FrontierSWE leaderboard
- GPT-6 Astra leads
- 65.5%
- models measured
- 20
- lead over #2
- +3.2 pts
- median
- 29.8%
Leader on FrontierSWE · read October 22
# FrontierSWE, read October 22
leaders = {
"GPT-6 Astra": 65.5%, "Claude Opus 5.5": 62.3%, "Claude Sonnet 5.5": 61.9%,
}
On October 22, 2026, 20 of our catalog's models had a FrontierSWE result, and OpenAI GPT-6 Astra led with 65.5%, ahead of Anthropic Claude Opus 5.5 at 62.3%.
Epoch AI collects its results; its benchmarking hub describes the tasks and links the authors.
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 | Average cost per task | Best of 5 runs | Hours per task | Evidence | Coding index | Output $/Mtok |
|---|---|---|---|---|---|---|---|---|
| 1 | GPT-6 AstraOpenAI | 65.5% | $1029.65 | 73.0% | 12.1 h | 85.1 | $50.00 | |
| 2 | Claude Opus 5.5Anthropic | 62.3% | $98.87 | 71.5% | 14.3 h | 88.8 | $20.00 | |
| 3 | Claude Sonnet 5.5Anthropic | 61.9% | $110.45 | 71.7% | 16.0 h | 87.3 | $10.00 | |
| 4 | Claude Fable 5.1Anthropic | 56.3% | $138.55 | 66.6% | 11.6 h | 84.2 | $50.00 | |
| 5 | Gemini 4 ArgonGoogle | 55.0% | $129.36 | 64.3% | 10.6 h | 81.0 | $10.00 | |
| 6 | Claude Opus 5Anthropic | 52.0% | $196.71 | 60.7% | 16.1 h | 83.0 | $25.00 | |
| 7 | Claude Fable 5Anthropic | 47.0% | $301.28 | 61.8% | 15.8 h | 79.3 | $50.00 | |
| 8 | Claude Haiku 5.5Anthropic | 43.8% | $20.55 | 52.8% | 12.4 h | 70.7 | $0.50 | |
| 9 | GPT-5.6 SolOpenAI | 32.2% | $179.64 | 44.2% | 8.6 h | 73.5 | $20.00 | |
| 10 | GLM 5.3Z.ai | 30.2% | $97.22 | 40.6% | 17.0 h | 70.6 | $4.40 | |
| 11 | Grok 4.7xAI | 29.5% | $318.77 | 41.1% | 12.1 h | 66.7 | $6.00 | |
| 12 | Kimi K3Moonshot AI | 25.9% | $109.71 | 37.5% | 18.4 h | 62.3 | $15.00 | |
| 13 | Muse Spark 1.3Meta | 25.8% | $243.34 | 40.9% | 17.0 h | 57.4 | $4.25 | |
| 14 | Grok 4.6xAI | 25.3% | $243.43 | 37.3% | 13.8 h | 65.7 | $6.00 | |
| 15 | Gemini 3.7 FlashGoogle | 20.3% | $34.14 | 31.2% | 7.9 h | 58.0 | $3.75 | |
| 16 | Gemini 3.8 FlashGoogle | 19.6% | $38.75 | 31.6% | 7.2 h | 61.6 | $3.75 | |
| 17 | GLM 5.3 FlashZ.ai | 18.1% | $11.03 | 28.4% | 16.4 h | 57.7 | $0.50 | |
| 18 | Qwen3.8 Max (0902)Qwen | 17.8% | $51.27 | 26.8% | 19.4 h | 62.4 | $6.00 | |
| 19 | Muse Spark 1.2Meta | 12.0% | $27.81 | 18.4% | 4.6 h | 55.7 | $4.25 | |
| 20 | T InklingThinking Machines | 4.1% | $9.15 | 10.8% | 1.2 h | 31.5 | $4.05 |
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 22, 2026, and rows that match no catalog model are left out. Average cost per task is the average spend on one task, in US dollars, as Epoch AI records it beside the result. Every other column is a figure the board publishes for the same run.
| Model | Result | Implementation | Performance | Research |
|---|---|---|---|---|
| OpenAI GPT-6 Astra | 65.5% | 87.0% | 48.3% | 59.1% |
| Anthropic Claude Opus 5.5 | 62.3% | 80.4% | 45.3% | 57.3% |
| Anthropic Claude Sonnet 5.5 | 61.9% | 81.1% | 48.1% | 54.5% |
| Anthropic Claude Fable 5.1 | 56.3% | 73.2% | 43.1% | 51.3% |
| Google Gemini 4 Argon | 55.0% | 74.5% | 35.0% | 47.7% |
| Anthropic Claude Opus 5 | 52.0% | 70.5% | 37.8% | 42.5% |
| Anthropic Claude Fable 5 | 47.0% | 63.7% | 33.7% | 40.5% |
| Anthropic Claude Haiku 5.5 | 43.8% | 54.3% | 28.7% | 37.5% |
| OpenAI GPT-5.6 Sol | 32.2% | 40.4% | 22.0% | 28.2% |
| Z.ai GLM 5.3 | 30.2% | 34.6% | 20.6% | 25.6% |
| SpaceXAI Grok 4.7 | 29.5% | 36.4% | 18.2% | 23.8% |
| MoonshotAI Kimi K3 | 25.9% | 29.9% | 18.8% | 21.1% |
| Meta Muse Spark 1.3 | 25.8% | 28.8% | 17.1% | 22.1% |
| SpaceXAI Grok 4.6 | 25.3% | 24.6% | 16.4% | 25.2% |
| Google Gemini 3.7 Flash | 20.3% | 24.9% | 15.0% | 14.3% |
| Google Gemini 3.8 Flash | 19.6% | 23.0% | 14.9% | 16.5% |
| Z.ai GLM 5.3 Flash | 18.1% | 19.5% | 15.8% | 13.3% |
| Qwen3.8 Max (0902) | 17.8% | 13.8% | 16.2% | 16.0% |
| Meta Muse Spark 1.2 | 12.0% | 2.6% | 11.5% | 14.7% |
| Thinking Machines Inkling | 4.1% | 0.4% | 3.9% | 5.4% |
The parts as Epoch AI benchmarking hub publishes them, for the same run as the result.