Mathematical Reasoning on AIME 2025 (Seed, VeRA-E, and ∆) benchmark leaderboard · SOTA2 Research

Mathematical Reasoning on AIME 2025 (Seed, VeRA-E, and ∆)

92Seed ScoreGLM-4.6

Evaluation Results

Method	Links
GLM-4.62026.01		92	76.3	-15.7
Gemini-3-Pro-Preview2026.01		90.7	85.7	-5
DeepSeek-V3.2-thinking2026.01		90	87.7	-2.3
GPT-5-high2026.01		89.3	85.3	-4
GPT-5.1-high2026.01		88.7	88.3	-0.3
DeepSeek-V3.1-thinking2026.01		86.7	83	-3.7
Gemini-2.5-Pro2026.01		85.3	73	-12.3
Seed-1.6-1015-high2026.01		80.7	76.7	-4
Seed-1.6-Thinking-07152026.01		80	75.3	-4.7
Claude-Sonnet-4.5-thinking2026.01		78.7	68.7	-10
Kimi-K2-thinking2026.01		78.7	71.7	-7
Seed-1.6-Lite-1015-high2026.01		77.3	70	-7.3
Minimax-M22026.01		76.7	57	-19.7
qwen3-max-09232026.01		74.7	69.7	-5
GPT-5.1-chat-latest2026.01		50.7	41.7	-9
Kimi-K2-09052026.01		47.3	44	-3.3