Gemini Personal Intelligence Comes to Android Auto and Google Built-in Cars
Gemini's Personal Intelligence is now available on Android Auto and coming soon to cars with Google built-in, letting it securely connect your physical journey…
Global AI Digest
-
-
OpenAI Frontier Models and Codex Coming to Google Cloud Model Garden
OpenAI announced that its frontier models and Codex agentic workflows will be available in Google Cloud's Model Garden, letting enterprises use OpenAI technolog…
-
MiniMax M3.1 Flash Integrates with SGLang Omni for Interruptible Voice Agents
MiniMax M3.1 Flash has been integrated into SGLang Omni's cascaded voice interaction system, replacing the locally hosted LLM while adding streaming, conversati…
-
Anthropic Opens Research Preview of Claude Duet, a Proactive Agent App
Anthropic launched a waitlisted research preview of Claude Duet, a proactive agent app that surfaces tasks needing attention, works while users are away, and as…
-
Perplexity Introduces JOLT: Self-Distillation Post-Training That Matches GRPO With 6.5x Fewer Tokens
Perplexity researchers developed JOLT, a post-training method where a single model acts as both teacher and student. It matched GRPO's MATH accuracy with 6.5x f…
-
Command Code Says Its Coding Agent Has Scaled Past 950T Tokens
Ahmad Awais said at Cloudflare Connect in San Francisco that Command Code, a coding agent built for open models, has processed more than 950 trillion tokens.
-
TypeSafe AI Brings Jev Models to UiPath Workflows via New Connector
TypeSafe AI launched a UiPath Integration Service connector that lets Jev models make judgment calls inside automation workflows, claiming roughly an order-of-m…
-
Command Code CLI Adds /btw for Mid-Session Side Questions
Command Code added a /btw command to its CLI that lets users ask side questions mid-session and get instant overlay answers without interrupting the main task.
-
Flint Small MoE Model Hits 54.9% on Terminal-Bench 2.1, Runs Locally in Cline
Cline highlights Flint, a ~2B active-parameter MoE model that scores 54.9% on Terminal-Bench 2.1, needs about 24GB of memory, and runs at 125+ tok/s on a MacBoo…
-
MiniMax M3.1 Flash Lands on Runware with 1M Context and 524K Output
MiniMax M3.1 Flash is now available on Runware, supporting text, image and video input with a 1M context window, up to 524K output tokens, adjustable reasoning…
