AI Agents

OpenAI

GPT-Live-1 in the API: Pricing, Full-Duplex Voice, and the Configuration Behind Every Benchmark

GPT-Live-1 is now available to developers at $0.05 per minute, billed per second, with backend models charged separately. The voice model listens while it speaks and delegates real work to a second model, which means every performance number OpenAI published describes a pair, not the voice model alone. Pricing, the delegation architecture, the full benchmark table with its configurations, and what to verify before shipping.

DeepSeek

DeepSeek V4.1 Flash Replaces V4 Pro: Pricing, Benchmarks, and What Changed Since V4 Flash

DeepSeek released V4.1 Flash on September 10, 2026 and is routing its own V4 Pro flagship to it four days later at Flash prices. A 552B backbone that activates 8B parameters to read and 16B to write, native vision, an 890-byte-per-token KV cache, and a scaffold table showing the same model swing 8.7 points on one benchmark depending on the agent harness.

ChatGPT

ChatGPT Operator in 2026: What It Became and When to Use It

ChatGPT Operator moved into agent mode and now lives in ChatGPT Work’s Cloud browser. Which tasks to try, how sign-in and approvals work, and how to stay safe.

Meta Muse: The Personal AI Agent That Books, Buys, and Negotiates for You

Meta’s Muse is a personal AI agent that carries out tasks rather than answering questions, free for most use and distributed inside WhatsApp. What it does, how the per-user virtual machine, Sentinel supervisor, and credential substitution work, what the paid tiers cost, where the privacy promise is policy rather than architecture, and how to try it without handing over your inbox.

Claude

Claude Commerce Agents: Anthropic's Open-Source Blueprint for AI Shopping Assistants and Merchant Agents

Anthropic published free, Apache-licensed reference code for retailer-hosted shopping and merchant agents. What is in the blueprint, how the guardrails keep the model away from money, what retailers still have to build, and who should worry.

Claude Code

Claude Code Subagents: What They Do and When to Use Them

How Claude Code subagents work: the context problem they solve, built-in Explore and Plan agents, writing your own, and a decision framework for delegating.

OpenClaw

OpenClaw 2.0: The Free Personal AI Agent Just Had Its Biggest Release Ever

OpenClaw 2.0 rebuilds the free, open-source personal AI agent: simple installation, a conversation-first app, shared multiplayer sessions — plus a security posture that still depends on the user flipping the switches.

Anthropic

Anthropic's Model Hardware Standard: AI Agents Step Into the Physical World

MHS is Anthropic’s specification for connecting AI agents to physical equipment. Pilots ran a drug-discovery assay, unified a seven-program microscope stack, and pushed laser recovery from 58 to 99.3 percent success.

Salesforce

Claudeforce: Salesforce Just Put Its Entire CRM Inside Claude

Salesforce put its CRM inside Claude: pipeline reviews, meeting prep, and record updates by conversation. What shipped, the economics, the security model, and the debate about what Salesforce is now.

Grok

Grok Bot: What It Costs, What It Can Access, and How to Build an Agent Team With It

Grok Bot hands work off across apps, files and websites instead of answering questions. What it costs and which plans qualify, the access boundary most write-ups miss (every Bot on an account shares one computer), how to run a first bounded task, and the orchestrator pattern behind the 10 to 20 agent setups.