A busy Tuesday in AI saw Anthropic push a new frontier model into general use, OpenAI open a fresh channel for studying AI’s economic impact, and the global funding race intensify in both the West and China. Here are the developments worth knowing.
Today’s top stories
Anthropic ships Claude Fable 5, its new frontier model
Anthropic announced Claude Fable 5, now its leading model, with the company highlighting gains in software engineering, knowledge work, and vision that grow more pronounced on longer, multi-step tasks. In high-risk domains such as cybersecurity, biology, and chemistry, the model is designed to withhold responses and fall back to an earlier, more constrained model. Anthropic listed pricing at $10 per million input tokens and $50 per million output tokens. The release was paired with a more capable, restricted-access variant offered only to vetted research partners. Read Anthropic’s announcement.
OpenAI launches an Economic Research Exchange
OpenAI introduced the OpenAI Economic Research Exchange, a platform inviting selected outside researchers to run structured, privacy-protected studies on how AI affects workers, firms, and the broader economy. The move signals growing pressure on AI labs to build credible, independent evidence about real-world economic effects rather than relying on internal claims. See OpenAI’s announcement.
DeepSeek reportedly nears a $7.4 billion raise
Chinese AI lab DeepSeek is reported to be closing in on a roughly $7.4 billion funding round, which would rank among the largest startup raises in China to date and underscore Beijing’s continued backing of open-source frontier models. The reported deal lands as Western labs pursue their own mega-rounds and public listings, keeping global compute and capital firmly in focus. More on the reported round.
NVIDIA deepens its physical-AI and robotics push
NVIDIA continued to expand its robotics and “physical AI” footprint, releasing new models and partnering across the hardware stack as global manufacturers unveil next-generation robots built on its platforms. The effort spans simulation tooling, memory partnerships, and on-device chips aimed at closing the persistent gap between training in simulation and deploying in the real world. NVIDIA’s newsroom has the details.
What to watch: with frontier models, agent infrastructure, and physical AI all advancing at once, expect the next round of competition to hinge as much on safety guardrails and compute access as on raw benchmark scores.