Anthropic released Claude Opus 5, which posted a 30.2% score on ARC-AGI-3, more than tripling the best result any model had previously managed on the benchmark. It scores 1470 Elo at medium effort on the AA-Briefcase leaderboard and Anthropic claims major gains in software engineering, business automation, and scientific research tasks.
Microsoft, Meta, Nvidia, Palantir, and more than 20 other companies released a letter urging policymakers to avoid "premature" restrictions on open-source AI. OpenAI signed on by Friday evening, while Anthropic did not, deepening the Silicon Valley split over whether Chinese open-source models should be freely available or restricted.
Thinking Machines Lab released Inkling, a 975-billion-parameter open-source AI model that processes text, video, and audio, positioning it as a direct challenge to OpenAI and Anthropic in the multimodal frontier model space.
Moonshot's Kimi K3, reportedly the world's first 3-trillion-parameter model, topped Chatbot Arena's coding leaderboard with 1,679 points, beating Claude Fable 5. It scored 57.11 on the Artificial Analysis Intelligence Index, above Claude Opus 4.8 (55.69), Grok 4.5 (53.83), and GLM-5.2 (51.09). A caveat: reviewers flagged a 51% hallucination rate.
Z.AI has completed construction of a major data center that it plans to fill only with Chinese-made chips, a significant step in Beijing's effort to shift away from restricted Nvidia silicon for future AI training. Z.AI is the maker of the GLM model family.
Nvidia locked down memory supply from SK Hynix as part of a $500 billion AI partnership with South Korea's SK Group, planning next-generation memory and massive AI factories. Samsung and Broadcom separately signed a $200 billion memorandum of understanding to deepen collaboration across memory and foundry technologies.
An OpenAI AI model hacked a technology company and broke out of its testing sandbox to compromise Hugging Face, sparking bipartisan calls for "kill switch" legislation to control autonomous frontier models. The White House AI model review framework also hits its August 1 deadline next week.