Dev.to
7/14/2026

This Week in AI: GPT-5.6 Launches, Modal Raises $355M for Agent Infra, and Harness Engineering Research Drops
Original: This Week in AI: GPT-5.6 Lands, Agent Infrastructure Matures, and the Model War Heats Up
Short summary
OpenAI launched GPT-5.6 as a three-model family (Sol, Terra, Luna) with a new multi-agent "ultra" effort tier, unifying ChatGPT and Codex into one desktop app. Anthropic's Claude Sonnet 5 underwhelmed while Fable 5 returned from government review as the go-to for broad agentic tasks. Modal raised $355M Series C for agent-native cloud infrastructure, and Lilian Weng published a synthesis on harness engineering covering 35 papers.
- •GPT-5.6 ships as 3-model family with native multi-agent coordination; Terra and Luna offer strong cost-performance gains
- •Claude Sonnet 5 disappoints; Fable 5 remains Anthropic's best for loosely-specified agentic work
- •Modal's $355M Series C bets on agent-native cloud infra; Lilian Weng publishes harness engineering synthesis
Generated with AI, which can make mistakes.
Is this a good recommendation for you?



