Text and images in. Labels, categories, structured fields and decisions out. Meet decision-machine-1, the small model behind milliseconds.ai. Route emails, extract invoice data, apply return policies (or build your hot-dog identification empire). One fast API. $0.04 per million input tokens. Output tokens are free. Get 125M free input tokens every month on test keys.
Today we’re launching milliseconds.ai: fast AI decisions on text and images, through one API.
We spend a lot of time building software that needs small answers:
Which team should get this email?
What are the invoice number, supplier and total?
Does this return request meet our policy?
Is that a hot dog? (A critical enterprise workload)
Those answers usually end up in an if statement, a database field or a routing rule, and a whole paragraph beginning with “Certainly!” adds surprisingly little.
For months, we’ve been using small models inside our main product, CloudRaker’s Paperwork API, a platform to parse and classify documents, extract fields and check statements against source text. Our open-weight Vision Language model, rakedoc-nano, still ranks #2 among VLMs on the ParseBench benchmark, right behind Fable.
We always thought of all these features as components of something much bigger. Then TypeSafe AI launched Jev (I believe there's zero chance you missed this launch!)
Credit where it’s due: their launch made us realize how many developers wanted those capabilities on their own. We’d been using them for paperwork; people could use them throughout their apps.
Getting the standalone API live took two days. The models, authentication and billing already existed. The work was rebuilding the request path so our infrastructure stopped taking longer than the inference itself. I wrote up the engineering story here.
What you can build with it today:
Classify, extract and make decisions with decision-machine-1 Send text or images; get labels, structured fields and decisions your app can use Get started with the API, TypeScript/Python SDKs or CLI
$0.04 per million input tokens. Output tokens are free. 125M free input tokens every month on test keys. No card required.
It runs on our existing, battle-tested infrastructure, is SOC-2 Type 2 compliant, and since we run our own inference, we can answer the craziest compliance and enterprise needs (we've been there...)
We’re a small, bootstrapped team, and this grew directly out of tools we built because we needed them ourselves. I’d love to see where you take it.
the pricing framing is what caught my eye, output tokens free and only paying for input makes sense for a classify/route/decide use case since the whole point is a short label back, not a paragraph. we have a similar problem mid-call, deciding in real time whether to route, escalate, or keep going, and the small-model angle matters there because a bigger model's latency alone can blow the moment. is decision-machine-1 something you can self-host or run at the edge for that kind of sub-second constraint, or is it API-only today?
About Milliseconds.ai on Product Hunt
“Fast AI decisions on text and images, via one API”
Milliseconds.ai launched on Product Hunt on September 21st, 2026 and earned 89 upvotes and 6 comments, placing #14 on the daily leaderboard. Text and images in. Labels, categories, structured fields and decisions out. Meet decision-machine-1, the small model behind milliseconds.ai. Route emails, extract invoice data, apply return policies (or build your hot-dog identification empire). One fast API. $0.04 per million input tokens. Output tokens are free. Get 125M free input tokens every month on test keys.
Milliseconds.ai was featured in API (98.7k followers), Developer Tools (519.8k followers) and Artificial Intelligence (479.1k followers) on Product Hunt. Together, these topics include over 219.2k products, making this a competitive space to launch in.
Who hunted Milliseconds.ai?
Milliseconds.ai was hunted by Baptiste Laget. A “hunter” on Product Hunt is the community member who submits a product to the platform — uploading the images, the link, and tagging the makers behind it. Hunters typically write the first comment explaining why a product is worth attention, and their followers are notified the moment they post. Around 79% of featured launches on Product Hunt are self-hunted by their makers, but a well-known hunter still acts as a signal of quality to the rest of the community. See the full all-time top hunters leaderboard to discover who is shaping the Product Hunt ecosystem.
Want to see how Milliseconds.ai stacked up against nearby launches in real time? Check out the live launch dashboard for upvote speed charts, proximity comparisons, and more analytics.
Baptiste Laget
👋 Hey hunters!
Today we’re launching milliseconds.ai: fast AI decisions on text and images, through one API.
We spend a lot of time building software that needs small answers:
Which team should get this email?
What are the invoice number, supplier and total?
Does this return request meet our policy?
Is that a hot dog? (A critical enterprise workload)
Those answers usually end up in an if statement, a database field or a routing rule, and a whole paragraph beginning with “Certainly!” adds surprisingly little.
For months, we’ve been using small models inside our main product, CloudRaker’s Paperwork API, a platform to parse and classify documents, extract fields and check statements against source text. Our open-weight Vision Language model, rakedoc-nano, still ranks #2 among VLMs on the ParseBench benchmark, right behind Fable.
We always thought of all these features as components of something much bigger. Then TypeSafe AI launched Jev (I believe there's zero chance you missed this launch!)
Credit where it’s due: their launch made us realize how many developers wanted those capabilities on their own. We’d been using them for paperwork; people could use them throughout their apps.
So we gave them a home: milliseconds.ai.
Getting the standalone API live took two days.
The models, authentication and billing already existed. The work was rebuilding the request path so our infrastructure stopped taking longer than the inference itself. I wrote up the engineering story here.
What you can build with it today:
Classify, extract and make decisions with decision-machine-1
Send text or images; get labels, structured fields and decisions your app can use
Get started with the API, TypeScript/Python SDKs or CLI
$0.04 per million input tokens. Output tokens are free.
125M free input tokens every month on test keys. No card required.
It runs on our existing, battle-tested infrastructure, is SOC-2 Type 2 compliant, and since we run our own inference, we can answer the craziest compliance and enterprise needs (we've been there...)
We’re a small, bootstrapped team, and this grew directly out of tools we built because we needed them ourselves. I’d love to see where you take it.
Try the demos · Read the docs
Where in your app are you currently waiting for an LLM just to get a category, a field or a yes/no?
Bring us that use case. Super happy to answer questions about the tech behind, the rebuild or the economics!
— bap + the CloudRaker team