shipfeedAI news, curated daily

22:53:42 CET
13 AUG22:53:42shipfeed
pull to refreshlast sync
Just in — 30 new
§ evals · storyline

DeepSeek-V4 Flash 0731 vs GPT-5.6 Luna on DeepSWE: Cost and Coding

DeepSeek-V4 Flash and GPT-5.6 Luna benchmarked on DeepSWE show Luna leading pass@1 by 14 points and DeepSeek delivering 4.8x the solves per dollar.

Aug 6 · · primary fetch1 sourceupdated Aug 6 ·

We ran 900 DeepSWE rollouts on DeepSeek-V4 Flash and GPT-5.6 Luna. Luna leads pass@1 by 14 points; DeepSeek delivers 4.8x the solves per dollar.

read full article on together.ai
§ sources1 publication · timeline below
  1. together.aiDeepSeek-V4 Flash 0731 vs GPT-5.6 Luna on DeepSWE: Cost and Codingprimary