SWE Atlas: Refactoring leaderboard
- GPT-6 Astra leads
- 59.1%
- models measured
- 14
- lead over #2
- +2.4 pts
- median
- 44.5%
Leader on SWE Atlas: Refactoring · read October 22
# SWE Atlas: Refactoring, read October 22
leaders = {
"GPT-6 Astra": 59.1%, "Claude Fable 5.1": 56.7%, "Claude Fable 5": 54.8%,
}
On October 22, 2026, 14 of our catalog's models had a SWE Atlas: Refactoring result, and OpenAI GPT-6 Astra led with 59.1%, ahead of Anthropic Claude Fable 5.1 at 56.7%.
SWE Atlas Refactoring tests whether a coding agent can restructure code in real repositories while keeping its behavior. Scale AI runs each model inside an agent harness; we show the best result per model and name the harness.
Role in the index
A coding board we show beside the Vibe Coding Index, not inside it: the index keeps the six boards it was defined on, so a score does not move every time a board is added. Read the methodology
| # | Model | Result | Evidence | Coding index | Output $/Mtok |
|---|---|---|---|---|---|
| 1 | GPT-6 AstraOpenAI | 59.1% | 85.1 | $50.00 | |
| 2 | Claude Fable 5.1Anthropic | 56.7% | 84.2 | $50.00 | |
| 3 | Claude Fable 5Anthropic | 54.8% | 79.3 | $50.00 | |
| 4 | Claude Opus 4.7Anthropic | 48.6% | 67.5 | $25.00 | |
| 5 | Claude Opus 4.8Anthropic | 46.7% | 65.0 | $25.00 | |
| 6 | GPT-5.5OpenAI | 44.8% | 71.3 | $30.00 | |
| 7 | Gemini 3.8 FlashGoogle | 44.8% | 61.6 | $3.75 | |
| 8 | GPT-5.4OpenAI | 44.3% | 64.4 | $15.00 | |
| 9 | GLM 5.2Z.ai | 42.4% | 61.1 | $4.40 | |
| 10 | Claude Opus 4.6Anthropic | 35.6% | 57.2 | $25.00 | |
| 11 | Claude Sonnet 4.6Anthropic | 32.2% | 57.3 | $15.00 | |
| 12 | GLM 5Z.ai | 24.2% | 41.7 | $3.20 | |
| 13 | Kimi K2.5Moonshot AI | 20.9% | 38.6 | $2.85 | |
| 14 | MiniMax M2.5MiniMax | 19.5% | 37.2 | $1.20 |
Only models in our catalog are listed, and each appears once with its best published run. 66 rows were read from the source on October 22, 2026, and rows that match no catalog model are left out.