OpenAI's New AI Lands on Amazon — Cheaper, Smarter at Real Work
This AI costs 5x less per task — and it's better at finishing the job
Amazon and OpenAI just made a powerful new AI widely available on Amazon Bedrock, Amazon's cloud service where businesses rent AI models the way you might stream shows. It's called GPT-6.1 Sol, and it's designed for tasks that take many steps — reading a document, checking a database, fixing a mistake, and trying again. Most AI tools handle one question at a time. This one is built to finish the whole job.
Here's why that matters to your wallet. Doing a task with AI costs money every time the AI "thinks" or calls a tool. If the AI reasons well, it takes fewer steps — so the same job costs less. OpenAI says GPT-6.1 Sol matches its top-tier Astra model on software engineering tasks at roughly one-fifth the cost per task. Companies that save money on AI often pass those savings along as cheaper or more generous products.
The new model is also better at admitting when it's stuck. If a tool fails or information is missing, it's more likely to say so instead of guessing and continuing. That means fewer confident-but-wrong answers, and more chances for a human to step in before something goes wrong at work.
Finally, privacy. Amazon says the model runs on hardware-isolated servers where "not even AWS operators can access your prompts and completions." If you've worried about sensitive work data being seen, that's a meaningful promise — though it applies to the business version, not necessarily the free chat apps you use at home.
- GPT-6.1 Sol matches OpenAI's top-tier Astra model on coding tasks at about one-fifth the cost per task.
- It's better at flagging when a tool fails or information is missing, so it's less likely to barrel ahead with a wrong answer.
- Amazon says the model runs on secure servers where even AWS staff can't see your data during use.
Why It Matters
Cheaper, more reliable AI could mean better tools for your job — and less frustration when AI gets things wrong.