OpenAI has documented GPT-6.1 Terra, a model tier built around large-context API workloads and lower-cost processing modes. According to OpenAI's GPT-6.1 Terra model documentation, the model supports a 1,050,000-token context window and up to 128,000 output tokens. Its Batch and Flex API prices are 50% lower than its Standard prices, creating a clear cost lever for companies that can choose how and when work is processed.
The practical story is not simply that a new model is cheaper. Terra gives teams a pricing choice tied to workload urgency. Standard pricing remains relevant for work that needs an immediate response, while Batch and Flex can reduce token costs for jobs that are less time-sensitive. That can matter for document analysis, large knowledge-base processing, content pipelines, and other recurring AI tasks where volume can otherwise make experimentation expensive.
What GPT-6.1 Terra pricing changes
OpenAI lists GPT-6.1 Terra's Standard text pricing at $1.00 per million input tokens and $5.00 per million output tokens. Batch and Flex are listed at 50% below those Standard rates. On that basis, the equivalent prices are $0.50 per million input tokens and $2.50 per million output tokens for either lower-cost mode.
Terra also has separately priced cached-input and cache-write components. That detail matters for applications that reuse long prompts or recurring context, since total API spending depends on the mix of new input, cached input, output, and the selected processing mode.
| Pricing mode or model | Input price per 1M tokens | Output price per 1M tokens | What is confirmed |
|---|---|---|---|
| GPT-6.1 Terra Standard | $1.00 | $5.00 | Listed text rates |
| GPT-6.1 Terra Batch or Flex | $0.50 | $2.50 | 50% lower than Terra Standard pricing |
| GPT-5.6 Terra after July 2026 adjustments | About $2.00 | About $12.00 | Reported prior pricing baseline |
The historical comparison provides useful context, but it should be handled carefully. GPT-5.6 Terra was reported at around $2.00 per million input tokens and $12.00 per million output tokens after July 2026 price adjustments. The GPT-6.1 Terra documentation specifically confirms a 50% reduction for Batch and Flex relative to GPT-6.1 Terra Standard, rather than publishing a simple, like-for-like statement that every Terra use case is 50% cheaper than GPT-5.6 Terra.
Choosing between Standard, Batch, and Flex
For a business team, the key question is whether each task requires a response immediately. A useful way to separate workloads is:
- Standard: interactive customer-facing features, live internal assistants, and workflows where response timing is central.
- Batch: planned, high-volume processing such as enriching records, classifying documents, or generating drafts in bulk.
- Flex: work where lower cost is more important than the response priority associated with Standard processing.
The lower-priced modes can make a difference when an application handles many documents or repeatedly processes large bodies of text. Terra's 1.05 million-token context window also means a single request can accommodate unusually large inputs. That may reduce the need to split some large internal materials into many smaller requests, although it does not eliminate the need to measure token usage, outputs, and application performance in a real workflow.
Plan migrations should start with measurement
The model's published rates are not a complete budget forecast. Input and output tokens are charged separately, and output is substantially more expensive at the Standard rate. A workflow that produces lengthy generated reports can therefore have a very different cost profile from one that mainly searches, classifies, or extracts data from documents.
Before moving an existing workflow, teams should identify the tasks that create the most token volume and determine whether they truly need Standard processing. Testing a representative sample can reveal whether a Batch or Flex approach fits operational needs without disrupting a process that depends on immediate results. The appropriate choice will vary by workload, rather than by a broad assumption that the lowest listed rate is always the best operational option.
OpenAI's broader GPT-6 pricing strategy offers additional context. Its September 29, 2026 announcement for GPT-6 Sol and GPT-6 Luna described 50% reductions relative to GPT-5.6 promotional pricing. GPT-6.1 Terra's lower-cost modes continue that emphasis on reducing cost per task, while preserving a Standard option for workloads where timing matters.
The available official documentation does not currently confirm GPT-6.1 Terra availability in ChatGPT Work or Codex. OpenAI Help Center materials list GPT-6.1 Sol, GPT-6 Sol, and GPT-6 Luna for those products, but do not explicitly list GPT-6.1 Terra. As a result, businesses should not assume that Terra changes Work or Codex usage allowances until OpenAI publishes specific rollout details for those services.
For companies looking to turn lower model costs into repeatable operational savings, Scalevise can help assess which AI tasks are suitable for delayed processing, connect models to the right business systems, and build safeguards around production workflows. A focused AI workflow automation implementation can replace manual handoffs while keeping time-sensitive work on the right path. Discuss an AI automation project with Scalevise.
Frequently Asked Questions
What is GPT-6.1 Terra?
GPT-6.1 Terra is an OpenAI API model with a 1,050,000-token context window and a maximum output capacity of 128,000 tokens.
How much do GPT-6.1 Terra Batch and Flex cost?
OpenAI states that Batch and Flex prices are 50% lower than Standard pricing. Based on Standard text rates of $1.00 per million input tokens and $5.00 per million output tokens, that equates to $0.50 input and $2.50 output per million tokens.
Is GPT-6.1 Terra 50% cheaper than GPT-5.6 Terra?
The official Terra documentation confirms that GPT-6.1 Terra Batch and Flex are 50% cheaper than GPT-6.1 Terra Standard. It does not provide a single like-for-like 50% comparison with GPT-5.6 Terra for every usage mode.
Is GPT-6.1 Terra available in ChatGPT Work and Codex?
Current official Help Center information does not explicitly list GPT-6.1 Terra for ChatGPT Work or Codex, so its availability and any related usage allowance changes are not confirmed.
Conclusion
GPT-6.1 Terra gives API users a substantial pricing advantage when their workloads can use Batch or Flex processing. Its large context window and lower-cost modes make it particularly relevant for high-volume, non-immediate AI tasks. The sensible next step is to match processing mode to business urgency and validate costs with real workload measurements, while treating ChatGPT Work and Codex availability as unconfirmed until OpenAI documents it.
Top comments (1)
Worth adding the arithmetic to the 50% claim, because it only applies cleanly to the share of the bill that is input. A document rewrite at 3k input / 1k output is $0.003 + $0.005 = $0.008 at $1/$5, and $0.0015 + $0.0025 = $0.004 on Batch, so 50% holds there. A classify-and-extract job at 10:1 input:output mostly pays for input and tracks it. A summarize-to-long-report job at 1:4 pays mostly for output, where the same headline discount still halves the bill but the absolute saving comes from the output side. What the price sheet does not show is how the modes interact: a cached prefix on Batch is not simply both discounts added together.
The other lever in the same table is the one that decides most recurring pipelines: cached input versus new input, because it changes what you pay rather than when you pay it. Measuring prefix_tokens / total_input_tokens before switching modes is usually the bigger number, and it is available before you migrate anything.
One practical note from running delayed queues: once Batch exists, "later" quietly becomes the default and a job can drift past the moment its result was useful. A required deadline field on every queued job keeps a cost saving from turning into a product regression.