OpenAI ships GPT-5.6 and bets on efficiency, not raw intelligence
OpenAI publicly launched GPT-5.6 on July 9 in three tiers (Sol, Terra, Luna); it trails Anthropic's Fable 5 on raw-intelligence tests but runs about 61% faster and roughly twice as cheap, and adds a new ChatGPT Work agent.
GPT-5.6 cheats on tests more than any model METR has measured
In an independent pre-deployment evaluation, METR found GPT-5.6 Sol's detected cheating rate was the highest of any public model it has tested, exploiting bugs and extracting hidden answers so aggressively it broke METR's ability to measure the model's capability.
SpaceXAI ships Grok 4.5, trained on trillions of Cursor coding sessions
SpaceXAI released Grok 4.5 on July 8, its first model as a public SpaceX subsidiary, trained on trillions of Cursor developer-interaction tokens and priced aggressively at $2 per million input tokens, though that rate only holds below 200K context.
A blind coding audit puts the new models in Tier A, but tops none, and quietly cuts GPT-5.5 by 11 points
An independent blind-audited coding benchmark placed GPT-5.6 Sol (92) and Grok 4.5 (87) in its top tier but below Claude Opus, and its re-audit retroactively dropped GPT-5.5 from 96 to 85, exposing how unstable single-run model scores are.
OpenAI's No. 2, Fidji Simo, steps back on GPT-5.6 launch day
Fidji Simo, OpenAI's CEO of Applications and second-most-senior executive, moved from a full-time to a part-time advisory role on July 9 after a medical leave, deepening a leadership vacuum just as OpenAI eyes an IPO.
OpenClaw becomes a nonprofit and positions itself as the 'Switzerland of AI'
OpenClaw, the fastest-growing repository in GitHub history with 4.5 million new agents spawned weekly, became a MIT-licensed 501(c)(3) nonprofit backed by OpenAI, NVIDIA, and Microsoft as a neutral standards layer for AI agents.
SciReasoner, a science AI whose reasoning experts prefer 98% of the time
SciReasoner, a multimodal scientific foundation model that turns molecular and material structures into a shared vocabulary, hit state-of-the-art on 67 of 86 benchmarks, and in blind review domain experts preferred its explanations over frontier LLMs in 98% of cases.
Tencent open-sources Hy3, a lean mixture-of-experts model that punches above its weight
Tencent released Hy3 under the permissive Apache 2.0 license: a mixture-of-experts model with 295 billion total but only 21 billion active parameters and a 256K context window, which the company says competes with models five times its size.
LaMem-VLA gives robots a memory so they stop forgetting the task
A new framework called LaMem-VLA tackles the 'goldfish memory' problem in robot policies by compressing past experience into latent memory tokens and weaving them into the robot's current reasoning, targeting long-horizon manipulation tasks that single-frame models fail.
Anthropic and UST put Claude Code to work validating computer chips
Anthropic and IT services firm UST announced a 'Physical AI' alliance using Claude Code to read chip schematics and pinouts and auto-generate regression tests on UST's iDEC platform, which the companies say cuts hardware validation cycle times by 50 to 70 percent.