OpenAI rolls out GPT-5.6 with model stratification and agent UX; Meta releases Muse Spark 1.1

OpenAI has launched GPT-5.6, introducing a new model ladder of Luna, Terra, and Sol alongside distinct effort levels for Max and Ultra modes. The release brings significant changes to the ChatGPT Work and Codex interface, though it initially faced user backlash regarding confusing navigation and rapid usage depletion.

• GPT-5.6 features explicit model/compute stratification with community guidance to start at lower tiers than previous versions.
• OpenAI issued multiple usage-limit resets and promised to restore familiar sidebar patterns after UX regressions were reported.
• Early benchmarks show GPT-5.6 tying in Code Arena: Frontend with Claude Fable 5 while being roughly twice as cheap, and gaining ~500 points in Presentation Elo over GPT-5.5.
• The update emphasizes orchestration and computer use, with Sol acting as a strong planner and verifier for high-throughput GUI automation.

Meta released Muse Spark 1.1, a model noted for strong UI generation and aggressive pricing. Artificial Analysis scored it at 51 on its Intelligence Index, tying it with GLM-5.2 and GPT-5.4 Luna.