Trending now

• OpenAI dots: Always-on ChatGPT agents, each with its own cloud computer and browser, that keep working on a goal between conversations.
• Meta Muse: Meta's consumer personal agent: a Muse Spark model running in your own cloud VM, with a separate guard agent that must approve anything leaving the machine.
• Claude Tag: A shared Claude agent that lives in Slack channels, works under its own service accounts, and turns a tagged request into a finished result in the thread.
• Grok Bot: Persistent named agents that share one cloud computer with a browser, terminal and file system, sign in to your apps, and keep working after you close the laptop.
• Jev: A hosted model that answers typed questions about your data with calibrated probabilities instead of generating text.
• Decision models and OpenAI's Decisions API: A new class of model that cannot write text: you give it state and a set of typed questions, and it returns probabilities over the answers you defined.
• Claude Opus 5.5 and Sonnet 5.5: Anthropic's new mid-cycle Claude models: Opus 5.5 at a lower price than Opus 5, and Sonnet 5.5 at the old Sonnet price, both with thinking you steer rather than switch off.
• GPT-6 Astra: OpenAI's top GPT-6 model, a reasoning model built for long agentic work such as computer use, coding and research, and the first OpenAI model rated Critical for cybersecurity capability.
• GPT-6.1 Sol, GPT-6 Sol and GPT-6 Luna: OpenAI's cheaper tiers below GPT-6 Astra: Sol for coding and professional agent work, Luna for fast high-volume tasks, and a GPT-6.1 Sol refresh one week later.
• Gemini 3.8 Flash and the Gemini managed agents harness: Google's mid-priced Flash model retuned for long-running coding and agent work, now the default brain of the Antigravity managed agent in the Gemini API.
• OpenAI Agents API: OpenAI's own Codex agent harness offered as a managed API: you send a task to a durable session, and OpenAI runs the loop, the context management and the recovery.
• Cloud sessions for coding agents: Coding agents that run on the vendor's virtual machines instead of your laptop, so a task keeps going after you close the lid and you can check on it from a phone.
• MCP specification 2026-07-28 (stateless MCP): The largest revision of the Model Context Protocol so far: no handshake, no sessions, every request self-describing, so an MCP server can run as an ordinary stateless HTTP service.
• Policy-based agent approvals: Agent tools are replacing the click-yes-on-every-command prompt with written deny, ask and allow rules, often with a second model deciding the cases in between.
• Qwen3.8 open weights: Alibaba's Qwen team published downloadable weights for its Max-class 2.4 trillion parameter model and for a 27B dense model that runs on one workstation.
• DeepSeek V4.1-Flash: An MIT-licensed, 552B-parameter mixture-of-experts model that reads text and images, takes a one million token context, and runs on DeepSeek's API for cents per million tokens.
• Muse Glimmer 30B: A 30B dense open-weight model from Meta, under Apache 2.0, built to run agent loops on one consumer GPU or a high-memory Mac.
• Sign in with ChatGPT: An OpenID Connect sign-in that can also let a ChatGPT subscriber spend their plan's included usage inside a third-party app instead of an API key.
• Gated frontier releases for cyber capability: The three largest labs now ship their strongest models in two versions: one with cyber safeguards for everyone, and one with those safeguards relaxed for vetted defenders only.
• TRACE runtime attestation specification: An open specification for a signed record that says which model an AI agent ran, where, under which policy, on what class of data and with which tool calls.