BabyAI: leaderboard
BALROG BabyAI task-completion percentages mirrored by BenchmarkList, with text-only LLM and image-plus-text VLM runs in the same source table.
Metric: BabyAI score (%). Source: benchmarklist.com. 44 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 100 | #58 |
| 2 | Gemini 3.1 Pro (Preview) (Thinking) | 98 | #58 (Gemini 3.1 Pro (Preview)) |
| 3 | Gemini 3 Pro | 96 | #80 |
| 4 | Gemini 3 Flash | 86 | #99 |
| 5 | Gemini 2.5 Pro (Preview 03-25) | 80 | #123 |
| 6 | Claude Opus 4.5 | 80 | #83 |
| 7 | GPT-5 (Minimal) | 80 | #98 (GPT-5) |
| 8 | GPT-4o (2024-05-13) | 77.6 | #325 |
| 9 | DeepSeek R1 | 74 | #251 |
| 10 | Claude Opus 4.5 (Thinking) | 72 | #83 (Claude Opus 4.5) |
| 11 | Gemini 2.5 Flash | 68 | #245 |
| 12 | Claude Haiku 4.5 | 66 | #280 |
| 13 | Grok 3 | 62 | #304 |
| 14 | Gemini 1.5 Pro (002) | 58.4 | #349 |
| 15 | Claude 3.5 Haiku | 52 | #568 |
Interactive version: theaggregate.ai/benchmark?slug=babyai · How It Works · Data refreshed daily, snapshot 2026-10-20.