Mistral's Voxtral 3 Adds End-of-Turn Detection for Voice Apps
Mistral announced that Voxtral 3 includes end-of-turn detection, helping developers determine when a user has finished speaking to build more responsive voice i…
Global AI Digest
-
-
OpenRouter Launches Next-Generation Auto Router and Auto Code
OpenRouter introduced two new routers, Auto Router for general Q&A and agents and Auto Code for coding agents, claiming benchmark parity with Opus 5.5 at 74% lo…
-
Requesty Data: Claude Haiku Token Usage Grew 381% From September to October
Requesty reports Claude Haiku token volume rose 381% month-over-month, with Claude's share of successful chat tokens climbing from 17% to 23%, while Kimi droppe…
-
LiteLLM Adds -latest Model Aliases and Claude/Copilot Fixes
LiteLLM introduced -latest aliases that automatically point to the newest model in a family, plus merged fixes for Claude parameter listing, Claude limit fallba…
-
Ahmad Awais Experiments With a 'Life Agent' Built on Command Code
Ahmad Awais describes a personal 'life agent' built on Command Code that handles ops work, merged about 100 PRs, replaced six apps, and rebuilt his Mac setup in…
-
Hugging Face Now Streams Live RL Training Runs
Hugging Face lets users follow live training runs, starting with LoRA adapters trained via RL on LFM2.5/LFM3 and Twil-9B for agentic tasks.
-
B.AI Weekly Token Ranking: MiniMax Climbs to #2, DeepSeek Holds 89.70% Share
B.AI's weekly token ranking for Oct 30–Nov 5 shows MiniMax rising two spots to second place, Qwen's momentum up 35.91 percentage points, and DeepSeek holding 89…
-
B.AI Weekly Token Ranking: MiniMax Climbs to #2, DeepSeek Holds 89.70% Share
B.AI's weekly Chinese-model token ranking for Oct 30–Nov 5 shows MiniMax rising two spots to second place, Qwen rebounding 35.91 percentage points, and DeepSeek…
-
Command Code Adds /import to Migrate Existing Coding Agent Configs
Command Code introduced an /import command that transfers skills, agents, slash commands, MCP configs, and memory files from an existing coding agent setup.
-
Hy4 Ranks #4 on OpenDesign Design Benchmark, 0.1 Points Behind GPT-6.1 Sol at 41% Lower Cost
Tencent's Hy4 scored 0.1 points behind GPT-6.1 Sol on OpenDesign's real-world design benchmark while costing an estimated 41% less, ranking #4 among 23 tested m…
