Mistral AISaba
Saba is a Mistral AI proprietary model, released 17 Feb 2025. It is not yet ranked: 1 independent result so far, and the index needs at least five across two categories. At $0.300 per million tokens blended it is cheaper than most ranked models. Output speed of 93 tokens per second puts it faster than most, with a first token in 0.5 s. Last measured 17 Feb 2025.stale: no new result in six months
- Blended price
- $0.300/M
- $0.200 in · $0.600 out
- Output speed
- 93 tok/s
- OpenRouter traffic, 7-day median; not yet measured by Artificial Analysis
- First answer
- 0.47 s
- Context
- 33k
- Full answer
- –
- Cost per run
- –
- Released
- 17 Feb 2025
- Composite37
Benchmark results
See all LLM benchmarks. One column per reasoning effort. Rank is among every configuration of every model on that benchmark. Hover a score for the run it came from.
Reasoning
| Benchmark | not stated | Source |
|---|---|---|
| GPQA Diamond (AA)not in index | 42.4%#468 | Artificial Analysis |
| Humanity's Last Exam (AA)not in index | 4.3%#513 | Artificial Analysis |
Composite
| Benchmark | not stated | Source |
|---|---|---|
| AA Intelligence Index v4.3.2 | 6.5#524 | Artificial Analysis |
Where to run it
Every provider serving this model through OpenRouter, with throughput and first-token latency measured on live traffic over the last 30 minutes and each provider’s own price. Purple marks the best in each column.
| Provider | Speed | First token | Input $/M | Output $/M | Context | Quantisation |
|---|---|---|---|---|---|---|
| Mistral | 99 tok/s | 0.47 s | $0.200 | $0.600 | 33k | – |
What a task costs
Estimates from list price, output speed and time to first answer for the best configuration. “With caching” assumes three-quarters of the input is served from the prompt cache. Reasoning tokens are not modelled.
| Workload | Tokens in / out | Cost | With caching | Time |
|---|---|---|---|---|
| Chat reply | 400 / 300 | $0.0003 | – | 3.7 s |
| Summarize a 30-page report | 12,000 / 600 | $0.0028 | – | 6.9 s |
| Code edit | 6,000 / 1,500 | $0.0021 | – | 16.6 s |
| Agentic coding session | 60,000 / 4,000 | $0.014 | – | 43.5 s |
| Structured extraction | 2,000 / 200 | $0.0005 | – | 2.6 s |
More from Mistral AI
See also
Questions people ask
- Why does Saba have no BenchLeader Index score?
- The index needs at least five independent results across at least two categories; as of 25 Oct 2026 Saba has 1 that count. Launch-post scores are never used, so it is ranked once independent publishers have measured it.
- How much does Saba cost per 1M tokens?
- As of 25 Oct 2026, Saba costs $0.200 per 1M input tokens and $0.600 per 1M output tokens, $0.300 blended (three input tokens per output token).
- When was Saba released?
- Mistral AI released Saba on 17 Feb 2025.
Data as of 25 Oct 2026. Compare with another model.
Cite as: BenchLeader, “Saba: benchmarks, pricing, speed and rank”, data as of 25 Oct 2026, https://www.benchleader.com/models/mistral-saba.