Multitask Language Understanding on MMLU-Pro
89.31AccuracyGemini-3-Pro-Preview
Evaluation Results
| Method | Links | |
|---|---|---|
| Gemini-3-Pro-Preview2026.05 | 89.31 | |
| GPT-5 (High)2026.02 | 87.1 | |
| Gemini 3-Pro2026.02 | 86.88 | |
| GPT-5-2025-08-07Setting=high2026.05 | 86.21 | |
| Gemini 2.5 Pro2026.02 | 86.2 | |
| Gemini 2.5-Pro2026.02 | 86.2 | |
| Kimi-K2.52026.05 | 86.2 | |
| Claude Opus 4Reasoning=Thinking2026.02 | 86 | |
| Gemini-2.5-Pro2026.05 | 85.85 | |
| DeepSeek-V3.22026.05 | 85.81 | |
| JTMode=Think2026.05 | 85.57 | |
| DeepSeek-V3.2-Speciale2026.05 | 85.45 | |
| GLM-5Precision=FP82026.05 | 85.23 | |
| DS V3.2-Thinking2026.02 | 85 | |
| Kimi-K2Mode=Thinking2026.05 | 84.34 | |
| TeleChatMode=Thinking2026.05 | 84.09 | |
| GLM-4-72026.05 | 83.97 | |
| ERNIE 5.02026.02 | 83.8 | |
| Claude Sonnet 42026.02 | 83.7 | |
| Claude Sonnet 4Reasoning=Thinking2026.02 | 83.7 | |
| Interfaze-Beta2026.02 | 83.6 | |
| DeepSeek-R1-05282026.05 | 83.52 | |
| Step-3.5-Flash2026.05 | 83.46 | |
| Qwen3Parameters=235B, Active Parameters=A22B, Mode=Thinking, Version=25072026.05 | 83.46 | |
| DeepSeek-Chat-V3-03242026.05 | 83.27 | |
| Mimo-V2-Flash2026.05 | 83.11 | |
| doubao-seed-1.6Mode=thinking, Version=2506152026.05 | 83.11 | |
| o3-2025-04-162026.05 | 83.02 | |
| Claude Sonnet 4Mode=Thinking2026.05 | 83.02 | |
| GLM-4.62026.05 | 82.95 | |
| GLM-4.52026.05 | 82.72 | |
| Finix-P1-32BMode=Thinking2026.05 | 82.65 | |
| o4-mini-2025-04-16Setting=high2026.05 | 82.42 | |
| Qwen3-NextParameters=80B, Active Parameters=A3B, Mode=Thinking2026.05 | 82.02 | |
| MiniMax-M2.52026.05 | 81.74 | |
| MiniMax-M22026.05 | 81.6 | |
| Qwen3-NextParameters=80B, Active Parameters=A3B, Type=Instruct2026.05 | 81.31 | |
| GPT-4.1-202504142026.05 | 81.02 | |
| LongCat-Flash-Chat2026.05 | 80.95 | |
| Gemini 2.5 Flash2026.02 | 80.9 | |
| Qwen3Parameters=235B, Active Parameters=A22B, Mode=Thinking2026.05 | 80.89 | |
| GLM-4.5-Air2026.05 | 80.68 | |
| Seed-OSS-36BType=Instruct2026.05 | 80.62 | |
| GPT-4.12026.02 | 80.6 | |
| GPT-5Reasoning=Minimal Reasoning2026.02 | 80.6 | |
| PEARBackbone=Qwen-3-30B-A3B2026.05 | 80.5 | |
| gpt-oss-120bSetting=high2026.05 | 79.68 | |
| Llama4-MaverickParameters=17B, Experts=128E, Type=Instruct2026.05 | 79.56 | |
| Qwen3Parameters=30B, Active Parameters=A3B, Mode=Thinking, Version=25072026.05 | 79.46 | |
| o3-mini-2025-04-16Setting=high2026.05 | 79.24 | |
| Qwen3Parameters=235B, Active Parameters=A22B, Type=Instruct, Version=25072026.05 | 79.22 | |
