LLM Stats (FrontierSWE): leaderboard
LLM Stats mirror of FrontierSWE, an ultra-long-horizon coding agent benchmark with 20-hour software engineering, performance, and ML research tasks.
Metric: Score (%). Source: llm-stats.com. 16 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Fable 5 | 90 | #16 |
| 2 | Kimi K3 | 81.2 | #30 |
| 3 | GLM-5.3 | 78.1 | #39 |
| 4 | Claude Opus 4.8 | 75 | #37 |
| 5 | GLM-5.2 | 74 | #73 |
| 6 | Qwen3.8 Max | 73.5 | |
| 7 | GPT-5.5 | 73 | #31 |
| 8 | Claude Opus 4.7 | 63 | #49 |
| 9 | Claude Opus 4.6 | 56 | #63 |
| 10 | GPT-5.4 | 54 | #79 |
Interactive version: theaggregate.ai/benchmark?slug=llm-stats-frontierswe · How It Works · Data refreshed daily, snapshot 2026-10-20.