OpenScience launches as an open-source agent, with claimed benchmark leads over Codex and Claude Code
A post describes Synthetic Sciences’ OpenScience as a free, Apache 2.0-licensed workbench for any provider’s model. It reports that OpenScience solved 53 of 70 Terminal-Bench-Science tasks, scoring 75.7%.
TLDR
A post says Synthetic Sciences launched OpenScience as an open-source agent. It reports a 75.7% score on Terminal-Bench-Science, compared with 68.1% for Codex with GPT-6 Astra on a September 23 mirror of the public leaderboard. On Terminal-Bench 4.0’s 14 science tasks, the post reports 71.4% for OpenScience and 60.0% for Claude Code on Claude Fable 5.1.
Combined views
4K
2 Sources, first seen 1h ago