Beat
Models
At 1,068 tokens per second on NVIDIA GPUs, Inception Labs' Mercury 3 establishes a new performance baseline that renders the fast-cheap tier of frontier models obsolete.
◆ Lena Park
Models
Exclusion and Removal Order #2026-01 reaches third-party products developed using DeepSeek implementations — and the bill for removal lands on the integrators who embedded them, not on DeepSeek.
◆ Lena Park
Models
A US federal exclusion, China's export-registration rule, and four private $20M license gates all landed in the same week. None of them stop at the weights; they follow the model into what you built on it.
◆ Lena Park
Models
An anonymous model topped OpenRouter's rankings for ten days and processed 56.7 trillion tokens before MiniMax attached its name to it. The stealth-to-scale playbook is now the deployment model.
◆ Lena Park
Models
DeepSeek's $0.75 per 1,000 Search & Fetch calls is 6 to 9 times cheaper than every established retrieval provider — and it introduces a billing model that no frontier lab has adopted for inference.
◆ Lena Park
Models
The company that built its reputation on subsidized inference is now selling the search infrastructure agents need to function — at a price point 6 to 9 times below every established provider.
◆ Lena Park
Models
Modal crossed $1B in annualized run rate and raised $750M at a $15.75B valuation — inside the same layer our coverage has documented commoditizing. What the pricing actually captures.
◆ Lena Park
Models
With a $2 trillion valuation target and $518 billion in compute commitments, Anthropic's Q3 financial performance is the critical test for its upcoming IPO.
◆ Lena Park
Models
At $2,599 across major OEMs, NVIDIA is engineering the hardware layer that turns open-weight models from downloadable weights into locally deployable agents – while earning from silicon, not models.
◆ Lena Park
Models
Step-5, Spark Open, Mistral Large 4, and NVIDIA Bolt all released within a day of each other. Only one of them means what you think 'open' means.
◆ Lena Park
