Saba

Provider: Mistral AI EU Official API price

Input price
$0.20
/ 1M tokens
Output price
$0.60
/ 1M tokens
Cache price
$0.02
/ 1M tokens (read)
Context
33K tokens
Max output 26K tokens

Overview

Saba is a premium model from Mistral AI (EU), designed for advanced applications requiring structured interactions. It sits at the higher end of Mistral's pricing spectrum, targeting sophisticated workloads that benefit from its specific capabilities.

Strengths

  • Supports robust function-calling for integrating external tools and APIs.
  • Offers JSON-mode for reliable, structured output generation.
  • Features a solid 32,768 token context window for moderately complex tasks.
  • Provides an extremely low cache read price of $0.02 per million tokens.

Best use cases

  • Developing agentic applications requiring tool use and API orchestration.
  • Automating data extraction into structured formats like JSON.
  • Implementing conversational AI with complex state management and multi-turn interactions.
  • Enhancing RAG systems with efficient cached document retrieval for repeated queries.

Pricing position

Saba is positioned as Mistral AI's most expensive input token model at $0.20/M, significantly higher than models like Mistral Nemo ($0.02/M) or Mistral Small 3 ($0.05/M). Its output price of $0.60/M is competitive among peers with similar input costs, being less expensive than OpenAI's GPT-5.4 Nano ($1.25/M) and DeepSeek V3 0324 ($0.77/M), though slightly higher than xAI's Grok Fast models ($0.50/M).

Caveats

The maximum output token length for Saba is currently unknown, which can impact planning for long-form generation tasks. While its 32,768 token context is solid, other Mistral models like Nemo or Small 3.2 offer significantly larger contexts up to 131,072 tokens at a much lower cost. Users should consider cheaper Mistral alternatives for tasks that do not specifically require Saba's function-calling or JSON-mode capabilities, especially when large context or cost optimization is paramount.

Notes drafted by an LLM from the figures on this page (2026-05-21) and rewritten when the price changes.

Key facts

Price basis
Official API price
Max output
26K tokens
Knowledge cutoff
2024-09-30
Listed on OpenRouter
2025-02-17
Tracked since
2026-05-08
Last updated
2026-11-10

Features

Function callingJSON output

Monthly cost estimate

Scale Input (M) Output (M) Monthly
0.1M (hobby) 0.1 0.05 $0.05
1M (small app) 1 0.5 $0.50
10M (mid scale) 10 5 $5.00
100M (large) 100 50 $50.00

※ Cache and batch discounts not factored in. Detailed calculator

Official resources

Other models from Mistral AI

Similar-priced models

Source: OpenRouter API (mistralai/mistral-saba)