Terminal-Bench 4.0 (mini-swe-agent): leaderboard
Terminal-Bench 4.0 resolution rate (%) for runs that the tbench.ai leaderboard lists under the lowercase agent label mini-swe-agent; runs listed as mini-SWE-agent appear on a separate board. Each row keeps its source reasoning effort, success and trial counts and confidence interval; the literal source effort none is not taken to mean that reasoning was disabled.
Metric: Resolution Rate (%). Source: www.tbench.ai. Saturation forecast: Rough model projection: around 2026. 4 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Beam 1.0 (Terminal-Bench Effort: none) | 6.36 | |
| 2 | Inkling Small (Terminal-Bench Effort: none) | 0.91 | |
| 3 | Nemotron 3.5 Super VL (Terminal-Bench Effort: none) | 0.3 | |
| 4 | Inkling (Terminal-Bench Effort: none) | 0 |
Interactive version: theaggregate.ai/benchmark?slug=terminal-bench-4-0-mini-swe-agent-a5tm7i · How It Works · Data refreshed daily, snapshot 2026-11-12.