The pulse of AI this week: fresh model upgrades, talent moving between labs, and new capital flowing into generative 3D.
Anthropic’s Claude Opus 4.8 Now Live on AWS Bedrock
Anthropic’s latest flagship model, Claude Opus 4.8, is now available through Amazon Web Services as of June 1, completing its rollout to all major cloud platforms — including Google Cloud Vertex AI and GitHub Copilot. Released on May 28, the model brings a roughly four-times improvement in code honesty (flagging its own bugs far more reliably), a parallel-subagent “dynamic workflow” mode, and a fast mode that runs at 2.5× speed at one-third the previous cost. Pricing stays flat at $5/$25 per million tokens. AWS Weekly Roundup →
Boston Dynamics Loses Its Robotics Research Chief to DeepMind
Scott Kuindersma, Boston Dynamics’ VP of Robotics Research and the engineer most responsible for turning Atlas into a commercial product through reinforcement learning and foundation models, is expected to join Google DeepMind in June. He follows Aaron Saunders, who left for DeepMind’s hardware engineering team last November. The moves underscore DeepMind’s aggressive push to lead in physical AI — the lab already has a $80–90M licensing deal with Contextual AI and Atlas fleet deployments scheduled at Hyundai’s robotics center later this year. TechTimes →
AI 3D Startup Vast Raises $200M at $1B+ Valuation
Beijing-based Vast, which lets users generate 3D assets from text and image prompts, announced a roughly $200 million raise at a valuation exceeding $1 billion. The company reports 20 million global users and sits at the intersection of two fast-moving trends: generative AI and real-time 3D content for games, film, and simulation. The round signals continued investor appetite for specialized AI applications even as the general-model arms race intensifies. LLM Stats →
Gemini 3.5 Pro on Deck for This Month
Google confirmed at I/O 2026 (May 19) that Gemini 3.5 Pro — its frontier multimodal model targeting a 2-million-token context window and Deep Think reasoning — is in limited Vertex preview and on track for general availability in June. Sundar Pichai’s on-stage line, “Give us until next month to get it to you,” sets expectations for a late-June launch. Meanwhile, Gemini 3.5 Flash, which already surpasses last year’s Pro on coding and agent benchmarks, continues to expand enterprise availability. Google Blog →
What to watch: Gemini 3.5 Pro’s general availability date will be the clearest signal yet of whether Google can close the frontier gap with Anthropic and OpenAI — keep an eye on Vertex AI release notes later this month.