Mistral: Saba
Provider: mistralai · mistralai/mistral-saba
How Mistral: Saba pricing compares
At $0.6 / 1M output tokens, Mistral: Saba is the 11th-cheapest of 20 mistralai models we track. It's cheaper than 76% of the 392 paid models on computetrail (94th-cheapest overall). Output costs 3.0× its input price ($0.2 / 1M in). For a typical 3:1 input-to-output workload that works out to about $0.30 / 1M blended. Reading cached input costs 90% less than fresh input, so prompt caching cuts repeat-context cost sharply. Its 32,768 tokens context window ranks 368th-largest of 392 models with a published limit.
Output price history
Across 60 daily snapshots since 2026-06-29, the output price has held steady at $0.6 / 1M — no change so far.
Similar-priced models
| Model | Provider | Output $/1M | Context |
|---|---|---|---|
| Cohere: Command R (08-2024) | cohere | $0.6 | 128,000 |
| WizardLM-2 8x22B | microsoft | $0.62 | 65,535 |
| Mistral: Mistral Small 3.1 24B | mistralai | $0.555 | 128,000 |
| DeepSeek: DeepSeek V4 Flash Vision Exp | deepseek | $0.66 | 1,048,576 |
| Meta: Llama 3.3 70B Instruct | meta-llama | $0.71 | 131,072 |
| Qwen: Qwen3 8B | qwen | $0.455 | 131,072 |
Prices are recorded daily from public provider data via OpenRouter and normalized to USD per 1M tokens. See methodology.