RABBITS MedQA G2B: leaderboard
RABBITS MedQA generic-to-brand substituted accuracy on four-option medical exam questions.
Metric: Accuracy (%). Source: huggingface.co. Saturation forecast: Estimated already saturated. 26 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 1.5 Flash | 96.03 | #618 |
| 2 | GPT-3.5 Turbo | 96.03 | #865 |
| 3 | GPT-4 | 89.95 | #438 |
| 4 | GPT-4o | 88.36 | #343 |
| 5 | Gemini 1.5 Pro | 87.3 | #445 |
| 6 | Claude 3 Opus | 83.33 | #571 |
| 7 | Qwen 2 72B | 74.07 | #521 |
| 8 | Llama 3 70B | 72.75 | #631 |
| 9 | Mixtral-8x22B-v0.1 | 67.46 | #711 |
| 10 | Mixtral 8x7B (v0.1) | 60.05 | #1083 |
| 11 | Yi 1.5 34B | 59.79 | #807 |
| 12 | c4ai-command-r-plus | 56.61 | #724 |
| 13 | Llama 3 8B | 55.03 | #1242 |
| 14 | Qwen 2 7B | 53.7 | #1095 |
| 15 | Phi-3-medium-4k-instruct | 53.44 | #952 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=rabbits-medqa-g2b · How It Works · Data refreshed daily, snapshot 2026-10-20.