Cognition's SWE-2 achieves 92.8 on Terminal-Bench 2.1

ccdnsteve about 2 hours ago 3 commentsRead Article on tokenstead.ai

ZH version is available. Content is displayed in original English for accuracy.

Discussion (3 Comments)Read Original on HackerNews

varispeed•20 minutes ago

These tests are pointless when often models get nerfed few days after release. Astra today is way dumber than just few days ago.

cbg0•3 minutes ago

Not Astra but Sol hasn't been nerfed: https://marginlab.ai/trackers/codex/

d_tr•3 minutes ago

How and why do they get nerfed? To save money?