OpenAI's GPT-6.1 Terra Costs $1 per Million Input Tokens and Scores 50.3% on Terminal-Bench 4.0, Behind Sonnet 5.5's 70.6%

OpenAI released GPT-6.1 Terra on Monday, October 26, pricing the mid-tier model at $1 per million input tokens and $5 per million output tokens through the API as gpt-6.1-terra. That is half the standard rate of both GPT-5.6 Terra ($2 and $12) and GPT-6.1 Sol ($2 and $10), according to launch coverage by iThinkDifferent and Kingy AI.

The Terminal could not retrieve OpenAI's launch post, which returned HTTP 403 to its fetch on Monday evening. Every OpenAI figure below is therefore as reported by those two outlets, and each is vendor-supplied unless an independent party is named.

What GPT-6.1 Terra costs and where it runs

Terra is available in ChatGPT Work and Codex for Plus, Pro, Enterprise and Edu users from October 26. Codex CLI needs version 0.159.0 or later, selected with codex -m gpt-6.1-terra.

Kingy AI lists a 1,050,000-token context window, 922,000 tokens of maximum input and 128,000 of output, with a June 1, 2026 knowledge cutoff. It takes text and images and returns text. OpenAI has not published a parameter count or weights, and fine-tuning is not supported, Kingy AI said.

Tier (up to 272K input)	Input	Cached input	Output
Standard	$1.00	$0.05	$5.00
Batch or Flex	$0.50	$0.025	$2.50
Fast	$2.00	$0.10	$10.00

Above 272K input tokens the standard rate rises to $2.00 input and $7.50 output. Prices are per million tokens, from Kingy AI.

Where the chart shows Terra behind

OpenAI's headline claim is computer use. On Agents' Last Exam, Terra at medium effort scores 57.6% at $0.29 a task, and 60.0% at max effort for $1.01 a task, Kingy AI reported. The same table puts Claude Opus 5.5 at 63.2% and $8.82 a task.

Coding is the weaker column. On Terminal-Bench 4.0 at max effort Terra scores 50.3%, up from 21.5% for GPT-5.6 Terra, according to iThinkDifferent.

1. GPT-5.6 Terra21.5 %
2. GPT-6.1 Terra50.3 %
3. GPT-6.1 Sol55.9 %
4. GPT-6 Astra58.2 %
5. Claude Sonnet 5.570.6 %

Source: OpenAI results as reported by Kingy AI and iThinkDifferent, accessed 2026-10-26

OpenAI says GPT-6.1 Sol and GPT-6 Astra remain better for the most complex coding, iThinkDifferent reported. Commenters on Hacker News made the cost point sharper. User Starlevel004 wrote that Terra at the highest setting is more expensive than Sol at a low one and still does worse.

An OpenAI employee posting as tedsanders called it "a nice midrange option" for API developers and "not a bombshell release" for subscribers.

What an independent index says

Artificial Analysis, which Kingy AI cites, scored Terra 48 at max effort and 41 at medium on its Intelligence Index v4.3.2. The same snapshot lists GPT-6.1 Sol at 52, Claude Opus 5.5 at 58 and Gemini 4 Argon at 53. Kingy AI said Terra costs about $0.40 a task at max effort on that index.

OpenAI also reported a drop in factual errors on prompts built to induce them. At low effort the rate fell from 53.6% for GPT-5.6 Terra to 16.7%, and at max effort to 4.3%. Kingy AI said these are selected error-prone prompts, not everyday error rates.

OpenAI classifies Terra as Critical for cybersecurity capability under its Preparedness Framework and High for biological and chemical capability, per Kingy AI.

What the price follows

The cut continues the pattern covered in OpenAI's 80% Luna price cut. Google's Gemini 4 Argon went generally available on October 13 at its own promotional rate.

The comparison OpenAI's chart does not settle is Sonnet 5.5, which leads Terra by 20.3 points on Terminal-Bench 4.0. OpenAI did not publish Sonnet's price alongside it in the material The Terminal could read. Independent speed figures for Terra were blank in the Artificial Analysis snapshot Kingy AI used, and are the next numbers to check.