Hugging Face Shares 'Training Agents 4': From Reward Functions to Environments
Hugging Face has shared the next installment of its AI agent training series, focusing on moving from reward functions to environments.
Hugging Face has shared the next installment of its AI agent training series, focusing on moving from reward functions to environments.
OpenRouter's new Shell tool lets any model run commands in a hosted Linux container with file support; try it in the chatroom by enabling the shell tool or send…
OpenRouter's new stateful Shell tool costs $0.0001 per active second, including file usage, with a 30-second minimum for cold container starts.
OpenRouter has launched a new beta tool, openrouter:shell, that lets any model on the platform run commands in a hosted Linux container and move files in and ou…
Requesty argues that enterprise AI policy should be enforced at request time, not stored as documentation. Governance becomes a decision chain: Who → Model → Pr…
SiliconFlow has published DeepSeek-V4.1-Flash pricing, with off-peak input at $0.15/M tokens, output at $0.60/M tokens, and cache reads as low as $0.003/M token…
DeepSeek-V4.1-Flash is live on SiliconFlow on day one, featuring a 552B MoE architecture, native vision, a 1M context window, and an MIT license.
WeKnora, an MIT-licensed knowledge agent platform with 22k stars, released v0.8.0, moving from RAG toward agentic RAG with cross-session memory and default sand…
DeepSeek's new V4.1 Flash model is reported to be back on top of the open-source model leaderboard, offering very low cost while staying highly capable. A compa…
StepFun shares a positive wrap-up of Day 1 at AGNTCon + MCPCon Japan, where it met builders and showcased its work. The team will return for Day 2 at Booth T6.