GPT-6.1 Terra Replaces GPT-5.6 Terra — Same Workhorse Band
October 29, 2026
GPT-6.1 Terra is in the model picker, in the seat GPT-5.6 Terra held. Use it when the work is visual — matching a product photo to a catalog, reading a drawing, or acting on an application screen — or when you want clear writing without stepping up to Sol. Same 1.05M-token context, 128K max output, tools, Swarm, and complexity-aware reasoning, on the same Workhorse 3× band. Open a GPT-5.6 Terra conversation and the next turn runs on GPT-6.1 Terra. You do not re-pin.
What you can do
- Pick GPT-6.1 Terra under Closed SOTA. It reads text and images, calls tools, and can lead or work inside a Swarm.
- Start from the new-chat card. It says GPT-6.1 Terra is live. Try it today. and opens the picker on that row.
- Match the effort to the job. Fast and Medium map to low and medium reasoning, the same way they do on GPT-6.1 Sol and GPT-6 Luna. Medium is the default. A pinned Beast asks for max effort and adds pro reasoning for the hard pass.
- Trust the visual pass more than GPT-5.6 Terra. OpenAI reports GPT-6.1 Terra ahead of GPT-5.6 Terra on visual reasoning at every effort setting, and close to GPT-6 Astra and GPT-6.1 Sol on that work at a lower token band.
- Get fewer factual slips on a quick pass. On OpenAI's error-inducing prompts, the error rate at low effort fell from 53.6% on GPT-5.6 Terra to 16.7% on GPT-6.1 Terra. At maximum effort it fell to 4.3%.
- Keep working in an existing GPT-5.6 Terra thread. The picker shows GPT-6.1 Terra and the next message runs on it. Saved model preferences roll forward the same way.
Where this shows up
You are checking supplier photos against the catalog before a buyer meeting. Pin GPT-6.1 Terra, attach the shots, and ask which SKUs match and which ones do not.
You have a marked-up floor plan and need a client note that says what changed. The same thread holds the drawing and the write-up.
You already have a GPT-5.6 Terra review open from last week. Open it and keep going — the picker says GPT-6.1 Terra, and the next reply runs on the new model at the same Workhorse 3×.
Try it
- Pick GPT-6.1 Terra, attach the photos, then: "Match each product photo to a catalog row. Flag any photo that does not match a live SKU, and say what is different."
- Pick GPT-6.1 Terra in Medium, then: "Read this marked-up floor plan and write the client note: what changed, what it costs us, and what I should ask them to confirm."
- Pick GPT-6.1 Terra in Fast, then: "From these application screenshots, list the steps a new hire follows to issue a refund, in the order the screens show."
Heads up
- Same band — Workhorse 3×. GPT-6.1 Terra sits where GPT-5.6 Terra did. Your plan quota and token balance do not change.
- GPT-5.6 Terra is retired from the picker, Auto, and Swarm. Sessions and saved preferences pinned to it roll forward to GPT-6.1 Terra. Stored pins update when you open the chat or send the next message. GPT-5.4 pins, already rolling forward to GPT-5.6 Terra, now land on GPT-6.1 Terra too.
- Auto on Pro Max and Black can select it for workhorse turns. Pro's Auto stays inside its curated pool, so pin GPT-6.1 Terra there when you want it every turn. GPT-6.1 Sol, GPT-6 Luna, and GPT-6 Astra stay on their current bands. Argus heartbeats still default to GPT-6 Luna.
- For the hardest coding, stay on Sol or Astra. OpenAI still points complex software-engineering work at GPT-6.1 Sol and GPT-6 Astra. Terra's gain is largest on visual work, computer use, and everyday writing.
- No harness numbers yet. Can it Run Alfrada? does not show GPT-6.1 Terra until it has a full run.