshot · Modified MIT · Fireworks most battle-tested 1 2.97% ±2.91%

18 Kimi K2.7 Code Moonshot · Modified MIT · Fireworks most battle-tested 1 0.45% ±5.13%

19 DeepSeek V4 Flash DeepSeek · MIT · SiliconFlow most battle-tested 2 1.60% ±3.76%

20 Minimax M3 MiniMax · Proprietary · Fireworks most battle-tested 2 6.97% ±4.07%

21 Qwen 3.6 Plus Alibaba · Proprietary · Fireworks most battle-tested 4 6.90% ±3.21%

22 Grok Build 0.1 xAI · Proprietary most battle-tested 6 11.34% ±2.99%

23 Grok 4.3 (High) xAI · Proprietary most battle-tested 7 15.20% ±2.87%

24 Nemotron 3 Ultra Nvidia · OpenMDW-1.1 most battle-tested 7 2.30% ±13.04%

25 Minimax M2.7 MiniMax · Modified MIT · Fireworks most battle-tested 8 15.42% ±2.57%

26 Gemini 3 Flash Google · Proprietary most battle-tested 8 13.09% ±2.18%

27 Gemma 4 31B Google · Apache 2.0 most battle-tested 13 7.94% ±2.97%

28 Grok 4.3 xAI · Proprietary most battle-tested 18 14.98% ±2.25%

Artificial Analysis

Índice independente de inteligência, código e capacidade agêntica, com velocidade e preço.

Ver fonte → 19 de jun. de 26, 20:04

1 Claude Fable 5 (with fallback) 1

2 Claude Opus 4.8 (max) 1

3 GPT-5.5 (xhigh) 922

4 Claude Opus 4.7 (max) 1

5 GPT-5.5 (high) 922

6 GLM-5.2 (max) 1

7 Gemini 3.5 Flash 1

8 Claude Sonnet 4.6 (max) 1

9 GPT-5.5 (medium) 922

10 Gemini 3.1 Pro Preview 1

11 Qwen3.7 Max 1

12 Gemini 3.5 Flash (medium) 1

13 MiniMax-M3 1

14 DeepSeek V4 Pro (Max) 1

15 GPT-5.3 Codex (xhigh) 400

16 Muse Spark 262

17 Kimi K2.6 256

18 Claude Opus 4.7 (Non-reasoning, high) 1

19 MiMo-V2.5-Pro 1

20 Kimi K2.7 Code 256

21 GPT-5.5 (low) 922

22 DeepSeek V4 Pro (High) 1

23 DeepSeek V4 Flash (Max) 1

24 GLM-5.1 200

25 MiMo-V2.5 1

26 GPT-5.4 mini (xhigh) 400

27 Qwen3.6 Plus 1

28 Qwen3.7 Plus 1

29 GPT-5.4 nano (xhigh) 400

30 MiniMax-M2.7 205

31 GLM-5-Turbo 200

32 Nemotron 3 Ultra 262

33 Grok 4.3 (high) 1

34 DeepSeek V4 Flash (High) 1

35 Qwen3.6 27B 262

36 MiMo-V2-Omni-0327 256

37 Grok 4.3 (medium) 1

38 Claude Sonnet 4.6 (Non-reasoning) 1

39 Grok 4.3 (low) 1

40 GLM-5.1 200

41 MiMo-V2-Omni 256

42 Gemini 3.5 Flash (minimal) 1

43 Kimi K2.6 256

44 GLM 5V Turbo 200

45 Claude Sonnet 4.6 (Non-reasoning, Low Effort) 1

46 Qwen3.5 397B A17B 262

47 Hy3-preview 256

48 GPT-5.5 Instant (May 2026) 400

49 MiMo-V2-Flash (Feb 2026) 256

50 GPT-5.5 (Non-reasoning) 922

51 Qwen3.5 122B A10B 262

52 Qwen3.5 397B A17B 262

53 Qwen3.6 35B A3B 262

54 DeepSeek V4 Pro 1

55 Qwen3.5 Omni Plus 256

56 Ring-2.6-1T 262

57 o3 200

58 GPT-5.4 nano 400

59 Mistral Medium 3.5 256

60 GPT-5.4 mini (medium) 400

61 Step 3.7 Flash 256

62 Claude 4.5 Haiku 200

63 Gemma 4 31B 256

64 Command A+ 192

65 Qwen3.6 27B 262

66 DeepSeek V4 Flash 1

67 JT-35B-Flash 256

68 Qwen3.5 122B A10B 262

69 MiMo-V2.5-Pro 1

70 Gemini 2.5 Pro 1

71 Hy3-preview 256

72 Ling-2.6-1T 262

73 Step 3.5 Flash 2603 256

74 Doubao Seed Code 256

75 Gemma 4 26B A4B 256

76 NVIDIA Nemotron 3 Super 1

77 Mercury 2 128

78 Gemini 3.1 Flash-Lite 1

79 Qwen3.5 9B 262

80 Gemma 4 31B 256

81 Grok 4.3 (Non-reasoning) 1

82 K-EXAONE 256

83 Trinity Large Thinking 512

84 Qwen3.6 35B A3B 262

85 gpt-oss-120b (high) 131

86 Claude 4.5 Haiku 200

87 Qwen3.5 35B A3B 262

88 MiMo-V2-Flash 256

89 EXAONE 4.5 33B 262

90 HyperNova 60B 2605 131

91 Gemma 4 12B 256

92 ERNIE 5.0 Thinking Preview 128

93 Nova 2.0 Pro Preview (medium) 256

94 Nemotron Cascade 2 30B A3B 1

95 Qwen3 Coder Next 256

96 Nova 2.0 Omni (medium) 1

97 Mistral Small 4 256

98 North Mini Code 256

99 Nova 2.0 Lite (high) 1

100 Qwen3.5 9B 262

LLM Stats

Agregador de benchmarks compostos (GPQA, SWE-bench, HLE) com preço e throughput em tempo real.

Ver fonte → 19 de jun. de 26, 20:05

1 Claude Mythos PreviewUNRELEASEDAnthropic 65

2 GPT-5.2 ProOpenAI 63

3 Claude Opus 4.8Anthropic 61

4 GLM-5.2NEWZhipu AI 59

5 Claude Fable 5UNRELEASEDAnthropic 59

6 GPT-5.4OpenAI 57

7 GPT-5.5OpenAI 56

8 GPT-5.1OpenAI 56

9 Seed 2.0 ProByteDance 56

10 GPT-5.1 InstantOpenAI 55

11 Qwen3.7 MaxAlibaba Cloud / Qwen Team 54

12 Claude Opus 4.7Anthropic 54

13 GPT-5 MediumOpenAI 54

14 Gemini 3.5 FlashGoogle 53

15 Gemini 3.1 ProGoogle 53

HuggingFace · LMArena

O leaderboard do LMArena hospedado no HuggingFace, rankeado por volume de votos da comunidade.

Ver fonte →

Dados não disponíveis ainda.
Aguarde a próxima atualização.

Ranking de Modelos de IA.

OpenRouter

Chatbot Arena

Artificial Analysis

LLM Stats

HuggingFace · LMArena