GPT-6.1 Terra ships, Liquid AI opens the 2B-class LFM3 family, and Copilot gets server-to-server auth

GPT-6.1 Terra is OpenAI's mid-tier model for computer use (understanding what's on screen and acting across apps) and everyday writing. OpenAI says that at medium reasoning effort it scores 15 points higher than GPT-5.6 Terra on Agents' Last Exam at about a third of the costVendor claim, and that it solves 50.3% of Terminal-Bench 4.0 at max effortVendor claim. The API name is gpt-6.1-terra, with a 1.05M-token context window and up to 128K output tokens.

Standard pricing is $1 per million input tokens and $5 per million output, half of SolVendor claim, which makes high-volume screen automation like reading invoices or filling forms much cheaper to run. OpenAI reports its internal error rate at low effort dropped from 53.6% to 16.7%Vendor claim, so it's worth testing a cheap low-effort setup first and raising effort only where needed. OpenAI still points to Sol or Astra for the hardest coding work.

LFM3 puts six models on one LFM3-1.7B-Base backbone: text (LFM3-1.7B), vision (LFM3-VL-2B), audio (LFM3-Audio-2B), an omni model that takes text, images, video and audio (LFM3-Omni-2B), plus d1-2B and d1-omni-2B for fast structured decisions. It was pretrained on 22T tokens with a 256K context window, and the text model is about 1GB at 4-bit, small enough for phones or a Raspberry Pi 5. You can load the models straight from Transformers.

pip install -U transformers

Liquid AI says the text model scores 38.42 on WebArena-Verified, far above LFM2.5-2.6B's 3.02Vendor claim, and that the omni model beats larger models such as Qwen2.5-Omni-7B on some benchmarksVendor claim. If your app needs offline tool calling or structured output on the device, it belongs on the shortlist, and every model can be fine-tuned on a single GPU.

Services and agents built on the Copilot SDK can now call Copilot with the permissions of a GitHub App installed on an organization, without a signed-in user's credentials. Give the app the read-only Copilot Requests organization permission, install it on the org, and pass a short-lived installation access token to the Copilot SDK.

Until now, internal bots and batch agents usually borrowed one employee's account, and broke whenever that person left or lost access. Usage is now billed to the organization, and its model policies, budgets and spending limits apply. Admins can track per-app usage through the app_client_id field in AI usage CSV reports and cut access by suspending or uninstalling the app.