Microsoft AI MAI-Thinking-1 85.0% microsoft.ai for MAI-Thinking-1 on MMLU-Pro Original (opens in a new tab) Accuracy Reported August 12, 2026 Technical report Microsoft AI Self-reported Microsoft Record for MAI-Thinking-1 on MMLU-Pro Original 70003 Google (DeepMind) Gemma 4 26B A4B Instruct (thinking mode)82.6% huggingface.co for Gemma 4 26B A4B Instruct (thinking mode) on MMLU-Pro Original (opens in a new tab) Accuracy Reported July 24, 2026 Model card Google Self-reported Google DeepMind Record for Gemma 4 26B A4B Instruct on MMLU-Pro Original 35001 Google (DeepMind) Gemma 4 31B Instruct (thinking mode)85.2% huggingface.co for Gemma 4 31B Instruct (thinking mode) on MMLU-Pro Original (opens in a new tab) Accuracy Reported July 24, 2026 Model card Google Self-reported Google DeepMind Record for Gemma 4 31B Instruct on MMLU-Pro Original 35002 Google (DeepMind) Gemma 4 E2B Instruct (thinking mode)60.0% huggingface.co for Gemma 4 E2B Instruct (thinking mode) on MMLU-Pro Original (opens in a new tab) Accuracy Reported July 24, 2026 Model card Google Self-reported Google DeepMind Record for Gemma 4 E2B Instruct on MMLU-Pro Original 35003 Google (DeepMind) Gemma 4 E4B Instruct (thinking mode)69.4% huggingface.co for Gemma 4 E4B Instruct (thinking mode) on MMLU-Pro Original (opens in a new tab) Accuracy Reported July 24, 2026 Model card Google Self-reported Google DeepMind Record for Gemma 4 E4B Instruct on MMLU-Pro Original 35004 Google (DeepMind) Gemma 4 12B Instruct (thinking mode)77.2% huggingface.co for Gemma 4 12B Instruct (thinking mode) on MMLU-Pro Original (opens in a new tab) Accuracy Reported July 24, 2026 Model card Google Self-reported Google DeepMind Record for Gemma 4 12B Instruct on MMLU-Pro Original 35005 Alibaba (Tongyi) Qwen3.7-Plus 88.5% alibabacloud.com for Qwen3.7-Plus on MMLU-Pro Original (opens in a new tab) Accuracy Reported June 3, 2026 Launch post Alibaba Cloud Community Self-reported Qwen Team Record for Qwen3.7-Plus on MMLU-Pro Original 130007 Alibaba (Tongyi) Qwen3.7-Max 89.6% qwen.ai for Qwen3.7-Max on MMLU-Pro Original (opens in a new tab) Accuracy Reported May 20, 2026 Launch post Qwen Team Self-reported Record for Qwen3.7-Max on MMLU-Pro Original 130006 DeepSeek AI DeepSeek-V4-Flash (non-thinking)83.0% huggingface.co for DeepSeek-V4-Flash (non-thinking) on MMLU-Pro Original (opens in a new tab) Accuracy Reported April 24, 2026 Model card DeepSeek Self-reported DeepSeek AI Record for DeepSeek-V4-Flash on MMLU-Pro Original 110001 DeepSeek AI DeepSeek-V4-Flash (max)86.2% huggingface.co for DeepSeek-V4-Flash (max) on MMLU-Pro Original (opens in a new tab) Accuracy Reported April 24, 2026 Model card DeepSeek Self-reported DeepSeek AI Record for DeepSeek-V4-Flash on MMLU-Pro Original 110001 DeepSeek AI DeepSeek-V4-Flash (high)86.4% huggingface.co for DeepSeek-V4-Flash (high) on MMLU-Pro Original (opens in a new tab) Accuracy Reported April 24, 2026 Model card DeepSeek Self-reported DeepSeek AI Record for DeepSeek-V4-Flash on MMLU-Pro Original 110001 DeepSeek AI DeepSeek-V4-Pro (max)87.5% huggingface.co for DeepSeek-V4-Pro (max) on MMLU-Pro Original (opens in a new tab) Accuracy Reported April 24, 2026 