Understand this story
On June 29, 2026, it was reported that the United States has formed a de facto model licensing regime, and OpenAI's latest model, GPT-5.6, is awaiting government approval. Previously, on June 12, the Trump administration forced Anthropic to take down its two most powerful models, and the industry is watching whether this will be extended to the entire AI industry. On July 3, the DeepLearning.AI weekly newsletter showed that OpenAI has released three closed-source vision-language models, GPT-5.6 Sol, Terra, and Luna, which are currently available only to about 20 U.S. government-approved institutions, with input prices ranging from $5 to $1 per million tokens; OpenAI said it will expand access in the coming weeks.
AI-generated from public reports
Developments
2 confirmed developments
- Source:English text is being prepared.
- Source:Timothy B. Lee：Understanding AI
Reports and discussion timeline
Browse reports and discussion by time. Their evidence levels remain distinct.
Jul 3
DeepLearning.AI · Letter from Andrew NgFirst-partySelected
DeepLearning.AI Weekly: GPT-5.6 Family Limited Release, Sakana Model Orchestration, and Microsoft's In-House Reasoning Model
The DeepLearning.AI weekly roundup covers several developments: OpenAI released three closed-source vision-language models, GPT-5.6 Sol, Terra, and Luna, currently available only to about 20 U.S. government-approved organizations, with prices ranging from $5 to $1 per million input tokens; OpenAI says it will expand access in the coming weeks. Sakana AI released Fugu and Fugu-Ultra, two models that orchestrate other models and agents, claiming they surpass Claude Fable 5, Claude Mythos Preview, and GPT-5.5 on benchmarks such as Terminal-Bench 2.1 and GPQA-Diamond. Microsoft launched MAI-Thinking-1, its first reasoning model not distilled or fine-tuned from another vendor, using a MoE architecture with 1 trillion total parameters and 35 billion activated per token, claiming it surpasses Claude Sonnet 4.6 with 97.0% on AIME 2025. The roundup also introduced RoboReward, a robot reward model from teams at Stanford and Berkeley.
Jun 29
timothy_b_lee_understanding_aiSelected
English translation is being prepared.