Medical Calculation and Tool Use on MedMCP-Calc Overall benchmark leaderboard · SOTA2 Research

Medical Calculation and Tool Use on MedMCP-Calc Overall

100Task Success RateClaude Opus 4.5

Evaluation Results

Method	Links
Claude Opus 4.52026.01		100	66.66	71.08	33.92
Claude Sonnet 4.52026.01		100	65.6	53.86	27.8
GPT-52026.01		100	59.81	53.12	26.7
Gemini-2.5-Pro2026.01		100	55.96	40.45	20.77
Qwen3-235B-Instruct-25072026.01		100	41.74	34.24	10.76
Qwen3-235B-Thinking-25072026.01		100	38.82	26.33	11.14
Qwen3-235B non-thinking2026.01		100	31.73	38.26	7.44
Qwen3-235B thinking2026.01		100	33.42	30.6	7.91
Llama-3.3-70B2026.01		100	11.6	22.55	2.83
Qwen3-30B-Instruct-25072026.01		100	28.91	23.94	5.95
Qwen3-30B-Thinking-25072026.01		100	29.74	19.23	7.6
Qwen3-4B-Instruct-25072026.01		100	19.95	9.34	3.08
CalcMate-4B2026.01		100	56.97	71.06	15.02
CalcMate-30B2026.01		100	65.66	75.31	20.07
Deepseek-V3.12026.01		99.15	44.33	41.04	12.37
Qwen3-VL-235B2026.01		99.15	33.74	25.67	9.94
Qwen3-Next-80B-Instruct2026.01		99.15	38.57	21.64	7.53
Qwen3-Next-80B-Thinking2026.01		99.15	33.3	34.1	11.03
Llama-3.1-70B2026.01		99.15	11.87	19.3	1.01
Gemini-3-Pro2026.01		97.46	54.05	55.45	26.49
Qwen2.5-72B2026.01		97.46	29.37	38.49	7.41
MedGemma-27B-IT2026.01		97.46	30.33	18.45	3.61
GLM-4.72026.01		96.61	51.89	50.06	15.59
Kimi-K2-Instruct2026.01		96.61	41.95	28.14	8.27
Baichuan-M2-32B2026.01		79.66	33.95	33.98	7.77