Technology

Last Week in AI: OpenAI Rolls Out GPT-5.6, Grok 4.5 Launches, Meta Releases Muse Spark 1.1

OpenAI quietly launches GPT-5.6 and rebrands its coding product, while SpaceX AI's Grok 4.5 and Meta's Muse Spark 1.1 enter the competitive AI landscape.

The past week saw significant AI releases from OpenAI, SpaceX AI, and Meta, including new models and products. Infrastructure, policy, and safety developments are also shaping the future of AI, from data center energy demands to international coordination on AI alignment.

In a week of rapid AI developments, OpenAI publicly rolled out GPT-5.6, including its Sol and Luna variants, and rebranded its desktop agentic coding product as ChatGPT Work. The release comes amid disputed claims about whether the US government effectively green-lit and delayed the launch, as well as ongoing concerns about inconsistent frontier-model oversight and jailbreakability.

New Models in the Arena

New model releases intensified pricing and capability competition. SpaceX AI's Grok 4.5 launched as a very low-cost, Opus-class coding model, albeit with minimal safety documentation. Meta, in turn, released Muse Spark 1.1 with aggressive pricing, large coding and cyber benchmark gains, and a lengthy safety evaluation. Meta also previewed Muse Video and rolled out Muse Image before quickly backtracking after backlash over easy generation of images of public Instagram accounts.

Infrastructure, Policy, and Safety

Infrastructure, policy, and safety developments were also in the spotlight. Meta is exploring selling AI compute as a cloud business, a move that would put it in direct competition with cloud providers. US energy regulators are pressing grid operators on large-load data-center connections, signaling a growing focus on the enormous energy demands of AI.

In safety and interpretability research, Anthropic published a "global workspace" interpretability method for verbalizable internal representations. This method aims to provide a better understanding of how language models process and represent information. Meanwhile, reports emerged that China may restrict overseas access to its top models, and the 'AI 2040' project is proposing US-China coordination to slow progress until alignment improves.

A Note on the News Cycle

This news comes as Andrey from Last Week in AI returns from a silent pause. In his note to subscribers, he apologized for the gap in posting and has paused paid subscriptions until he can get back to consistent posting, underscoring the challenge of keeping pace with the breakneck speed of AI news.