BALROG TextWorld (LLM): leaderboard
Text adventure games requiring language comprehension, exploration strategy, and multi-step reasoning to complete quests through natural-language descriptions.
Metric: Progress (%). Source: balrogai.com. Saturation forecast: Around January 2027. 39 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) (Thinking) | 75.69 | #58 (Gemini 3.1 Pro (Preview)) |
| 2 | Claude Opus 5 (Max) | 71.57 | #12 (Claude Opus 5) |
| 3 | Gemini 3.1 Pro (Preview) | 66.47 | #58 |
| 4 | GPT-5.6 Sol (Max) | 63.53 | #19 (GPT-5.6 Sol) |
| 5 | Grok 4 | 62.94 | #181 |
| 6 | Gemini 3 Pro | 60.2 | #80 |
| 7 | Claude Opus 4.5 (Thinking) | 59.02 | #83 (Claude Opus 4.5) |
| 8 | GPT-6 Astra (Max) | 54.51 | #6 (GPT-6 Astra) |
| 9 | Claude Opus 4.5 | 51.37 | #83 |
| 10 | Gemini 3 Flash | 50.2 | #99 |
| 11 | Gemini 2.5 Pro (Preview 03-25) | 49.22 | #123 |
| 12 | Claude Haiku 4.5 | 49.02 | #280 |
| 13 | Claude 3.5 Sonnet (20241022) | 42.06 | #291 |
| 14 | GPT-4o (2024-05-13) | 39.31 | #325 |
| 15 | Grok 3 | 32.55 | #304 |
No result here: #3 Claude Opus 5.5, #7 GPT-6.1 Sol, #8 Gemini 4 Argon.
Interactive version: theaggregate.ai/benchmark?slug=balrog-textworld-llm · How It Works · Data refreshed daily, snapshot 2026-10-20.