OpenAI launches GPT-5.6 and ChatGPT Work as voluntary AI review faces scrutiny

OpenAI’s GPT-5.6 family (Sol, Terra, Luna) and new ChatGPT Work agent launched publicly after a controversial voluntary government review. Anthropic revealed a groundbreaking interpretability technique (J-lens) that peers inside Claude’s hidden reasoning, and also launched a usage-tracking dashboard called Reflect. Meta fired back with Muse Spark 1.1, offering multi-agent orchestration and aggressive API pricing that undercuts competitors.

Written by AI from the sources it cites; it can be wrong — follow any citation to confirm.

10 sources1 min read

OpenAI launches GPT-5.6 family and ChatGPT Work globally after government review The three-model tier (Sol, Terra, Luna) is now publicly available, and OpenAI also introduced ChatGPT Work, an agentic tool for automating business workflows across apps. The release follows a 12-day voluntary government review that critics say remains opaque and lacks clear standards. 1 2 3 4 5

Anthropic reveals Claude's hidden 'J-space' and launches Reflect dashboard Using a new technique called the Jacobian lens, Anthropic discovered a hidden internal space in Claude Opus 4.6 that reveals words the model is 'thinking' before it speaks. Separately, Claude's new Reflect feature lets users track their AI usage habits and encourages mindful interaction. 6 7

Meta releases Muse Spark 1.1 with multi-agent AI and aggressive API pricing Meta's new flagship model supports multi-agent orchestration, a 1M-token context window, and costs $4.25 per million output tokens, undercutting Grok 4.5 and putting pressure on OpenAI and Anthropic. The model is available via the new Meta Model API in public preview. 8 9 10