News · 27 Oct 2026

Tuesday roundup: 27 Oct 2026

OpenAI releases GPT-6.1 Terra for computer use at half the API price of GPT-5.6 Terra, the Cursor Agent SDK is generally available, Claude Code 2.1.311 fixes long and unattended sessions, new Gemini API projects store requests by default, and Grok Bot 0.75.8 tightens Team Bot sharing.

  1. OpenAI

    GPT-6.1 Terra

    Released 26 Oct 2026

    OpenAI's new fast, cheaper model is aimed at computer use, visual reasoning and writing. It's live in ChatGPT Work and Codex for Plus, Pro, Enterprise and Edu.

    In the API it's gpt-6.1-terra, with a 1.05M-token context window, reasoning effort from none to max, and tool calling through the Responses API only.

    OpenAI says standard API prices are half those of GPT-5.6 Terra and GPT-6.1 Sol ($1 per million input tokens), and that it scores 50.3% on Terminal-Bench 4.0 at max effort, more than double GPT-5.6 Terra. OpenAI still points people to GPT-6.1 Sol or GPT-6 Astra for the hardest coding work.

    Who it affects Teams running high-volume screen and form automation, such as invoice entry. It's worth re-running the cost figures for any computer-use agent you've put on hold.

    Source: OpenAI announcement ↗

  2. Cursor

    Cursor Agent SDK, generally available

    Released 20 Oct 2026

    You define an agent as a folder (instructions.md, a tools/ folder of typed actions, skills, MCP connections and channels), run it locally with a playground, test it with agent-sdk eval, and deploy it to Cursor or your own infrastructure.

    Agents can wake on GitHub, GitLab and Bitbucket PR, comment and CI events, Slack mentions and DMs, webhooks, or a cron schedule. agent-sdk github replay turns a real PR into test payloads.

    Who it affects Teams building review bots and triagers, which can now be versioned and tested in CI as ordinary code rather than kept as prompts in a dashboard.

    Source: Cursor changelog ↗

  3. Anthropic

    Claude Code 2.1.311

    Released 26 Oct 2026, 21:51 UK

    Standard-context sessions that went past 200K tokens no longer stop with "Usage credits required for 1M context". Claude Code now compacts and carries on.

    Subagents can use MCP tools from servers that finish connecting after they start, and Claude Code no longer retries requests the API has marked as not retryable, which could hang unattended runs.

    The self-hosted runner gets --hold-idle-file and --max-hold-min to keep an idle session alive while your host reports leftover processes.

    Who it affects Anyone running long Claude Code sessions or unattended runners. These are mostly reliability fixes for the cases where a hung turn costs the most.

    Source: Claude Code changelog ↗

  4. Google

    Gemini API stores requests by default for new paid-tier projects

    Released 26 Oct 2026

    New paid-tier projects using the Gemini API's Generate Content endpoints now store requests by default (store=true), matching the Interactions API.

    Existing projects keep their current setting, and you can opt out per request or in the Logs and Datasets settings.

    Who it affects Anyone creating new Gemini API projects for regulated data, who should check the logging default before anything sensitive goes through.

    Source: Gemini API release notes ↗

  5. SpaceAI

    Grok Bot 0.75.8

    Released 24 Oct 2026

    Grok Bot can now offer to add, change or remove your Auto-review rules on a card you confirm, and Auto-review takes a second look before asking for approval.

    Connectors that add new tools or fields work without reconnecting, and Team Bots are now only reachable by people on their share list.

    Who it affects Teams leaving agents running unattended, who should see fewer approval prompts, and anyone sharing Team Bots, which are now limited to their share list.

    Source: Grok Bot changelog ↗

← All news