modelgrep
M

Mistral: Saba

mistralai/mistral-saba

ToolsJSON
Use via OpenRouter ↗
Intelligence
Design Elo
Speed
101
59th fastest
Latency
333ms
first token
Input price
$0.200
100th cheapest
Context
33K

How it compares

Faster than81%
of all ranked models
Cheaper than66%
of all ranked models

Overview

Mistral Saba is a 24B-parameter language model specifically designed for the Middle East and South Asia, delivering accurate and contextually relevant responses while maintaining efficient performance. Trained on curated regional...

Benchmarks

independent · via OpenRouter
Artificial Analysis
GPQA Diamond
42%
Humanity's Last Exam
4%
SciCode
24%

Providers & pricing (1)

ProviderIn $/MOut $/MUptime
Mistral$0.200$0.600

Specifications

Context window33K
Max output
Knowledge cutoffSep 2024
Input modalitiestext, file
Output modalitiestext
Prompt caching
Cache read price$0.020/M
ModeratedNo

Saba FAQ

How much does Saba cost?

Saba costs $0.200 per million input tokens and $0.600 per million output tokens via OpenRouter, making it 100th cheapest of 298 paid models.

How fast is Saba?

Saba generates around 101 tokens per second with 333ms time-to-first-token (p50), the 59th fastest tracked model.

What is Saba's context window?

Saba supports a 33K-token context window. It accepts text, file input.

More from mistralai

All Saba alternatives →

Compare head-to-head