← Benchmark report cards
Agentic
SWE-bench Verified
Published by Epoch AI. Source page ↗ · Data we read (file download) ↗
Epoch's own runs of the public 500-task subset; stale since mid-2026, see the live board.
Facts
| Fact | Value | Where it comes from |
|---|---|---|
| Who runs the evaluation | an independent organisation | curated by aiqbench, reviewed 2026-10-09 |
| Test set exposure | public | curated by aiqbench, reviewed 2026-10-09 |
| Data licence | CC-BY-4.0 | the board's attribution record |
| Newest published result | Jun 25, 2026 | the ingested rows |
| Last pulled by aiqbench | 57 minutes ago · every 12 h | the sync log |
| Models on this axis | 35 | the ingested rows |
| Rows with a published interval | every row | the ingested rows |
| Headroom at the top | Best 83.5%, 16.5% below the ceiling; top 10 within 6.8%. | computed from the ingested rows |
| Axes it tracks closely | Spearman ρ over shared models (see Clusters) |
Last frontier jumps
- Sep 13, 2026 · claude-opus-4-7 · 83.5%
- Sep 13, 2026 · gemini-3.5-flash · 79.3%
- Jun 16, 2026 · glm-5.2 · 78.7%
Where it appears
Weighted in the built-in templates: All-rounder ×1, Coding-focused ×3, Agentic ×3.
View this axis alone in the console →Attribution
Source: Epoch AI, 'AI Benchmarking Hub'. Published online at epoch.ai. Retrieved from https://epoch.ai/benchmarks. Licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Modified by aiqbench: normalised into aiqbench's common board format; we ingest the 'swe_bench_verified.csv' table and rescale its 0-1 scores to percent. Evaluation run by Epoch AI.