ByteBrief
We're a portrait publication through and through. Turn your phone back and your briefing picks up right where you left it.
(We tried widescreen once. It wasn't us.)

Abacus.AI released the Smaug line of open-weight models fine-tuned for agentic workloads. Smaug Flash gained 14.3 points in agentic coding, while Smaug Mini beat GPT-5.6 Luna on IFBench. Smaug Agentic ran 113 tasks over 7+ hours with zero infrastructure errors, prioritizing behavioral stability over raw score deltas.
Tap to vote and see what everyone thinks.
Summary by ByteBrief
Anthropic Releases Mythos 5.2 AI Model