Devstral Small (May '25)
May 2025
Older versionReplace withMistral Medium 3.5Newer version
GPT OSS 120BCheaper, as goodGPT-6 LunaBetter
MiMo v2.6 ProOpen weights
Mistral Medium 3.5Newer versionGPT OSS 120BCheaper, as goodGPT-6 LunaBetterMiMo v2.6 ProOpen weightsDevstral Small (May '25)Compare inon
Benchmarks of Devstral Small (May '25)12
BenchmarkByScoreRank
Overall1
Knowledge4
Science2
Coding1
Documents1
Legacy · saturated or no longer reported3
No benchmark matches
Ranking?
Retired models are left out of the current rankings.
Alternatives?
ModelScoreCostPer taskWhat changes
CheaperQwen3.5 4BthinkingOpen+50%-20%$0.03 / $0.15BetterGPT-6 Lunalow×2.5+167%$0.005Function calling only with none · Chat Completions+3OpenMiMo v2.6 Prothinking×5.3+625%$0.13Pass back reasoning_content+3
CheaperNemotron 3.5 LightningthinkingOpen+47%-35%$0.079CheaperGPT OSS 120BhighOpen+33%-20%$0.092Context: 256K → 131KModelScoreCostPer taskWhat changes
Qwen3.5 4BthinkingOpen+50%-20%$0.03 / $0.15Nemotron 3.5 LightningthinkingOpen+47%-35%$0.079GPT OSS 120BhighOpen+33%-20%$0.092Context: 256K → 131KGPT OSS 20BlowOpen+14%-52%$0.018 / $0.09Context: 256K → 131K
Granite 4.2 3BthinkingOpensame-30%$0.006Context: 256K → 128KModelScoreCostPer taskWhat changes
GPT-6 Lunalow×2.5+167%$0.005Function calling only with none · Chat Completions+3GPT-5.6 Lunalow×2.4+500%$0.01Mistral Small 4thinkingOpen+29%+250%$0.015MiMo v2.5thinkingOpen×2.5+133%$0.019Granite 4.2 8BthinkingOpen+27%+43%$0.024Context: 256K → 128KModelScoreCostPer taskWhat changes
MiMo v2.6 ProthinkingOpen×5.3+625%$0.13Pass back reasoning_content+3
DeepSeek V4.1 Flash 1017maxOpen×5.2+600%$0.26Pass back reasoning_content · Chat Completions+4
GLM 5.3maxOpen×5.1+2,767%$2.01Reasoning cannot be disabled+3
Step 5 PreviewthinkingOpen×5+1,800%$1.03web_search not available · Chat Completions+2
Kimi K3maxOpen×5+7,900%$2.00Sampling parameters are fixed+3ModelScoreCostPer taskWhat changes
Mistral Large 4thinking×4.4—$1.13Infinite whitespace stream+4Mistral Medium 3.5highOpen+62%—$0.50Mistral Small 4thinkingOpen+29%—$0.015Mistral Large 3non-thinkingOpensame—$0.031Magistral Medium 1thinkingsame——Context: 256K → 40KDevstral Mediumnon-thinkingOpensame—$0.40 / $2Context: 256K → 128KMagistral Small 1thinkingOpensame—$0.70 / $1.16Context: 256K → 128KMistral Small 3non-thinkingOpen-24%—$0.05 / $0.08Context: 256K → 32KAll Mistral models →Run it yourself
Fits onhardware needs not verified yet
Licencenot verified
History
3 versionsBenchmarks
all categoriesChat & Knowledge
AA Omniscience Index-56.8
GPQA Diamond43.4
Coding
Agents & Tool Use
Writing & Language
FAQ
How much does Devstral Small (May '25) cost?›
$0.06 per 1M input tokens and $0.12 per 1M output tokens (cheapest host: DeepInfra).
Is Devstral Small (May '25) free to use?›
There is no free API; the cheapest paid price is $0.06 per 1M input tokens. The weights are open, so you can run it yourself.
How can I use Devstral Small (May '25)?›
Through the Mistral API.
What is the context window of Devstral Small (May '25)?›
256K tokens.
When was Devstral Small (May '25) released? Is there a newer version?›
Released 2025-05-21. It has been superseded by Devstral Small (Jul '25).
Is Devstral Small (May '25) open source?›
Yes, the weights are open.
Can I run Devstral Small (May '25) locally?›
The weights are open and can be self-hosted; hardware needs are not confirmed yet.
What is Devstral Small (May '25)?›
Devstral Small (May '25) is an open-weights language model from Mistral, released May 2025.
What are the best alternatives to Devstral Small (May '25)?›
Qwen3.5 4B (cheaper), GPT-6 Luna (better), MiMo v2.6 Pro (open weights).