Model card DeepSeek Self-reported DeepSeek AI Record for DeepSeek-V4-Pro on MMLU-Pro Original 110002 DeepSeek AI DeepSeek-V4-Pro (high)87.1% huggingface.co for DeepSeek-V4-Pro (high) on MMLU-Pro Original (opens in a new tab) Accuracy Reported April 24, 2026 Model card DeepSeek Self-reported DeepSeek AI Record for DeepSeek-V4-Pro on MMLU-Pro Original 110002 DeepSeek AI DeepSeek-V4-Pro (non-thinking)82.9% huggingface.co for DeepSeek-V4-Pro (non-thinking) on MMLU-Pro Original (opens in a new tab) Accuracy Reported April 24, 2026 Model card DeepSeek Self-reported DeepSeek AI Record for DeepSeek-V4-Pro on MMLU-Pro Original 110002 Alibaba (Tongyi) Qwen3.6-35B-A3B 85.2% alibabacloud.com for Qwen3.6-35B-A3B on MMLU-Pro Original (opens in a new tab) Accuracy Reported April 17, 2026 Launch post Alibaba Cloud Community Self-reported Qwen Team Record for Qwen3.6-35B-A3B on MMLU-Pro Original 130005 Alibaba (Tongyi) Qwen3.6-Plus 88.5% alibabacloud.com for Qwen3.6-Plus on MMLU-Pro Original (opens in a new tab) Accuracy Reported April 2, 2026 Launch post Alibaba Cloud Community Self-reported Qwen Team Record for Qwen3.6-Plus on MMLU-Pro Original 130004 Mistral AI Mistral Small 4 (high)78.0% huggingface.co for Mistral Small 4 (high) on MMLU-Pro Original (opens in a new tab) Accuracy Reported March 16, 2026 Model card Mistral AI Self-reported Record for Mistral Small 4 on MMLU-Pro Original 90004 Mistral AI Mistral Small 4 (none)73.5% huggingface.co for Mistral Small 4 (none) on MMLU-Pro Original (opens in a new tab) Accuracy Reported March 16, 2026 Model card Mistral AI Self-reported Record for Mistral Small 4 on MMLU-Pro Original 90004 NVIDIA NVIDIA Nemotron 3 Super 120B-A12B 83.7% huggingface.co for NVIDIA Nemotron 3 Super 120B-A12B on MMLU-Pro Original (opens in a new tab) Accuracy Reported March 11, 2026 Model card NVIDIA Self-reported Record for NVIDIA Nemotron 3 Super 120B-A12B on MMLU-Pro Original 60002 Alibaba (Tongyi) Qwen3.5-397B-A17B 87.8% alibabacloud.com for Qwen3.5-397B-A17B on MMLU-Pro Original (opens in a new tab) Accuracy Reported February 17, 2026 Launch post Alibaba Cloud Community Self-reported Qwen Team Record for Qwen3.5-397B-A17B on MMLU-Pro Original 130003 Moonshot AI Kimi K2.5 (thinking)87.1% huggingface.co for Kimi K2.5 (thinking) on MMLU-Pro Original (opens in a new tab) Accuracy Reported January 27, 2026 Model card Moonshot AI Self-reported Record for Kimi K2.5 on MMLU-Pro Original 120001 Alibaba (Tongyi) Qwen3-Max-Thinking 85.7% alibabacloud.com for Qwen3-Max-Thinking on MMLU-Pro Original (opens in a new tab) Accuracy Reported January 25, 2026 Launch post Alibaba Cloud Community Self-reported Qwen Team Record for Qwen3-Max-Thinking on MMLU-Pro Original 130002 Z.ai GLM-4.7 84.3% z.ai for GLM-4.7 on MMLU-Pro Original (opens in a new tab) Accuracy Reported December 22, 2025 Launch post Z.ai Self-reported Record for GLM-4.7 on MMLU-Pro Original 150001 NVIDIA NVIDIA Nemotron 3 Nano 30B-A3B 78.3% huggingface.co for NVIDIA Nemotron 3 Nano 30B-A3B on MMLU-Pro Original (opens in a new tab) Accuracy Reported December 15, 2025 Model card NVIDIA Self-reported Record for NVIDIA Nemotron 3 Nano 30B-A3B on MMLU-Pro Original 60001