GMI Cloud details the permission model of its gmi mcp serve command: 36 read tools by default, with --allow-write adding 42 write tools and --allow-destructive…
LiteLLM now officially supports OpenCode Go, letting users route Go subscription models through their existing LiteLLM setup with automatic upstream selection a…
GMI Cloud's gmi mcp serve exposes GMI Cloud as tools for agents; a demo shows Codex running gpt-6.1-sol starting a sandbox, running a command, querying a model…
B.AI has added Moonshot's Kimi K3.1 Pro and K3.1 Pro Highspeed to its platform, both with a 1M token context window and native vision, targeting coding, 3D mode…
Qoder extended the free usage period for DeepSeek-V4.1-Flash-1017 from Oct 24 to Nov 1, 00:00 UTC+8, keeping the model at 0x credits for both free and paid user…
Qoder's Community Boost unlocked three rounds faster than expected, so it is adding 100K total boosts and a bigger mystery pack; Pro-or-above subscribers can cl…
Qoder now supports Kimi K3.1 Pro across its full suite from day one, offering a 1M-token context window aimed at long-running agent work.
GMI Cloud ran Kimi K3.1 Pro for 4 hours 30 minutes at a cost of about $39 to generate an aquarium of Cthulhu mythos creatures, noting the model struggled with f…
OpenAI's Tibo shared Day 19 small updates for Codex: faster call connections, reasoning changes mid-conversation no longer invalidate the cache, plus new CLI fe…
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…
