North Small Translate API Guide: Model ID, Setup & Pricing

01

Overview

North Small Translate is available through Cohere's hosted API under the model ID north-small-translate-1-0. The API route is the simplest way to evaluate the model without provisioning the large GPU footprint required for self-hosting.

02

Current API status

Cohere's documentation currently says North Small Translate is free for both trial and production keys until rate limits are reached. That wording matters: it does not establish an unlimited permanent free tier, and it does not mean every account has identical throughput.

Before production use, check the live rate-limit documentation and your account limits.

03

Model ID

Use north-small-translate-1-0.

Cohere lists support through Chat V2, Chat V1 and Chat Completions. For a new integration, use the current recommended API version from Cohere's documentation rather than copying an older example blindly.

04

Basic workflow

1. Create or use a Cohere API key.
2. Send the source text to the Chat API with model north-small-translate-1-0.
3. Tell the model the source and destination language explicitly.
4. Include terminology or style constraints when they matter.
5. Validate output on representative domain content before scaling.

For localization systems, keep stable terminology rules outside the prompt when possible so they can be versioned and tested.

05

Context and output limits

The model supports up to 16K input tokens and 16K output tokens. That makes it suitable for long passages and some document-scale workflows, but very large documents may still need chunking or staged processing.

06

Agentic multi-pass workflow

Cohere reports a stronger WMT26 score when the model is used in an agentic translation flow that identifies and corrects errors in an additional pass. A practical implementation can therefore use two stages: first translation, then targeted review/correction.

Do not describe this as proof that every double-pass translation is better. The improvement is based on Cohere's own evaluation.

07

Pricing

The hosted model is currently documented as free until rate limits are reached. Cohere does not present that as a permanent universal token price.

For dedicated commercial deployment, Model Vault uses instance-based pricing. The current standard pricing table lists North Small Translate at $57.50 for its flex performance tier, while commitment and customized pricing require Cohere contact.

RWS Language Weaver is another enterprise route for organizations that want a full translation/localization platform rather than a raw model API.

08

When the API is the better choice

Use the hosted API when you want fast evaluation, do not want to operate large GPU infrastructure, or need to integrate translation into an application with minimal deployment overhead.

Use self-hosting or Model Vault when data control, sovereignty, predictable dedicated capacity or infrastructure isolation matter more than operational simplicity.

09

Production checklist

• Verify the live model ID and endpoint.
• Check current rate limits for your key type.
• Test your highest-value language pairs.
• Evaluate domain terminology, names and formatting preservation.
• Decide whether one-pass or review/correction workflows are appropriate.
• Recheck license/commercial terms if moving away from hosted API into downloaded weights.
• Track quality regressions when prompts, model versions or chunking logic change.

10

Important limitation

Cohere's public launch benchmarks are not a substitute for evaluating your own language pairs and domain. Machine translation can look strong on aggregate metrics while still failing on terminology, legal language, brand names or low-resource edge cases.