BALROG leaderboard
- GPT-6 Astra leads
- 68.3%
- models measured
- 22
- lead over #2
- +4.9 pts
- median
- 31.8%
Leader on BALROG · read October 21
On October 21, 2026, 28 of our catalog's models had a BALROG result, and OpenAI GPT-6 Astra led with 68.3%, ahead of Anthropic Claude Opus 5 at 63.4%.
Text and image games an agent has to play through; the number is the average progress across games.
Role in the index
BALROG is shown as published and does not feed any of our indexes. Read the methodology
| # | Model | Result | Evidence | Coding index | Output $/Mtok |
|---|---|---|---|---|---|
| 1 | GPT-6 AstraOpenAI | 68.3% | 85.1 | $50.00 | |
| 2 | Claude Opus 5Anthropic | 63.4% | 83.0 | $25.00 | |
| 3 | GPT-5.6 SolOpenAI | 60.0% | 73.5 | $20.00 | |
| 4 | Gemini 3.1 Pro PreviewGoogle | 57.0% | 32.3 | $12.00 | |
| 5 | GPT-5.6 TerraOpenAI | 53.2% | 63.9 | $12.00 | |
| 6 | Gemini 3 Flash PreviewGoogle | 48.1% | 43.7 | $3.00 | |
| 7 | GPT-5.6 LunaOpenAI | 45.6% | 58.0 | $1.20 | |
| 8 | Claude Opus 4.5Anthropic | 43.5% | 44.0 | $25.00 | |
| 9 | R1RetiredDeepSeek | 34.9% | Not benchmarked | $2.50 | |
| 10 | Gemini 2.5 FlashGoogle | 33.5% | Not benchmarked | $2.50 | |
| 11 | GPT-5OpenAI | 32.8% | 38.8 | $10.00 | |
| 12 | Claude 3.5 SonnetRetiredAnthropic | 32.6% | Not benchmarked | $15.00 | |
| 13 | GPT-4o (2024-05-13)OpenAI | 32.3% | Not benchmarked | $15.00 | |
| 14 | Claude Haiku 4.5Anthropic | 31.2% | 35.6 | $5.00 | |
| 15 | Grok 3RetiredxAI | 29.5% | Not benchmarked | $15.00 | |
| 16 | Reka Flash 3Reka | 29.2% | Not benchmarked | $0.20 | |
| 17 | Llama 3.1 70B InstructMeta Llama | 27.9% | Not benchmarked | $0.56 | |
| 18 | Llama 3.3 70B InstructMeta Llama | 23.0% | Not benchmarked | $0.72 | |
| 19 | Gemini 1.5 ProRetiredGoogle | 21.0% | Not benchmarked | $5.00 | |
| 20 | Claude 3.5 HaikuRetiredAnthropic | 19.3% | Not benchmarked | $4.00 | |
| 21 | GPT-4o-mini (2024-07-18)OpenAI | 17.4% | Not benchmarked | $0.60 | |
| 22 | Qwen2.5 72B InstructQwen | 16.2% | Not benchmarked | $0.40 | |
| 23 | Llama 3.1 8B InstructMeta Llama | 15.1% | Not benchmarked | $0.08 | |
| 24 | Gemini 1.5 FlashRetiredGoogle | 14.6% | Not benchmarked | $0.30 | |
| 25 | Phi 4Microsoft | 11.6% | Not benchmarked | $0.14 | |
| 26 | Llama 3.2 3B InstructMeta Llama | 10.1% | Not benchmarked | $0.3325 | |
| 27 | Qwen2.5 7B InstructQwen | 7.8% | Not benchmarked | $0.20 | |
| 28 | Llama 3.2 1B InstructMeta Llama | 6.6% | Not benchmarked | $0.201 |
Only models in our catalog are listed, and each appears once with its best published run. 215 rows were read from the source on October 21, 2026, and rows that match no catalog model are left out.