AutoMedBench Full
Leaderboard
Average model performance across all 96 task-tier combinations (48 tasks × Lite/Standard).
96-combination average
Model score comparison
Compare every model on the same 0–100 scale. Model-colored columns are generated directly from the leaderboard values below.
Scroll horizontally to compare every model.
Overall = 1/2 Agentic + 1/2 Task
Muted model colors · Higher is better · 0–100
Leaderboard
96-task average
Overall = 1/2 Agentic + 1/2 Task
| Rank | Model | Release | Overall | Agentic | Task | Avg cost | Avg time | Avg turns |
|---|---|---|---|---|---|---|---|---|
| 1 |
Claude Opus 4.6Anthropic
|
2026-02-05 | 73.04 | 87.97 | 58.10 | $10.49 | 13.2 min | 23.6 |
| 2 |
GLM-5Z.ai
|
2026-02-12 | 70.38 | 85.30 | 55.47 | $2.36 | 18.3 min | 39.3 |
| 3 |
Qwen3.5Alibaba Cloud
|
2026-02-15 | 66.43 | 81.28 | 51.59 | $1.96 | 14.9 min | 31.7 |
| 4 |
Gemini 3.1 ProGoogle
|
2026-02-19 | 66.36 | 80.00 | 52.72 | $4.27 | 17.2 min | 20.9 |
| 5 |
MiniMax M2.5MiniMax
|
2026-02-12 | 61.21 | 76.00 | 46.43 | $2.34 | 20.5 min | 35.0 |
| 6 |
ChatGPT 5.4OpenAI
|
2026-03-05 | 56.37 | 68.35 | 44.38 | $2.75 | 10.4 min | 10.0 |
Scores and operational metrics are equal-weight averages across all 96 task-tier combinations (48 tasks × Lite/Standard). Cost is USD; time is wall-clock; turns are conversational turns.