The Apache 2.0 framework runs on Nvidia Blackwell GPUs and Huawei Ascend 950 chips. DeepSeek kept some parts private, including cache management and load balancing.
DeepSeek released DeepInfer on 20 October 2026, the inference framework it uses to serve its models, as open source under the Apache 2.0 licence.
Open Source AIElevenLabs launched Eleven Realtime on 20 October 2026, two speech-to-speech models for voice agents.
Tools & AppsOpenAI began rolling out Locked Chats in ChatGPT on 20 October 2026.
Tools & AppsMistral released Voxtral 3 on 20 October 2026, a family of transcription and text-to-speech models with a voice agent SDK.
Models & LaunchesMistral and Nvidia announced on 21 October 2026 a collaboration on AI-native engineering workflows that connect simulation, AI model development and design.
Research
Transcription starts at $0.0035 a minute and covers 50+ languages; speech generation costs $13 per million characters. A…
Customers will be able to buy through Google Cloud Marketplace and count it against existing Google Cloud commitments. No launch…
The Chinese lab says the model is faster and 70% cheaper on agent tasks than its predecessor. Independent tests from Artificial…
The draft sets a threshold of more than 50 deaths or $1 billion in damage, relies on registered independent evaluators, and gives…
Postmedia, Torstar, The Globe and Mail and The Canadian Press allege Anthropic used their content without a licence. The claims…
The order, set out in a Federal Register notice, also bars agencies from buying DeepSeek products in future. DHS cites the risk…
The model reads text, images, documents and video and is built to be post-trained for specific business tasks. Nvidia also…
The model has a 1-million-token context window and is free for many developers under Meta’s new Muse Model License. Meta’s own…
Built with Meta’s PyTorch team, TorchTPU supports eager mode and torch.compile, and Google says training scaled to 8,192 Ironwood…
Meta’s engineers say the New Albany site trained its Muse models, now runs hundreds of thousands of GPUs, and cut training…
The open-source backend supports Trainium2 and Trainium3, works with Hugging Face and vLLM plugins, and starts public beta on 20…
PHP 8.5.12, 8.4.27, 8.3.36 and 8.2.35 fix the “PHP2Shell” bugs. Amazon saw an exploitation attempt four days before disclosure…
NorForge found the campaign through control panels left open online of an open-source agent tool. Deere, SharkNinja and Cummins…
AISLE would become Cohere Cyber, a new security unit. No price was disclosed and the deal has not closed.
Orchestra’s agents track studies, budgets and clinical trial schedules for pharma teams. Terms were not disclosed, and OpenAI…
Caijing reports investors including Tencent, CATL and SMIC are backing a round that values the Chinese AI lab at about 500…
The October report says most districts saw little net change in jobs. Minneapolis contacts said AI was not causing layoffs but…
Max and Mini models are available in ElevenAgents and an OpenAI-compatible Realtime API. ElevenLabs reports 98.8% on Big Bench…
Locked Chats hide sensitive conversations from anyone using your device. They do not change data retention or training settings…
AI Scan checks pull requests in languages GitHub’s CodeQL does not cover, such as PHP, Bash and Terraform. Agentic Autofix drafts…
Up to 10,000 more AI compute units can help Indian builders. Clear prices, simple sign-up and public usage data would make the…
OpenAI’s COBOL results are encouraging for banks and governments running old systems. Publishing MultiLangBench was a good step…
Risk scores, read-only modes and human approval steps make agents easier to trust. We think this kind of control should come…
The partners will link Mistral’s models with Nvidia’s Omniverse, CUDA-X and Warp tools, and publish datasets and tests for AI…
C-DAC’s tender asks AI service providers for GPUs, TPUs and other accelerators in clusters of at least 512 units in Indian data…
New controls in Microsoft 365 E7 and Project Perception can remove an agent’s write access, hold actions for human approval or…