Key takeaways
- OpenAI released GPT-6.1 Terra on October 26 in ChatGPT Work, Codex and the API.
- It has a 1,050,000-token context window and costs $1 per million input tokens and $5 per million output tokens.
- Requests over 272,000 input tokens are billed at 2x input and 1.5x output rates.
OpenAI released GPT-6.1 Terra on October 26, a mid-size model it describes as built for visual reasoning, precise computer use and clear writing. It is available in ChatGPT Work, in Codex and in the API, according to OpenAI's API changelog and 9to5Mac.
Terra is the cheaper tier below GPT-6.1 Sol, and the price is the main news. Input costs $1 per million tokens and output $5, with a context window of 1.05 million tokens.
What GPT-6.1 Terra is
Terra replaces GPT-5.6 Terra. It was the one model left out when OpenAI shipped the GPT-6 line last month. According to 9to5Mac, the lineup now has four tiers: Luna, Terra, Sol and Astra. OpenAI aims Terra at routine work such as drafting proposals and filling out forms.
The model page lists these specs:
- Context window of 1,050,000 tokens, with up to 922,000 as input and 128,000 as output.
- Text and image input, text-only output.
- Knowledge cutoff of June 1, 2026.
- Streaming, structured outputs, function calling, file search, web search and prompt caching. Fine-tuning and predicted outputs are not supported.
Pricing: $1 in, $5 out
Standard prices per million tokens are $1 for input, $0.05 for cached input and $5 for output. Cache writes cost $1.25, with a 30-minute retention. Batch and Flex requests are 50% cheaper than Standard. Fast mode costs twice the Standard rate.
There is a catch for long prompts. Once a request goes past 272,000 input tokens, the whole request is billed at twice the input rate and 1.5 times the output rate. The headline price applies to prompts under that size, so the full million-token window is not at $1.
OpenAI says Terra costs 50% less than GPT-5.6 Terra, per 9to5Mac. OpenRouter also lists the model, with the same $1 and $5 prices and a batch variant at half price.
What OpenAI says about quality
OpenAI calls Terra a significant step up from GPT-5.6 Terra. Its own tests, as relayed by 9to5Mac, measured error rates on prompts built to induce mistakes. At low reasoning effort the rate fell from 53.6% to 16.7%. At maximum effort it was 4.3%, close to GPT-6 Astra and GPT-6.1 Sol.
These are OpenAI's numbers on its own test. We found no independent benchmarks yet, and the full announcement page was not readable when this was written.
What to watch
Independent results on computer use and coding will show whether Terra holds up against Sol at a lower price. Developers on long prompts should test their real cost, since the 272,000-token threshold changes the math.