OpenAI's GPT-6.1 Terra is now generally available on Amazon Bedrock, and its input and output token prices are at least 50% lower than GPT-6.1 Sol and GPT-5.6 Terra, AWS said. Tokens are the chunks of text a model bills by.
Terra is the everyday-workload line of the GPT-6 family. On OpenAI's Agents' Last Exam evaluation, the company reports that it scored 15 percentage points higher than GPT-5.6 Terra, matched GPT-6 Astra, and finished tasks more than three times faster than both. Those are OpenAI's figures, not an independent test. The model accepts up to 1 million input tokens of text and images, and is reachable through the Bedrock APIs, ChatGPT Work, and Codex.
The price comparison deserves a narrow reading. It sets Terra against two other OpenAI models only. AWS gave no comparison with models from other labs on Bedrock, so the claim says where Terra sits inside OpenAI's lineup and nothing about the wider market.
Within that lineup, the pairing is the point. A cheaper tier that, by OpenAI's own account, matches Astra on one agent benchmark while running much faster is a pitch to teams running high volumes of routine tasks, where per-token cost adds up fastest. One benchmark from the vendor is thin evidence for a whole workload, though, and teams will want to run their own tasks before moving anything.
In a separate launch the same week, AWS made Step Functions workflows available as agent tools through Bedrock AgentCore Gateway. The agent decides when to run a workflow and what input to pass, while Step Functions executes each step in the order defined. AWS's example is a refund, where the eligibility check, manager approval, and credit always happen in sequence.
What the numbers establish is limited: inside OpenAI's own catalog, the everyday tier now costs at most half as much as the models around it. Whether that holds up against other labs' models needs independent testing.