New Standard One SH benchmark results on the Decision Index 0.3

Open-model inference for
products and coding

Pay per token. Or get a plan for your product or coding.

Get access View pricing

~/app/inferenceSSE

❯ python stream.py

↳ POST /v1/chat/completions

│ Authorization Bearer ••••••••

│ stream true

✓ Connected 200 OK

Response complete

data: {"delta": "Connect your product."}

data: {"delta": " Stream responses."}

data: {"delta": " Ship useful features."}

Connect your product. Stream responses. Ship useful features.

Usage 128 in / 4 out tokens

✓ Complete [DONE]

❯ python classify.py

↳ POST /v1/chat/completions

✓ Connected 200 OK

data: {"category": "support"}

data: {"priority": "normal"}

✓ Complete Usage recorded

Standard API

Use inference on demand, with no capacity commitment. Pay only for the tokens you use.

Explore API

Monthly base$1,200 / model

Product demandWithin base · $0 extraCapacity over time

+$5+$5+$5

Normal use2h peak2h peak2h peak

Within base · no extra charge+$0

Auto scaling: $5 / unit / 2 hours · when authorized

Product Plan

Reserve steady base capacity and scale for peaks. A data platform for your product’s data is coming soon.

Explore Product Plan

Fix the rounding bug in invoice.ts

Review complete···

Round once, after adding the items.

Keep the calculation precise, then round the final total to two decimal places.

⌘ invoice.tsSuggested change

12 const total = items.reduce(13   (sum, item) => sum + item.price, 0);14 − return total;14 + return Math.round(total * 100) / 100;

✓ Patch ready for your review

Want me to add these cases to invoice.test.ts?

＋Ask a follow-up…K3

Coding Plan

Switch between models in your coding tool. No 5-hour window, no weekly quota, no monthly token allowance. $149/week or $549/30 days per developer.

Explore Coding Plan

Standard One turns
text and images into decisions.

Ask a question about a message or an image. One 3B and One 8B answer with one of your own options and say how sure they are.

One 8B SH scores 57.4 on a public benchmark

Highlights from real-time runs. The model reads the game state as text, not the screen.

standard-one-3b

Dana Whitfield 9:41

Charged twice for my order

Hi, I was charged twice for order #48213 this morning. Please refund the extra charge — I need this sorted today.

RouteBilling 0.88

RefundYes 0.94

UrgencyHigh 0.58

Your code acts above 0.85Route to Billing · Start a refund

A person decidesUrgency

22.6 ms

Try it locally, or call our API. Download the open weights from Hugging Face under Apache 2.0, or use One 3B on our API at 45–53% off its list price. Output is free.

Text input (/1M toks): $0.018 ~~list price $0.038~~
Image input (/1M toks): $0.055 ~~list price $0.10~~
Output: $0

Start with free credit.

Try Standard One

Explore more models we offer.

Eight open models from six labs, plus Standard One 3B and 8B, all through one API.

Explore API models

Kimi K3, GLM-5.3, and DeepSeek V4.1 Flash are also in Product Plan and Coding Plan.

Then choose how you pay for them.

API and Product Plan both power AI features in your product. Choose between token-based billing and a monthly subscription.

API · Pay per token

Usage-based billingPay for tokens used

Your bill follows the input and output tokens your product uses.

Explore API

Product Plan · A monthly base

Guaranteed baseOptional extra capacity

No monthly token cap. Use free extra capacity when available, or enable auto scaling for traffic peaks.

Explore Product Plan

Your data stays under your control.

We process your inputs and outputs to serve your requests. Retention depends on the service and features you use.

Zero Data Retention. Standard API (including Standard One) and Product Plan inference does not keep your inputs or outputs as content logs, histories, or backups.

Operational metadata. API and Product Plan content-free service metadata is identifiable for up to 90 days, then retained only in sufficiently de-identified form. Account, billing, tax, and support records follow separate criteria.

Product Plan data platform (planned). It is not released. If offered, its availability, scope, pricing, storage terms, and controls will be announced separately.

Coding Plan. Coding Plan inputs and outputs may be used for our model training and evaluation under its separate terms. This does not apply to inference covered by ZDR.

Read the data terms

Standard API and Product Plan inference have ZDR terms. Any future Product Plan data platform would store data designated by the customer under separate terms.

Product Plan stored data is used to provide the customer-directed storage feature; storage will be optional if and when released.

Open-model inference for products and coding