| ERNIE-4.5Parameters=300B, Active Parameters=A47B, Type=PT2026.05 | 78.72 | |
| ChainBackbone=Qwen-3-30B-A3B2026.05 | 78.5 | |
| Kimi-K2Type=Instruct2026.05 | 78.39 | |
| CliqueBackbone=Qwen-3-30B-A3B2026.05 | 78 | |
| PEARBackbone=Claude-Haiku-4.52026.05 | 77.5 | |
| GLM-Z1-32B-04142026.05 | 77.42 | |
| RandomBackbone=Qwen-3-30B-A3B2026.05 | 77.2 | |
| Claude 3.5 Sonnetshot=5-shot, CoT=true2024.07 | 77 | |
| RingBackbone=Qwen-3-30B-A3B2026.05 | 77 | |
| StarBackbone=Qwen-3-30B-A3B2026.05 | 76.5 | |
| DCPOBackbone=Qwen3-4B, Optimization Method=DCPO2026.01 | 76.1 | |
| Trinity Large Previewtuning=instruct-tuned2026.02 | 75.25 | |
| GPT-5-nano-2025-08-07Setting=high2026.05 | 75.02 | |
| GPT-4oshot=5-shot, CoT=true2024.07 | 74 | |
| QwQ-32B2026.05 | 73.94 | |
| Qwen3Parameters=30B, Active Parameters=A3B, Type=Instruct, Version=25072026.05 | 73.93 | |
| Hunyuan-A13BType=Instruct2026.05 | 73.84 | |
| GLM-4-32B-04142026.05 | 73.47 | |
| Llama 3 405Bshot=5-shot, CoT=true2024.07 | 73.3 | |
| PEARBackbone=GPT-5.4-nano2026.05 | 73 | |
| gpt-oss-20bSetting=high2026.05 | 72.84 | |
| Qwen3-4BMode=Thinking, Version=25072026.05 | 72.78 | |
| Qwen3-4BBackbone=Qwen3-4B, Optimization Method=Base2026.01 | 72.67 | |
| MiniCPM-Sala2026.05 | 71.99 | |
| CliqueBackbone=Claude-Haiku-4.52026.05 | 71 | |
| ERNIE-4.5Parameters=21B, Active Parameters=A3B, Mode=Thinking2026.05 | 70.79 | |
| GPT-5-mini-2025-08-07Setting=high2026.05 | 70.24 | |
| Qwen3-4B-InstructBackbone=Qwen3-4B-Instruct2026.04 | 69.6 | |
| RandomBackbone=Claude-Haiku-4.52026.05 | 69 | |
| CliqueBackbone=GPT-5.4-nano2026.05 | 68.5 | |
| ChainBackbone=Claude-Haiku-4.52026.05 | 68 | |
| Gemma-3-27B-it2026.05 | 67.84 | |
| SpuriousModel=Qwen/Qwen3.5-27B2026.05 | 67.12 | |
| RTT-AONBackbone=Qwen3-4B-Instruct2026.04 | 66.76 | |
| GCWMModel Scale=Qwen3-8B2026.05 | 66.5 | |
| PEARBackbone=Qwen-2.5-14B2026.05 | 66.5 | |
| StarBackbone=Claude-Haiku-4.52026.05 | 66.5 | |
| Llama 3 70Bshot=5-shot, CoT=true2024.07 | 66.4 | |
| GCWMModel Scale=Qwen3-14B2026.05 | 66.2 | |
| GPT-4oAccess=Closed-source2026.01 | 66 | |
| RandomBackbone=GPT-5.4-nano2026.05 | 66 | |
| Gemma 2 27BShots=5-shot2026.02 | 65.15 | |
| ChainBackbone=GPT-5.4-nano2026.05 | 65 | |
| TAModel Scale=Qwen3-14B2026.05 | 64.9 | |
| GPT-4 (0125)shot=5-shot, CoT=true2024.07 | 64.8 | |
| PEARBackbone=Gemma-3-12B2026.05 | 64.5 | |
| CoT-SCBackbone=Qwen-3-30B-A3B2026.05 | 64.5 | |
| RingBackbone=GPT-5.4-nano2026.05 | 64.5 | |
| GCWMModel Scale=Qwen3-4B2026.05 | 63.9 |