Stanford
The first release of Terminal-Bench-Science contains 70 tasks drawn from work researchers actually performed. In the first results, the best agent completes 30 percent of them.
KubeCon
KubeCon + CloudNativeCon, OpenInfra Summit and PyTorch Conference China have merged into a single schedule. Attendees include people from Alibaba, ByteDance, Huawei, Tencent, NVIDIA, Google, Microsoft and Meta.
AI Infra Summit
Organisers expect 8,000 attendees, more than 400 speakers and 250 partners. Six tracks run across the programme, from the AI data centre through to physical AI.
CoreWeave
Three days, three tracks and more than 2,000 attendees expected. The programme is aimed at AI systems running in production, and early-bird pricing ends on 29 August.
Vercel
Anthropic runs the agent loop server-side; Vercel supplies the chat interface. Moving the same handler from Slack to WhatsApp takes a few line changes, and there is no separate database to run.
Tavus
The conventional system cleans the audio, isolates one speaker and throws away the rest. Sparrow-2 does the opposite, and the company says it measured a 2.1 percent conversation failure rate.
Anthropic
The Model Hardware Standard research preview covers programmable devices from microscopes to robotic arms. The company says integration drops from weeks to hours, and the standard is model-agnostic.
Meta
Each call searches the web and refines the output over iterative passes, with extra evaluation of the prompt for precise elements like charts and QR codes. Generation and editing sit in one model.
Wan
The model produces 30-second 1080p clips with audio in a single pass. Its distinguishing feature is Omni-Reference: documents, spreadsheets and slide decks count as inputs alongside text and images.
Apodex
The flagship targets long-horizon professional work; the 35-billion-parameter Mini is released under Apache 2.0. FrontierAgent, the agent workbench, is open source too.
AMD
The release adds a command-line tool, an AMD Skills pack that plugs into coding agents, and the open-source Hyperloom. AMD reports a 3.3x inference speedup over ROCm 7.0.
Cartesia
The company says it stopped fine-tuning the existing approach and rebuilt from the architecture up. The model leads both the Provider Voice and Controlled Voice arenas, but is in beta on Cartesia's API only.