sovereign AI

Mistral Large 4 beats Kimi K3 on Harvey's legal agent benchmark and trails it on the Vals Index. It is a discounted preview API for now, and the licence for its end-of-October weights is unpublished.

Reflection's Beam is a 501B Apache 2.0 open model from a US lab, and its own table has it losing to Kimi K3 on 13 of 14 shared benchmarks. Weights and pricing aren't out yet.

Cohere North 2 gives agents memory that persists across sessions, but Cohere has published no scope, retention or deletion terms for it. The token caps in North Admin are usable now; memory should wait for written answers.

Self-hosted IBM Bob swaps SaaS Bob's Claude, Mistral and Granite routing for one local model, Nemotron or Laguna. IBM has published no evals, GPU sizing or price for it.

Mistral's €3 billion Series D moved its lead investor to Samsung Electronics without changing one customer contract. The same announcement publishes a four-part sovereignty definition that maps to clauses you can actually test — and the European Commission's own reference rubric weights ownership at 15%.

Mistral's EU endpoint reached GA at a 10% upcharge, but its own docs say Agents, Batch and the Files API don't run there — and the 99.9% SLA is still in public preview. Price a multi-year European Compute Unit commitment against what ships today.

Seven routes off OpenAI, priced and deployment-matched on 12 August 2026. Residency is already solved almost everywhere — only Mistral, Cohere private deployment and self-hosted open weights clear a jurisdiction rule.

On July 1, 2026, Palantir CEO Alex Karp told The Information that multiple U.S. government customers — including agencies supporting critical national infrastructure — had switched from proprietary AI models developed by companies like Anthropic to Nvidia's open-source Nemotron alternatives. The agencies that handle the most sensitive data on Earth just decided that open-source AI is good enough. If the national security enterprise can trust open models in air-gapped classified environments, the question every CIO needs to answer isn't whether open-source AI is ready. It's why you're still paying 18x more for proprietary APIs when you don't have to.

DeepSeek V4 ships a 1.6T-parameter open-weight model with 1M-token context at $1.74/M tokens and Huawei chip optimization. What enterprise AI leaders do next.