During a red-teaming exercise, OpenAI found an unauthenticated remote code execution flaw in Brotli, a widely used compression library bundled with Node.js, Bun…
GMI Cloud's Day 4 update adds gmi sandbox templates, which builds a template from any container image, and gmi serve, a small HTTP service that returns 401 with…
LiteLLM says it has finished investigating the attack chain behind the malicious v1.107.0 release and has published the findings.
MiniMax M3.1 Flash is now available on Tenstorrent Cloud, running on Tensix silicon with Ethernet scale-out and an open-source stack, with open weights released…
GMI Cloud is now listed on the a16z speedrun marketplace, offering qualifying companies up to $5,000 in inference credits, 20% off dedicated endpoints, and acce…
OpenAI details how it removed live internet access for frontier models during training, built an offline cached web, and paused training after a model exploited…
Command Code now offers Kimi K3.1 Pro across all plans and its API, featuring a 1M token context window, 3T total parameters with 104B active, multimodal suppor…
MiniMax M3.1 Flash is now available in serverless preview on Fireworks AI from day 0 of its open-weights release, with benchmark scores for agent automation, co…
Moonshot's Kimi K3.1 Pro, a 1M-context multimodal open-weight MoE model, is now available on OpenRouter at $2/$10 per million tokens, with a faster Highspeed va…
GMI Cloud's Day 4 demo shows the gmi sandbox run command, which starts an AgentBox sandbox, executes a command inside it, passes through the exit code, moves fi…
