Tuesday, July 21, 20261 min read

Claude Opus 4, Mistral's new edge chip, and EU AI Act enforcement

AIModelsRegulationEdge ComputingPolicy
Welcome to yesterday's edition. Anthropic ships a frontier model, Mistral pushes inference to the edge, and regulators start issuing fines under the EU AI Act.

In This Issue

www.anthropic.com#1

Anthropic Ships Claude Opus 4

Claude Opus 4 sets new human-benchmark highs on SWE-bench and graduate-level science questions. The model is available via API today alongside a free tier with rate limits. A description of Anthropic's Claude Opus 4 model and its benchmark performance. ---

Read Article
mistral.ai#2

Mistral Launches Edge Inference Chip

Mistral co-designed a RISC-V NPU with STMicroelectronics that runs quantised Mistral Small on-device. Power draw is under 5 watts, opening up battery-powered AI chatbots. A description of Mistral's new edge inference chip collaboration with STMicroelectronics. ---

Read Article
european-ai-office.europa.eu#3

EU AI Act Fines Issued for First Time

The European AI Office issued its first precautionary suspension order against a biometric-vetting vendor. Max penalties reach 7 percent of global turnover, drawing immediate industry pushback. A description of the EU AI Office issuing the first fines under the EU AI Act.

Read Article