Meta released Llama 4 (405B and 70B) with native multimodal support, open weights and a 256K context window. Anthropic's Claude 4 Sonnet leads the SWE-bench and Aider leaderboards, dethroning GPT-4o on autonomous coding tasks. EU AI Act enforcement timelines tightened, with the first high-risk audits scheduled for Q3.
One weekly email. The signal, none of the noise. Unsubscribe anytime.