Command Code Teases Upcoming DeepSeek Integration
Command Code announces an exciting DeepSeek-related update is coming soon, following recent moves like free AI models and a low-cost coding plan.
LiteLLM AutoRouter Gets Four New Updates for Production Traffic
LiteLLM announced four new updates to AutoRouter: custom routing dimensions, automatic router configuration, per-model output limits, and updated presets for th…
LiteLLM Resignation Public Statement: AI Gateway Direction Questioned
A LiteLLM team member announced their resignation, saying both founders are not acting responsibly and that the company is racing toward an AI gateway without a…
OpenCode Improves Git Worktree Management with Pluggable Strategies for AI Agents
OpenCode is making Git worktree management more flexible by letting plugins register different strategies such as rift and lane, plus adding a built-in tool so…
GMI Cloud Signals Partnership with MiniMax AI
GMI Cloud has signaled a partnership with MiniMax AI, following plans to collaborate on video generation using MiniMax's H3 model on GMI Cloud's platform.
GLM 5.3 Flash Doubles Rate Limits on OpenCode Go
OpenCode announced that GLM 5.3 Flash now receives twice the rate limits on OpenCode Go, giving users more generous access to the model for coding workflows.
GMI Cloud and MiniMax Team Up for a Live H3 Video Generation Build
GMI Cloud is hosting a live building session with MiniMax to explore H3, including practical prompt strategies and real-time video generation on GMI Cloud, plus…
OpenAI Shares an AI-Powered 'Defense Factory' Playbook for Cybersecurity
OpenAI says it mobilized 250+ people to harden hundreds of systems, and its latest cyber models helped uncover and fix vulnerabilities that might otherwise have…
Perplexity Invites Retrieval Model Submissions for Q2D-Web Evaluation
Perplexity is accepting submissions of publicly available Hugging Face retrieval models for evaluation, with a technical report available.
Perplexity Evaluates 13 Retrieval Models: pplx-embed-v1-4b Tops Web Ranking and Combined Sets
Perplexity benchmarked 13 retrieval models on three relevance sets using Recall@1000. The pplx-embed-v1-4b model leads Web Ranking and Combined, while Nemotron-…