GPT-6.1 Terra at Half the Price: What a Cheaper Model Really Changes

In short: OpenAI launched GPT-6.1 Terra on 26 October 2026. According to OpenAI's launch page, standard API pricing is $1 per million input tokens, $0.05 per million cached input tokens and $5 per million output tokens, and the model is available to ChatGPT Plus, Pro, Enterprise and Edu users in ChatGPT Work and Codex, and in the API as gpt-6.1-terra. Reports put that at half the price of the model above it. Cheaper does not automatically mean better for your task, so test before switching.

Key Takeaways

• Launch date reported as 26 October 2026. Pricing from OpenAI's launch page: $1 per million input tokens, $0.05 per million cached input tokens, $5 per million output tokens.
• Access: Plus, Pro, Enterprise and Edu users in ChatGPT Work and Codex, and the API model name gpt-6.1-terra.
• Reports compare it with GPT-6.1 Sol at $2 and $10, and with GPT-5.6 Terra at $2 and $12, so roughly half the price of either.
• One independent index, Artificial Analysis as relayed by a third-party report, scores Terra 48 against Sol's 52, so it is cheaper and somewhat less capable on that measure.
• Secondary sources add details such as 50% lower Batch and Flex pricing and higher rates for very long prompts. Check OpenAI's pricing page before budgeting.

The Numbers

Cached input, which is text the model has seen recently and can reuse, is priced at $0.05 per million tokens on Terra according to OpenAI's page. For workloads that repeat a long instruction block or document many times, caching can matter more than the headline rate.

Third-party reports also describe 50% lower pricing for Batch and Flex modes, a 2x rate for Fast mode and higher billing for prompts above 272,000 input tokens. These come from secondary sources, so confirm them on OpenAI's own pricing page.

What "Half the Price" Does Not Tell You

A lower price per token is one input to cost, not the cost itself.

• Quality per task. An Artificial Analysis index of 48 for Terra against 52 for Sol, as relayed by one report, suggests a gap on that benchmark. Whether it matters depends on your task. For summarising or routing, a small gap may not change the result. For complex reasoning it might.
• Retries and corrections. If a cheaper model needs more attempts or more human checking, savings shrink.
• Output length. Output tokens cost five times input here, so a model that writes longer answers costs more per task.
• Volume. A halved price matters most at high volume. At small volume, the difference may be pennies.

The same lesson applied to Anthropic's price cut, which we covered in Claude Haiku 5.5 Price Cut: What "75% Cheaper" Really Means.

How to Test Before You Switch

1. Pick one real workflow, such as classifying support emails or drafting replies.
2. Collect 50 to 100 real examples with the outcome you would call correct.
3. Run both models on the same set and record accuracy, the number of outputs a person had to fix, and token usage.
4. Compute cost per correct result, not cost per token.
5. Decide per workflow. A cheaper model for routine steps and a stronger one for hard cases is a common, sensible split.

Keep your prompts and tool definitions separate from any one provider's model name, so switching is a configuration change, not a rewrite. The multi-cloud announcement in OpenAI Models Come to Google Cloud points the same way.

The Short Version

GPT-6.1 Terra is an aggressive price point from OpenAI, with pricing confirmed on its launch page. The decision is not "cheaper or not" but "cheaper for which tasks". Test it on a real workflow, compare cost per correct result and keep your systems portable.

To design an AI system that can switch models without a rebuild, see our AI development service.

Related Insights

• Claude Haiku 5.5 Price Cut: What "75% Cheaper" Really Means
• AI Agent Pricing: Per Seat, Per Task or Per Outcome?
• OpenAI Models Come to Google Cloud: What Multi-Cloud AI Means for Buyers

Sources

• OpenAI, Introducing GPT-6.1 Terra, October 26, 2026
• gHacks, OpenAI Launches GPT-6.1 Terra With API Pricing 50% Lower Than GPT-5.6 Terra, October 27, 2026
• DataCamp, GPT-6.1 Terra: Benchmarks, Pricing and Access