GPT-6.1 Terra is OpenAI's new model, released on 26 October 2026. It is the mid tier of the GPT-6.1 line: half the price of the flagship GPT-6.1 Sol, clearly smarter than the previous GPT-5.6 Terra and, according to OpenAI, more than three times faster than Sol on real tasks. The model is already live on Kumo — 15% below OpenAI's list price. Here is what it can do, what the measurements show and how to connect it in a couple of minutes.
Context
1.05M tokens
up to 128K tokens in one answer
Price on Kumo
$0.9 / $4.3
input / output per 1M tokens, 15% below OpenAI
Input
text, images, files
answers in text
Released
26 Oct 2026
on Kumo since 27 October
What GPT-6.1 Terra is
OpenAI splits its line-up into classes: Sol is the strongest, Terra is the mid tier for everyday work, Luna is the lightest and cheapest. GPT-6.1 Terra is the first Terra-class model since GPT-5.6 and the first built on the GPT-6.1 generation.
OpenAI focuses it on three things: computer use (the model understands what is on screen and takes actions across apps), visual reasoning over screenshots and documents, and more natural writing. Reasoning depth is adjustable from low to max: at low effort the model answers fast and cheap, at high effort it catches up with Sol on agent tasks.
In ChatGPT and Codex the model is rolling out to paid plans, and developers can call it over the API as gpt-6.1-terra. Codex CLI needs version 0.159.0 or newer.
GPT-6.1 Terra vs Sol and GPT-5.6 Terra
| Spec | GPT-5.6 Terra | GPT-6.1 Terra | GPT-6.1 Sol |
|---|---|---|---|
| Class | mid tier | mid tier | flagship |
| Context | 1.05M tokens | 1.05M tokens | 1.05M tokens |
| OpenAI price, input / output | $2 / $12 | $1 / $5 | $2 / $10 |
| Cache read, OpenAI | $0.2 | $0.05 | $0.1 |
| Kumo price, input / output | $1 / $6.1 | $0.9 / $4.3 | $1.1 / $5.5 |
What the measurements show
The headline number is Terminal-Bench 4.0, a test of agent work in a terminal: at maximum reasoning effort GPT-6.1 Terra solves 50.3% of tasks against 21.5% for GPT-5.6 Terra (OpenAI figures). On OpenAI's internal error-prompt test the error rate fell from 53.6% to 16.7% at low effort and to 4.3% at max effort — close to GPT-6 Astra and GPT-6.1 Sol.
Independent Artificial Analysis confirms the picture. On its Coding Agent Index the model scores 48 — 6 below Sol (54) and 17 above GPT-6 Luna. On the overall Intelligence Index, Terra at max effort gains 6 points over GPT-5.6 Terra and trails Sol by 4, while a task costs $0.40 instead of $1.40. The hallucination rate on AA-Omniscience dropped from 88% to 38%.
What GPT-6.1 Terra costs
| Where | Input | Output | Cache read |
|---|---|---|---|
| OpenAI | $1 | $5 | $0.05 |
| Kumo, paid from balance | $0.9 ($0.85) | $4.3 ($4.25) | $0.04 ($0.043) |
On Kumo GPT-6.1 Terra is 15% cheaper than at OpenAI itself, and in token packages it is charged with a ×1 multiplier — less than GPT-6.1 Sol (×1.1). Re-reading the same context — a repository, a conversation, a document — costs a twentieth of regular input, so in agent work the final bill is noticeably below the list price.
GPT-6.1 Terra in the Kumo catalogue
The model card with its current price is below; it leads straight to setup.
What it is good for
- Everyday development: code edits, reviews, tests and terminal work — where Sol is overkill and speed matters.
- Agents with tools: multi-step chains where the model calls functions itself and checks the result.
- Computer use and screenshots: the model understands interfaces and documents with images.
- Writing: emails, articles, customer replies — OpenAI specifically improved how natural the writing reads.
- High request volume: at half the price of Sol, Terra pays off for bulk processing and chatbots.
What Kumo has verified
Before launch Kumo runs probe requests against every model. For GPT-6.1 Terra, plain answers, streaming and tool calling are confirmed over both protocols — Chat Completions and Responses — along with image input and reasoning with adjustable effort. The same scenarios then passed as live requests through the api.kumorouter.com gateway, so the model also works in Responses API clients such as Codex CLI.
How to connect GPT-6.1 Terra
- Sign up in the Kumo console and top up your balance.
- Create an API key and set a spending limit on it.
- Point any OpenAI-compatible client or SDK at the gateway https://api.kumorouter.com/v1.
- Set the model to gpt-6.1-terra and send your first request.
curl -s https://api.kumorouter.com/v1/chat/completions \
-H "Authorization: Bearer $KUMO_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "gpt-6.1-terra",
"messages": [
{"role": "user", "content": "Find the bug in this function and suggest a fix: ..."}
],
"max_tokens": 4096
}'
Python example
The official OpenAI SDK works as is — just change the base URL and the key.
from openai import OpenAI
client = OpenAI(
base_url="https://api.kumorouter.com/v1",
api_key="KUMO_API_KEY",
)
stream = client.chat.completions.create(
model="gpt-6.1-terra",
messages=[{"role": "user", "content": "Write tests for this module: ..."}],
stream=True,
)
for chunk in stream:
if chunk.choices:
print(chunk.choices[0].delta.content or "", end="")
FAQ
- How does GPT-6.1 Terra differ from GPT-6.1 Sol?
- Sol is the flagship and somewhat stronger on hard tasks: 54 against 48 on the Artificial Analysis Coding Agent Index. Terra is half the price at OpenAI and, per OpenAI, more than three times faster, so it pays off for everyday work and high volume.
- Why is Kumo cheaper than OpenAI?
- That is Kumo's price for GPT-6.1 Terra: 15% below OpenAI's list price — $0.85 against $1 per million input tokens and $4.25 against $5 for output. You pay from your Kumo balance, with no separate OpenAI account.
- Does GPT-6.1 Terra work in Russia?
- Through Kumo, yes: the gateway works without a VPN, and the balance can be topped up with a Russian card, SBP or crypto.
- Can I use the model in Codex CLI, Cursor or my own agent?
- Yes. Kumo supports both OpenAI protocols — Chat Completions and Responses — so any OpenAI-compatible client works, Codex CLI included: set the base URL to https://api.kumorouter.com/v1, your Kumo key and the model gpt-6.1-terra.
Sources
- OpenAI announcement and figures: OpenAI on X and AI Weekly coverage, 26 October 2026.
- Independent measurements: Artificial Analysis — GPT-6.1 Terra pushes the cost efficiency frontier, 26 October 2026.
- Prices and parameters: GPT-6.1 Terra on OpenRouter, 27 October 2026.
- Kumo prices and specs: models page, price list v42.