Saba
Mistral Saba is a 24B-parameter language model developed by Mistral to provide accurate and contextually relevant responses for users in the Middle East and South Asia. It supports both text and file inputs, offering efficient performance for regional language tasks and tool-calling applications.
Saba specs and pricing
- Provider: Mistral
- Released: February 17, 2025
- Context window: 33K tokens
- Max output: 26K tokens
- Input: text, files
- Output: text
- Price: $0.20 input / $0.60 output per 1M tokens
- Tool calling: yes
What Saba is good at
- Regional language processing for Middle Eastern and South Asian contexts
- Automated tasks requiring tool integration
- Analyzing documents and text files
To-dos to try with Saba
- Summarize the key cultural themes in this provided document.
- Translate this business email into a formal tone suitable for a regional partner.
- Use tool calling to extract specific data points from this file and format them into a table.
FAQ
- Who developed the Saba model? Mistral developed Saba as part of their model lineup.
- What types of input can the model process? The model accepts both text input and file uploads.
- How much does Saba cost? Saba costs $0.20 per million input tokens and $0.60 per million output tokens through OpenRouter. In todo.is a typical agent step costs about 0.64 credits ($1 = 100 credits).
- What is the context window of Saba? Saba can read up to 33K tokens at once and write up to 26K tokens in one reply.
- Can I use Saba without coding? Yes. In todo.is you pick Saba as your agent's model and write to-dos in plain words, on the web, the Android app, WhatsApp or Telegram. No API key needed: credits are included, or you can bring your own key.
JavaScript is required to use the todo.is app.