Generative models

Language models built to scale with you

Command is a family of high-performance, scalable language models. Pick the size that fits the task, ship to production with confidence, and grow from prototype to global traffic on the same API.

Benchmark coverage

Reasoning78
Summarization61
Code44
Retrieval33

Powering products at the teams shipping with Command

MeridianQuantaNorthpeakOrbixCastlebankHalcyonVantageBrightlineLumenFoundryMeridianQuantaNorthpeakOrbixCastlebankHalcyonVantageBrightlineLumenFoundry
FoundryLumenBrightlineVantageHalcyonCastlebankOrbixNorthpeakQuantaMeridianFoundryLumenBrightlineVantageHalcyonCastlebankOrbixNorthpeakQuantaMeridian

Capabilities your product can rely on

Reason over long inputs, call your tools, and stream grounded answers — everything a real generative feature needs.

5/12, 9:15 AM

Command, summarize what the linked reports say this quarter.

Harriet 9:16 AM

@Command what should we take away from this quarter's reports?

Command APP 9:16 AM

Here is what the linked reports point to:

  • A clear summary of the quarter, drawn from every linked report
  • The three risks most likely to affect next quarter's targets
  • Suggested follow-up questions, each with a cited source

Cost per million tokens

0.0x lower

Serve more for less

Command's smaller tiers handle routine work at a fraction of the cost of a single large model.

Faster, cheaper, and steady under load. Performance that holds up from the first prototype to global traffic.

Want the full picture on Command performance?

Download the model benchmark report

Fits the way you already build

Call Command from your stack, scale it on your terms, and tune it to your domain.

SDKs & open API

Call Command from Python, TypeScript, and Go with a single API, or wire it into any language over plain HTTP.

Streaming & batch endpoints

Stream tokens for interactive features, or run high-volume jobs through the batch endpoint at a lower rate.

Fine-tuning & deployment

Adapt any model tier to your domain, then deploy it to a private endpoint with predictable, dedicated capacity.

We prototyped on the largest Command model, then dropped down two tiers for production. Quality held, latency improved, and our serving bill fell by more than half. The same API made the switch effortless.

MERIDIAN

Renata Voss

VP, Engineering

No training on your prompts

Inputs and outputs sent to Command are never used to train our models, by default and without exception.

Scoped API keys

Issue keys per project with fine-grained limits, so each Command integration only reaches what it needs.

SOC 2 Type II

Independent audits confirm the platform's security controls meet the highest industry standards, year after year.

Encryption & retention controls

Traffic is encrypted in transit and at rest, and you decide how long request data is retained.

Safety guardrails

Built-in classifiers and configurable filters keep Command's generations within the bounds you set.

HIPAA aligned

A HIPAA program and signed BAAs let regulated teams build healthcare features on Command with confidence.

BREVA

How a support team cut response times with Command

A global appliance maker drafts replies with Command and routes only the hard cases to people. Customers wait less, and agents focus where they matter most.

Build your next feature on Command