Amazon Web Services has made OpenAI’s GPT-6.1 Terra generally available on Amazon Bedrock, bringing the model to enterprise customers on the same day OpenAI launched it. The move puts one of OpenAI’s cheapest computer-use models inside the cloud platform many large companies already use to run artificial intelligence workloads.
GPT-6.1 Terra is designed for everyday work that needs a balance of capability, speed and cost, according to AWS. It accepts text and image inputs and supports a context window of up to 1 million tokens — the amount of material the model can take into account at once. AWS lists drafting proposals, interpreting technical drawings, investigating codebases and building agents that work across business systems among the jobs it is aimed at.
What Bedrock availability changes for enterprise teams
The significance of the announcement is less the model itself than where it can now run. Teams can call GPT-6.1 Terra directly through Amazon Bedrock APIs, or configure ChatGPT Work and Codex to use the Bedrock-hosted copy of the model. That matters for companies that already standardise on AWS, because they can apply existing AWS controls to secure workloads, govern access and audit model invocations, rather than adding a separate vendor relationship for this model.
It also continues a clear pattern: OpenAI’s models are no longer confined to OpenAI’s own platform. The same week, OpenAI’s models expanded onto Google Cloud, and GPT-6.1 Terra also reached GitHub Copilot and Microsoft Foundry on launch day. For buyers, the model is becoming a portable component that runs wherever their infrastructure already lives, not a destination they must move to.
Price, performance and the vendor numbers
Through OpenAI’s own API, GPT-6.1 Terra costs $1 per million input tokens and $5 per million output tokens — at least 50% lower than both GPT-6.1 Sol and its predecessor, GPT-5.6 Terra. Readers should note one distinction: AWS’s announcement does not list Bedrock token prices, and directs customers to the Bedrock documentation for pricing and supported regions. The $1 and $5 figures are OpenAI’s standard API rates, not confirmed Bedrock rates.
On performance, AWS restates OpenAI’s own evaluation claims. On the Agents’ Last Exam test of professional tasks, GPT-6.1 Terra scored 15 percentage points higher than GPT-5.6 Terra, performed on par with the flagship GPT-6 Astra, and completed tasks more than three times faster than GPT-6 Astra and GPT-6.1 Sol. OpenAI also reports that its error rate on an internal set of prompts designed to trigger mistakes fell from 53.6% with GPT-5.6 Terra to 16.7% at low reasoning effort. These are vendor-reported figures: the announcement provides no independent test results, so they should be read as OpenAI’s claims until third-party labs reproduce them.
What Terra is for — and what it is not
Terra sits in the middle of OpenAI’s GPT-6 family, between the cheaper Luna models and the more capable Sol and Astra tiers. Its pitch is “computer use”: an AI that looks at a screen and clicks, types and moves between applications the way a person would, for routine office and coding work. OpenAI still points customers with the hardest coding problems to GPT-6.1 Sol or GPT-6 Astra.
For high-volume agent work — form-filling, invoice processing, code investigation, document drafting — the combination of lower price, faster task completion and a million-token context window is the point of the product. Whether independent tests confirm the vendor numbers will decide how many teams actually switch their agents over. What is not in doubt is the distribution: within a day of launch, the model was available through OpenAI’s API, ChatGPT Work and Codex, GitHub Copilot, Microsoft Foundry and now Amazon Bedrock.
What to watch next
Three things will show whether the Bedrock launch matters beyond launch-week coverage. First, Bedrock pricing and regional availability, which AWS has not yet published in the announcement. Second, independent benchmark results that confirm or challenge OpenAI’s exam and error-rate claims. Third, whether enterprise teams actually move production agent workloads onto the Bedrock-hosted model, or treat it as one more option in an already crowded catalogue.
Frequently asked questions
What is GPT-6.1 Terra?
It is OpenAI’s mid-tier GPT-6 model for everyday computer-use work, including drafting, coding assistance and agents that operate business applications. It replaces GPT-5.6 Terra.
How can enterprises access it on AWS?
Through Amazon Bedrock APIs, or by configuring ChatGPT Work and Codex to use the Bedrock-hosted model. AWS says teams can start in the Bedrock console or through supported Bedrock APIs.
How much does it cost?
OpenAI’s standard API price is $1 per million input tokens and $5 per million output tokens. AWS has not published Bedrock-specific token prices in its announcement, so Bedrock customers should check the Bedrock documentation.
Is it better than GPT-6 Astra?
OpenAI claims it matches Astra on its Agents’ Last Exam test while running more than three times faster, but those are OpenAI’s own figures, not independent results. OpenAI still recommends Sol or Astra for the hardest coding work.
Sources
- AWS announcement of GPT-6.1 Terra general availability on Amazon Bedrock, via GIXtools
- Oossa: OpenAI GPT-6.1 Terra launches on Amazon Bedrock
- Subagentic: GPT-6.1 Terra lands for computer-use work
- Modeldex: OpenAI’s GPT-6.1 Terra halves the price of computer-use AI
Related coverage on Emadan: OpenAI to Sell Its Top Models on Google Cloud in Latest Enterprise Push
