deepseek
Saba vs DeepSeek V4 Flash Vision Exp
76.8 overall · $0.330 / 1M
// mistralai
Saba
mistralai/mistral-saba
tool callingfile inputprompt caching
Saba costs $0.200 per 1M input tokens and $0.600 per 1M output tokens with a 33K-token context window.
A tool-calling model from mistralai, released Feb 17, 2025. LiveBench hasn't published a run for it, so this page carries measured price and specs only — no benchmark numbers.
- Input / 1M
- $0.200
- Output / 1M
- $0.600
- Context
- 33K
- LiveBench overall
- Not evaluated
Specs and pricing
| Input price USD per 1M prompt tokens | $0.200 |
|---|---|
| Output price USD per 1M completion tokens, reasoning included | $0.600 |
| Blended price 3:1 input:output — #101 cheapest of 343 listed models | $0.300 |
| Cached input read Blank means not priced separately — not free | $0.020 |
| Cache write | — |
| Context window | 32,768 tokens |
| Max output | 26,214 tokens |
| Input modalities | text, file |
| Tool calling | Yes |
| Extended reasoning | No |
| Knowledge cutoff | 2024-09-30 |
| Listed | Feb 17, 2025 |
What Saba costs to run
Monthly list cost across five workload shapes, using the published cached-input rate where there is one. A floor, not a quote — batch discounts and cache writes aren't included.
| Workload | Per month |
|---|---|
| Support chatbot 1.2K in / 400 out × 200K requests | $83.04/mo |
| RAG assistant 8K in / 600 out × 100K requests | $124.00/mo |
| Coding agent 40K in / 4K out × 20K requests | $107.20/mo |
| Document extraction 20K in / 1.5K out × 50K requests | $236.00/mo |
| Bulk classification 500 in / 20 out × 5M requests | $470.00/mo |
Compare Saba with
Benchmarked models at a similar price.
Saba FAQ
How much does Saba cost?
Saba lists at $0.200 per 1M input tokens and $0.600 per 1M output tokens, with cached input reads at $0.020 per 1M. A RAG assistant handling 100K requests a month (8K tokens in, 600 out) comes to about $124.00 at list price.
What is the context window of Saba?
Saba accepts up to 32,768 tokens of context and can return up to 26,214 output tokens per response.
Does Saba support tool calling?
Yes — Saba supports tool (function) calling. It accepts text, file input.
How does Saba score on benchmarks?
LiveBench has not published a run for Saba yet. Until it does, compare it on price and specs, or run your own evaluation on your own tasks.
What is the API model ID for Saba?
On OpenRouter the model ID is "mistralai/mistral-saba". It was listed on Feb 17, 2025.
More from mistralai
How these numbers are produced
- Price and specs — provider list data from OpenRouter, refreshed every 15 minutes. “Blended” is a 3:1 input:output mix.
- Scores and ranks — LiveBench release 2026-06-25. Ranks count only models that have a published run; a blank means “not evaluated”, never “bad”.
- Cost per point — the measured dollars LiveBench spent on the run, divided by the score it earned.
Published benchmarks rank models on someone else's tasks. Before committing, see LLM & agent evaluation for building an eval on your own.