Saba
by Mistralmistralai/mistral-saba
LLMCompare cannot run Saba yet.No route to this model is configured on this site yet. Its prices and parameters are listed below for reference.
See the models that can be run nowSaba at a glance
- Context length
- 32,768 tokens
- Maximum output
- 26,214 tokens
- Accepts
- Text, File
- Returns
- Text
- Features
- JSON outputTool callingPrompt caching
- Reasoning effort levels
- Not supported
- Released
- 17 Feb 2025
- Knowledge cutoff
- 30 September 2024
Model facts synced from the OpenRouter model list on 21 Oct 2026.
Saba pricing by route
A route is the service LLMCompare calls to reach the model. Saba has 1 route. Prices are in US dollars.
| Route | On LLMCompare | Inputper 1M tokens | Outputper 1M tokens | Cache readper 1M tokens | Price verified on | Source |
|---|---|---|---|---|---|---|
| OpenRouterDefault | Not configured | $0.20 | $0.60 | $0.02 | openrouter.ai/mistralai/mistral-saba |
The cost shown in an LLMCompare report is calculated from these prices as they were at the time of the run.
Cost calculator
Estimate what a batch of calls to Saba costs. The starting numbers are examples. Change them to match your task.
Part of the input tokens above, billed at the cache read price.
- Cost per call
- $0.0005
- 1 credits
- Cost of 100 calls
- $0.05
- 100 credits
Calculated from the OpenRouter prices verified on 21 Oct 2026. Credits are the cost plus a 20% service fee, at $0.001 per credit, rounded up for each call, with a minimum of 1 per call. A real run is charged for the usage the provider reports.
Parameters you can set
These are the parameters LLMCompare can set when it calls the model through OpenRouter.
| Parameter | Type | Range or options | Default |
|---|---|---|---|
temperature | Number | 0 to 2in steps of 0.1 | Provider default |
top_p | Number | 0 to 1in steps of 0.05 | Provider default |
max_tokens | Integer | 1 to 26,214 | Provider default |
seed | Integer | -2,147,483,648 to 4,294,967,295 | Provider default |
response_format | One of | textjson_object | Provider default |
About Saba
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...
Description as published for this model. LLMCompare has not edited it.
Models to compare with Saba
Text models from the same provider and models at the nearest price. The order is not a ranking.
| Add to comparison | Model | Context | Max output | Input priceper 1M tokens | Output priceper 1M tokens | Features | Released | On LLMCompare |
|---|---|---|---|---|---|---|---|---|
Mistral Small 4Mistralmistralai/mistral-small-2603 | 262k | 210k | $0.15 | $0.60 | ReasoningJSONToolsCacheVision | No route yet | ||
Mistral Small 3.1 24BMistralmistralai/mistral-small-3.1-24b-instruct | 128k | 102k | $0.351 | $0.555 | JSONToolsVision | No route yet | ||
Codestral 2508Mistralmistralai/codestral-2508 | 256k | 205k | $0.30 | $0.90 | JSONToolsCache | No route yet | ||
Cydonia 24B V4.1TheDrummerthedrummer/cydonia-24b-v4.1 | 131k | 118k | $0.30 | $0.50 | JSONCache | No route yet | ||
UnslopNemo 12BTheDrummerthedrummer/unslopnemo-12b | 1.02M | 819k | $0.40 | $0.40 | JSON | No route yet | ||
Llama 3.1 70B InstructMetameta-llama/llama-3.1-70b-instruct | 131k | 16.4k | $0.40 | $0.40 | JSONTools | No route yet |
Questions about Saba
- How much does Saba cost?
- On the OpenRouter route, Saba costs $0.20 per 1 million input tokens and $0.60 per 1 million output tokens, in US dollars. Input tokens that Saba reads from the prompt cache cost $0.02 per 1 million tokens. LLMCompare verified this price on 21 Oct 2026.
- What is the context length of Saba?
- Saba has a context length of 32,768 tokens and returns up to 26,214 output tokens in one call.
- What inputs does Saba accept?
- Saba accepts these inputs: Text, File. Saba returns: Text.
- Can I test Saba on my own data?
- Not yet. LLMCompare lists the prices and parameters of Saba, but no route to Saba is configured on LLMCompare, so it cannot be selected for an evaluation.