Claude Opus 5.5 makes agent cost the main question
Anthropic's Claude Opus 5.5 lowers token and cache-read prices, shifting the buyer test toward agent-task cost rather than benchmark rank alone.
AI Agents
Gemini 3.8 Live keeps voice agents working while you talk
Google says Gemini 3.8 Live and 3.8 Live Extended Thinking make real-time voice agents more useful by combining visual grounding, background tool calls and deeper reasoning.
OpenAI’s math agents have a proof to answer for
OpenAI has released a Lean-checked Navier–Stokes proof from a 10,000-agent system, bringing independent mathematical review and questions of credit into focus.
Astra's ARC score depends on the harness
ARC Prize's Astra evaluation shows why frontier-agent scores need their harness, memory condition, action count and cost beside the headline number.
Gemini 3.8 Flash holds token prices as Cyber access stays limited
Google launched Gemini 3.8 Flash at the same introductory token prices as 3.7 Flash, while routing Gemini 3.8 Flash Cyber through a restricted defender program.
Claude Fable 5.1 ties coding gains to data custody
Anthropic's Claude Fable 5.1 launch links long-running coding claims, cheaper cache reads and customer-controlled monitoring data.
Gemini gives video analysis an agentic mode
Google has added agentic video understanding to Gemini, exposing targeted processing steps for developers handling longer video workloads.