← All reports

The open-source agent developed by the user outperformed other agents in the TerminalBench benchmark using the Gemini-3-flash-preview model.

AISoftware DevelopmentApr 27, 2026score 0.172 posts · 0 replies across 1 instances
A user shared an open-source agent they built that achieved top performance in the TerminalBench on the Gemini-3-flash-preview model, sparking discussion on Hacker News.

Claims

The open-source agent developed by the user outperformed other agents in the TerminalBench benchmark using the Gemini-3-flash-preview model.
Parent: AIEntity: Gemini-3-flash-previewImpact: positiveDate: Apr 27, 2026Target: The performance of the open-source agent in the TerminalBench benchmark using the Gemini-3-flash-preview model.

Source posts

@[email protected]
Show HN: OSS Agent I built topped the TerminalBench on Gemini-3-flash-preview L: https://github.com/dirac-run/dirac C: https://news.ycombinator.com/item?id=47920787 posted on 2026.04.27 at 08:35:55 (c=0, p=3)
0 boosts · 0 favs · 0 replies · Apr 27, 2026
@[email protected]
Show HN: OSS Agent I built topped the TerminalBench on Gemini-3-flash-preview Link: https://github.com/dirac-run/dirac Comments: https://news.ycombinator.com/item?id=47920787
0 boosts · 0 favs · 0 replies · Apr 27, 2026
@[email protected]
Show HN: OSS Agent I built topped the TerminalBench on Gemini-3-flash-preview https://github.com/dirac-run/dirac Discussion: https://news.ycombinator.com/item?id=47920787
1 boosts · 0 favs · 0 replies · Apr 27, 2026