Bring near-Astra intelligence to everyday work with GPT-6.1 Sol on Amazon Bedrock
Amazon Web Services Tanvi Girinath ● Covered by 20 sources
GPT-6.1 Sol is now on Amazon Bedrock. AWS says it’s cheaper per task and stronger at coding, docs, and agent work.
Based on reporting by Amazon Web Services, Tanvi Girinath — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
GPT-6.1 Sol is now generally available on Amazon Bedrock, and AWS is pitching it as a more capable engine for the kind of work agents actually get stuck doing: coding, using software, reading documents, and recovering when a plan goes sideways. The pitch is not just that the model is smarter. It’s that smarter reasoning can cut down the number of tool calls, retries, and human nudges that a task needs before it finishes well.
OpenAI says GPT-6.1 Sol matches GPT-6 Astra on DeepSWE v1.1 at about one-fifth the cost per task. It also beats the best result from GPT-6 Sol by 6.4 percentage points while using lower reasoning effort than that earlier result. That’s the sort of claim AWS wants to hang this release on: not a vague “better AI,” but a model that can do more work per dollar when tasks stretch across many steps.
For software teams, Codex is the obvious place this lands. It can be configured to use GPT-6.1 Sol on Amazon Bedrock for investigation, implementation, and testing, and it works with repositories, local files, terminals, and development tools. AWS also points to the Agent Toolkit for AWS, which connects Codex to AWS documentation, APIs, and services through a single terminal command. Access comes through the desktop app, CLI, and supported IDEs.
The model is also meant for work beyond code. AWS says it can help when agents have to interpret complex documents, choose the right tools, and adapt as conditions change. OpenAI says GPT-6.1 Sol comes close to GPT-6 Astra on complex document analysis and does better than GPT-6 Sol on multistep workflows across business tools. In the desktop app, ChatGPT Work can pull information from files and applications into finished deliverables, while Bedrock APIs can be used to build internal tools, coordinating agents, and customer-facing systems that weigh multiple inputs.
AWS is also leaning hard on controls. You can manage access with IAM, audit activity through CloudTrail, keep traffic inside your network with VPC endpoints powered by PrivateLink, and run inference on hardware-isolated infrastructure with zero-operator access. AWS says prompts and completions are not used for model training, and that you do not need to opt into sharing data with OpenAI. For teams worried about agents making confident mistakes, that mix of model capability and guardrails is probably the real story here.
My take — AI-written commentary, not fact-checked reporting
This is the boringly useful version of AI progress: fewer grand claims, more math. If a model can get closer to the right answer in fewer steps and at lower cost, that matters more than another shiny demo nobody can deploy. The industry keeps selling intelligence; the bill shows up in tool calls.
Read more about this at: Amazon Web Services