AI Models Compared: from GPT-4 to GPT-6, Claude Opus 5.5 and Gemini 4

Well-known AI models, from GPT-4 to today's frontier, in the order they were released, each logo as tall as its score on Epoch AI's Capabilities Index above 100.

• GPT-4: ECI 125.73
• GPT-3.5 Turbo: ECI 113.27
• Claude 2: ECI 120.13
• Llama 2 70B: ECI 113.71
• Falcon 180B: ECI 112.02
• Mistral 7B: ECI 112.12
• Mixtral 8x7B: ECI 118.41
• Gemini 1.0 Pro: ECI 117.06
• Gemma 7B: ECI 111.90
• Claude 3 Opus: ECI 126.88
• Phi-3 Small: ECI 121.77
• GPT-4o: ECI 128.99
• Qwen2 72B: ECI 125.25
• Llama 3.1 405B: ECI 128.75
• Mistral Large 2: ECI 127.56
• Claude 3.5 Sonnet: ECI 133.40
• Phi-4: ECI 130.63
• Grok-2: ECI 130.59
• o1: ECI 141.82
• DeepSeek-V3: ECI 132.32
• DeepSeek-R1: ECI 138.83
• Claude 3.7 Sonnet: ECI 141.03
• Gemma 3 27B: ECI 129.89
• Llama 4 Maverick: ECI 132.34
• o3: ECI 146.86
• Qwen3 235B: ECI 139.14
• Gemini 2.5 Pro: ECI 145.28
• Grok 4: ECI 146.33
• GPT-5: ECI 150.00
• Gemini 3 Pro: ECI 153.22
• Claude Opus 4.5: ECI 150.20
• GLM-4.7: ECI 143.65
• Nemotron 3 Ultra: ECI 146.02
• Kimi K3: ECI 158.33
• Qwen 3.8 Max: ECI 157.98
• GLM-5.3: ECI 156.17
• Claude Fable 5.1: ECI 166.36
• GPT-6 Astra: ECI 169.06
• Grok 4.7: ECI 154.55
• Claude Opus 5.5: ECI 168.40
• Gemini 4 Argon: ECI 166.07
• Beam 1.0: ECI 150.24
• Mistral Large 4: ECI 150.59
• DeepSeek V4.1 Flash: ECI 158.83