GPT-6.1 Terra API Pricing: Costs, Batch and Test Results

👁 114 reads

GPT-6.1 Terra costs $1 per million input tokens and $5 per million output tokens through the API. That is half the rate for GPT-6.1 Sol. I calculated what 16 tasks in my test cost, including one technical drawing and one form, and when the rate doubles.

GPT-6.1 Terra API pricing

These rates are from the official model page. Prices are before tax.

Two additional pricing details:

• Cache writes cost $1.25 per million tokens for a 30-minute cache and $2.50 for a 12-hour cache.
• Regional processing (for example, keeping data in the EU) adds 10% to the rate.

The 272,000-token threshold applies to the entire request. If the input exceeds it, the higher rate applies to all tokens, including output.

What GPT-6.1 Terra cost in my test

On the evening of October 26, I ran 12 tasks in Polish and four tasks based on OpenAI’s announcement, using the OpenAI Standard endpoint with medium reasoning. OpenRouter reports the cost of each task in its response, so the figures below are actual charges, not estimates.

Three takeaways from the bill:

• An image used about 1,800 tokens. My A4 drawing, saved at 150 dpi, cost more in input tokens than the entire form email.
• Reasoning counts as output. For the puzzle, Terra used 1,520 tokens to reason and just 335 for the answer, so 78% of that task’s bill went to reasoning.
• The long prompt was charged like a cache write. With 25,091 input tokens, the cost was $0.0315; at the $1 rate it would have been $0.0252. The difference matches the $1.25 cache-write rate. Caching is worthwhile only if you send the same text again within 30 minutes - then input costs $0.05 per million tokens.

GPT-6.1 Terra compared with other models

I ran the same 12 tasks in Polish through other models. The per-token rate is only part of the cost; the number of tokens a model uses also matters. Terra generated 5,873 tokens, GPT-6.1 Sol 4,627, and Claude Opus 5.5 13,371.

Terra costs half as much per token as Sol, but its higher token use meant the bill was 45% lower. It was 6.9 times cheaper than Opus 5.5 and 8.9 times cheaper than GPT-6 Astra. See my GPT-6.1 Terra guide for the task descriptions and scores.

According to Artificial Analysis, one task from its index costs $0.095 with Terra at medium reasoning and $0.40 at max. Terra scores 48.1 points at max and 41.3 at medium - the higher reasoning level adds 6.8 points for more than four times the cost.

GPT-6.1 Terra pricing compared with GPT-5.6 Terra

OpenAI said API rates are 50% lower than for GPT-5.6 Terra and GPT-6.1 Sol. I checked the pricing: GPT-5.6 Terra costs $2 for input and $12 for output, so input is 50% cheaper and output is 58% cheaper. Compared with GPT-6.1 Sol ($2 / $10), both rates are exactly half as high.

How much does Terra cost in ChatGPT?

You don’t pay per token in ChatGPT. Terra is included with Plus, Pro, Enterprise or Edu plans, but only in ChatGPT Work and Codex - it isn’t available in regular chat. Work and Codex share a usage pool, with limits over five-hour and weekly windows depending on your plan. OpenAI says the same pool will cover more tasks with Terra than with GPT-5.6 Terra, and “significantly more” than with GPT-6.1 Sol or GPT-6 Astra. It hasn’t provided specific figures.

When the pool runs out, personal-plan users can buy credits or switch to their own API key, in which case the rates above apply. See ChatGPT Plus vs Pro for a plan comparison and Codex limits on ChatGPT Plus for Codex limits.

Frequently asked questions

How much does GPT-6.1 Terra cost?

Through the API, it costs $1 per million input tokens and $5 per million output tokens. In ChatGPT, it’s included with Plus, Pro, Enterprise and Edu plans, in Work and Codex.

Is GPT-6.1 Terra free?

No. OpenAI’s announcement doesn’t include it in Free or Go plans, and every API request is charged.

How can you pay less for Terra through the API?

Use Batch or Flex (half price, with slower responses), keep the input below 272,000 tokens and repeat the same prompt prefix to use the cache at $0.05 per million tokens.

Is Terra cheaper than Claude Opus 5.5?

Yes. Its per-token rate is four times lower ($1 / $5 versus $4 / $20), and my 12-task test cost 6.9 times less: $0.064 versus $0.441.

Google · Your sourcesQuantslant higher in your Google - one clickAdd to preferred sources →

Sources and date checked

• OpenAI API: GPT-6.1 Terra model - rates, cache, 272,000-token threshold, Batch, Flex and Fast (checked October 26, 2026)
• OpenAI API pricing - GPT-5.6 Terra and GPT-6 Astra (checked October 26, 2026)
• OpenAI: Introducing GPT-6.1 Terra - availability and rate comparisons (checked October 26, 2026)
• Help Center: ChatGPT Work and Codex - Work and Codex models and billing (checked October 26, 2026)
• OpenRouter: GPT-6.1 Terra - cost per request in my test, and Opus 5.5 and Gemini 4 Argon rates (checked October 26, 2026)
• Artificial Analysis: GPT-6.1 Terra - benchmark task cost (checked October 26, 2026)

Kinetyra