Saba
Feb 2025
Retired 2025-09-30Replace withMistral Small 4Newer version
MiniMax M3.1 FlashCheaper, as good
GPT-6 LunaBetter
MiMo v2.6 ProOpen weights
Mistral Small 4Newer versionMiniMax M3.1 FlashCheaper, as goodGPT-6 LunaBetterMiMo v2.6 ProOpen weightsSabaCompare inon
Benchmarks of Saba3
BenchmarkByScoreRank
Overall1
Science1
Legacy · saturated or no longer reported1
No benchmark matches
Ranking?
Retired models are left out of the current rankings.
Alternatives?
ModelScoreCostPer taskWhat changes
CheaperClaude Haiku 5.5max×6.7-33%$0.21Remove sampling parameters+5BetterGPT-6 Lunalow×3.3-33%$0.005Function calling only with none · Chat Completions+3OpenMiMo v2.6 Prothinking×7.1+81%$0.13Pass back reasoning_content+3CheaperMiniMax M3.1 Flashthinking×6.6-13%$0.22Complete model response must be appended
CheaperGLM 5.3 FlashmaxOpen×6.4-21%$0.28Defaults to max · Self-hosted+4ModelScoreCostPer taskWhat changes
Claude Haiku 5.5max×6.7-33%$0.21Remove sampling parameters+5MiniMax M3.1 Flashthinking×6.6-13%$0.22Complete model response must be appendedGLM 5.3 FlashmaxOpen×6.4-21%$0.28Defaults to max · Self-hosted+4
Qwen3.8 FlashthinkingOpen×6.1-23%$0.37max_tokens range · QwenCloud+5GPT-6 Lunamax×5.9-33%$0.068Function calling only with none · Chat Completions+3ModelScoreCostPer taskWhat changes
GPT-6 Lunalow×3.3-33%$0.005Function calling only with none · Chat Completions+3
Granite 4.2 3BthinkingOpen+40%-83%$0.006
Gemma 4 E4BthinkingOpen+37%-87%$0.008GPT-5.6 Lunalow×3.2+50%$0.01GPT OSS 20BhighOpen+38%-88%$0.011ModelScoreCostPer taskWhat changes
MiMo v2.6 ProthinkingOpen×7.1+81%$0.13Pass back reasoning_content+3
DeepSeek V4.1 Flash 1017maxOpen×7+75%$0.26Pass back reasoning_content · Chat Completions+4GLM 5.3maxOpen×6.9+617%$2.01Reasoning cannot be disabled+3
Step 5 PreviewthinkingOpen×6.7+375%$1.03web_search not available · Chat Completions+2
Kimi K3maxOpen×6.7+1,900%$2.00Sampling parameters are fixed+3ModelScoreCostPer taskWhat changes
Mistral Large 4thinking×5.9—$1.13Infinite whitespace stream+4Mistral Medium 3.5highOpen×2.2—$0.50Mistral Small 4thinkingOpen+74%—$0.015Mistral Large 3non-thinkingOpen+43%—$0.031Magistral Medium 1thinking+40%——Devstral Mediumnon-thinkingOpen+39%—$0.40 / $2Magistral Small 1thinkingOpen+27%—$0.70 / $1.16Mistral Small 3non-thinkingOpensame—$0.05 / $0.08Context: 33K → 32KAll Mistral models →FAQ
How much does Saba cost?›
$0.20 per 1M input tokens and $0.60 per 1M output tokens; cached input $0.02 per 1M.
Is Saba free to use?›
There is no free API; the cheapest paid price is $0.20 per 1M input tokens.
How can I use Saba?›
Through the API of Mistral.
What is the context window of Saba?›
33K tokens, with up to 26K output tokens.
When was Saba released? Is there a newer version?›
Released 2025-02-17. It has been superseded by Mistral Small 4.
Is Saba open source?›
No, Saba is a closed model available through an API. The closest open-weights option here is MiMo v2.6 Pro.
Can Saba read images?›
It accepts text and PDF files as input.
What is Saba?›
Saba is a closed language model from Mistral, released Feb 2025.
What are the best alternatives to Saba?›
Claude Haiku 5.5 (cheaper), GPT-6 Luna (better), MiMo v2.6 Pro (open weights).