GPT-6.1 Terra
OpenAI's Terra update combines visual work and everyday writing with lower standard token prices than GPT-5.6 Terra. It also improves the family's reported coding results.
GPT-6.1 Terra is available on MiniRouter as . The model page has current pricing, provider routes and API setup.
A practical model for visual work
Terra's focus is everyday work involving images, applications and prose. OpenAI still recommends Sol or Astra for its hardest coding tasks.
- Context
- 1.05M tokens
- 128K maximum output, according to OpenAI.
- Publisher modalities
- Text and image input
- Text output. MiniRouter route capabilities may be narrower.
- Reasoning
- none → max
- Medium is the publisher default.
- Tools
- Responses API
- OpenAI requires Responses for tool calling.
A step forward for Terra's coding
Terminal-Bench 4.0
Share of command-line tasks solved in OpenAI's reported evaluation.
- 6.1 Terra50.3%
- 5.6 Terra21.5%
0–100% · higher is better
View chart data
| Model | Reasoning effort | Tasks solved |
|---|---|---|
| GPT-6.1 Terra | max | 50.3% |
| GPT-5.6 Terra | Not specified in announcement text | 21.5% |
Use this as a reason to test a familiar repository task. Give both models the same tools and acceptance checks, then compare the resulting change, cost and time.
Build around a result you can verify
The publisher highlights image and screen-based tasks alongside writing. On MiniRouter, use text input; image input is not supported. Start with a proposal revision or fields extracted from document text. Your application supplies any execution environment and decides which actions may run.
Give it the source
Provide the image, draft or screen state, together with a specific goal and clear constraints.
Inspect the proposed result
Check extracted values or the planned action before your application commits a change.
Verify the completed work
Compare the final document or application state with the original requirements.
Lower standard token prices
OpenAI lists $1 input and $5 output per million tokens, half the standard rates of GPT-5.6 Terra. Fast mode doubles those rates. Above 272K input tokens, the full request uses 2x input/cache rates and 1.5x output rates.
Publisher pricing and options · Current MiniRouter rates
Choose the lowest effort that reliably completes your task. Record the full request cost, including reasoning, instead of comparing output length alone.
Try a bounded writing task
Create a MiniRouter API key and set MINIROUTER_API_KEY. Check that the rewritten update preserves all dates and amounts.
curl https://api.minirouter.sh/v1/responses \
-H "Authorization: Bearer $MINIROUTER_API_KEY" \
-H 'Content-Type: application/json' \
--data '{
"model": "openai/gpt-6.1-terra",
"reasoning": { "effort": "low" },
"max_output_tokens": 2048,
"input": "Rewrite this project update for a client. Keep every date and amount unchanged: The draft is due on November 3. Design costs $450 and implementation costs $900."
}'Keep exploring
Compare published results in the model rankings and coding benchmarks, or read more model releases.
For MiniRouter product updates, visit the changelog.
Sources
- Read at OpenAI ↗
OpenAI GPT-6.1 Terra announcement
Published Oct 26, 2026
- Read at OpenAI ↗
GPT-6.1 Terra model documentation
Published Oct 26, 2026
Editorial reference revised 2026-10-26. Catalog status comes from the generated routing snapshot.