Mistral Large 4 Can’t Be Cut Off, and Its Guardrails Can’t Be Patched From Paris Either
Mistral released the Large 4 weights while admitting a potential for cyber uplift and promising guardrail updates. Those updates reach the API, not your downloaded copy. What that means for buyers.
Five Things to Check Before Your Team Lets Copilot Fusion Pick the Model
GitHub's Copilot Fusion routes each coding request across one or more models and bills every model it calls. What the docs say about Max costs, the November discount, AI credits and discarded drafts.
Gemini 4 Carbon Is Gated to Cyber Defenders: Seven Moves to Make on Argon Before It Opens
Gemini 4 Carbon tops the Artificial Analysis index at 59, but only Fairwind cyber defenders can call it. Its hallucination rate rose, its latency doubled, and Argon's price doubles December 1.
What Cloudflare’s Cache Leak and Cohere’s Qwen Finding Say About Where to Run Open Models
On October 26, a hosted open model returned other customers' data with HTTP 200, and a self-hosted one tried to reach a Beijing bucket while its trace denied calling tools. Both reported success.
Anthropic’s S-1 Shows Half of Claude’s Revenue Now Runs Through Amazon, Google and Microsoft
Anthropic filed its public S-1 on October 26. For Claude buyers, three disclosures matter: 48% of revenue billed through cloud marketplaces, $265.6 billion in compute commitments, and a hint of...
The Clause That Cost Anthropic Replit Wasn’t the Discount Rate
OpenAI's reported discounts are winning coding workloads from Anthropic. The terms deciding those deals are the overage cliff and share-of-wallet clauses, not the headline rate.
