Benchmarks / Professional Domains
MedQA
Medical licensing-exam question answering.
maintained by Vals AI · www.vals.ai · 95 observations on record
Versions
1 version
| Version | Tasks |
|---|---|
| MedQA current medqa | — |
Current leaderboard
MedQA · Accuracy · top 25 of 95 systems
| # | System | Accuracy % |
|---|---|---|
| 1 | O1 (2024-12-17)high effort OpenAI | 96.5% |
| 2 | GPT 5.1 (2025-11-13)high effort OpenAI | 96.4% |
| 3 | Gemini 3.1 Pro Previewhigh effort Google DeepMind | 96.4% |
| 4 | GPT 5 (2025-08-07)high effort OpenAI | 96.3% |
| 5 | GPT 5.4 (2026-03-05)xhigh effort OpenAI | 96.1% |
| 6 | O3 (2025-04-16)high effort OpenAI | 96.1% |
| 7 | GPT 5 Mini (2025-08-07)high effort OpenAI | 96.1% |
| 8 | Gemini 3 Pro Previewhigh effort Google DeepMind | 96.0% |
| 9 | O4 Mini (2025-04-16)high effort OpenAI | 96.0% |
| 10 | Claude Opus 4.5 20251101thinking Anthropic | 95.9% |
| 11 | Gemini 3 Flash Previewhigh effort Google DeepMind | 95.8% |
| 12 | Claude Opus 4.6thinking Anthropic | 95.4% |
| 13 | Qwen3 5 Plusthinking Alibaba | 95.2% |
| 14 | O3 Mini (2025-01-31)high effort OpenAI | 94.8% |
| 15 | Claude Sonnet 4.5 20250929thinking Anthropic | 94.7% |
| 16 | Grok 4.20 0309 Reasoning xAI | 94.5% |
| 17 | Kimi K2 5thinking Moonshot AI | 94.4% |
| 18 | GLM 5thinking Z.ai | 94.3% |
| 19 | GPT 5.2 (2025-12-11)xhigh effort OpenAI | 94.1% |
| 20 | Deepseek V3p2high effort Fireworks AI | 93.9% |
| 21 | GLM 4.7 Z.ai | 93.7% |
| 22 | Claude Opus 4.1 20250805thinking Anthropic | 93.6% |
| 23 | GPT 5 Nano (2025-08-07)high effort OpenAI | 93.3% |
| 24 | Claude Opus 4.5 20251101 Anthropic | 93.2% |
| 25 | Gemini 2.5 Pro Exp 03.25 Google DeepMind | 93.1% |
One row per evaluated system — reasoning-effort variants rank separately.