InfraHugging Face switches its own router for OpenRouter's for open-weight modelsHugging Face announced on 16 November that Inference Providers moves from its own router to OpenRouter's. A failed request is now retried on another provider, and your own provider API key no longer works with the language models.InfraAMD completes its World Labs acquisition, Fei-Fei Li joins as chief scientistAMD said on 16 November 2026 that its acquisition of World Labs is complete. Fei-Fei Li becomes executive vice president and chief scientist, reporting directly to Lisa Su.InfraNVIDIA swept every MLPerf Training v6.1 test, and Vera Rubin debuts with leading numbersNVIDIA said on 16 November it swept every MLPerf Training v6.1 benchmark and made its first MLPerf Endpoints submission. According to NVIDIA, Vera Rubin NVL72 delivers up to 2x more performance per GPU than GB300 NVL72, and the Endpoints results are still pending MLCommons verification.InfraAWS launched Emulator: test your application against AWS locally, with no deploymentAWS announced AWS Emulator on 11 November: a local emulator you run on your own machine. The AWS CLI, the SDKs and your coding agents speak to it over a local HTTPS endpoint using the same APIs as in the cloud. According to AWS, there is no additional service fee.InfraAnthropic shows FlashProve: cryptographic proof that an answer came from the model you were promised, open sourceAnthropic published a prototype that proves cryptographically that a response was produced by the model you were promised. By the company's figures the cost is about 70-95 times the GPU time of running the model itself, and the code is under Apache 2.0.InfraNVIDIA compresses DeepSeek's V4.1 Flash model for Blackwell: about 400 GiB instead of 475NVIDIA published an NVFP4 version of DeepSeek-V4.1-Flash-1017: about 400 GiB instead of the source's roughly 475. The card lists only Blackwell and needs a vLLM build with a patch that stock images lack.InfraCloudflare Radar logged 49 internet disruptions in a quarter: 18 were exam shutdowns in Iraq, 14 were power outagesAccording to Cloudflare, 49 disruptions were recorded from July to September; 19 were government-directed, 18 of them exam shutdowns in Iraq. Power outages account for 14, natural disasters for five.InfraApple put M6 in the MacBook Pro and iMac: by Apple's own numbers, up to 25% faster AI than M5 on the laptop and up to 5x faster than M4 on the iMacOn 27 October Apple announced a 14-inch MacBook Pro and a new iMac with the M6 chip and presents them as platforms for AI on the machine itself. By Apple's figures, AI workloads on the laptop are up to 25% faster than M5, and the iMac is up to 5x faster than M4. Orders open 28 October, stores from 6 November.
48 stories