Devstral Small 2
Dec 2025
Retired 2026-03-31Replace withMistral Medium 3.5Newer version
GPT-6 LunaBetter
Qwen3.8 27BOpen weights
Mistral Medium 3.5Newer versionGPT-6 LunaBetterQwen3.8 27BOpen weightsDevstral Small 2Compare inon
Benchmarks of Devstral Small 228
BenchmarkByScoreRank
Overall1
Knowledge4
Science3
Coding3
Agents4
Multimodal1
Documents2
Legacy · saturated or no longer reported10
No benchmark matches
Rankingin non-thinking mode?
Retired models are left out of the current rankings.
Alternatives?
ModelScoreCostPer taskWhat changes
BetterGPT-6 Lunalow×2.9—$0.005Function calling only with none · Chat Completions+3OpenQwen3.8 27Bxhigh×4.5—$1.01Memory (4-bit): 14 GB → 17 GB+5BetterGPT-5.6 Lunalow×2.8—$0.01BetterMistral Small 4thinkingOpen+49%—$0.015Memory (4-bit): 14 GB → 72 GB
BetterMiMo v2.5thinkingOpen×2.9—$0.019Memory (4-bit): 14 GB → 186 GBModelScoreCostPer taskWhat changes
Nothing here yet.
ModelScoreCostPer taskWhat changes
GPT-6 Lunalow×2.9—$0.005Function calling only with none · Chat Completions+3GPT-5.6 Lunalow×2.8—$0.01Mistral Small 4thinkingOpen+49%—$0.015Memory (4-bit): 14 GB → 72 GBMiMo v2.5thinkingOpen×2.9—$0.019Memory (4-bit): 14 GB → 186 GB
DeepSeek V3 0324non-thinkingOpen+29%—$0.021Context: 256K → 128K+2ModelScoreCostPer taskWhat changes
Qwen3.8 27BxhighOpen×4.5—$1.01Memory (4-bit): 14 GB → 17 GB+5IK2 Horizon MoVA 36B A4BthinkingOpen×3.4—$1 / $4Memory (4-bit): 14 GB → 22 GB+1Qwen 3.5 27BthinkingOpen×3—$0.30 / $2.40Memory (4-bit): 14 GB → 17 GBAG9v3 39A5BOpen×2.9——Context: 256K → 131K+2Qwen3.6 27BthinkingOpen×2.8—$0.62Memory (4-bit): 14 GB → 17 GBModelScoreCostPer taskWhat changes
Mistral Large 4thinking×5.1—$1.13Infinite whitespace stream+4Mistral Medium 3.5highOpen+88%—$0.50Memory (4-bit): 14 GB → 77 GBMistral Small 4thinkingOpen+49%—$0.015Memory (4-bit): 14 GB → 72 GBMistral Large 3non-thinkingOpen+23%—$0.031Memory (4-bit): not verifiedMagistral Medium 1thinking+20%——Context: 256K → 40K+1Devstral Mediumnon-thinkingOpen+19%—$0.40 / $2Context: 256K → 128K+2Magistral Small 1thinkingOpensame—$0.70 / $1.16Context: 256K → 128K+1Mistral Small 3non-thinkingOpensame—$0.05 / $0.08Context: 256K → 32K+1All Mistral models →Run it yourself
estimateFits on✗ 8 GB (phone, laptop)✓ 16 GB · compressed✓ 24 GB GPU · compressed✓ 48 GB GPU · 8-bit✓ 80 GB GPU · full quality
LicenceApache-2.0free for commercial use
History
3 versionsBenchmarks
all categoriesChat & Knowledge
AA Omniscience Index-57.7
GPQA Diamond53.2
Coding
SciCode32.4
Agents & Tool Use
τ³-Banking · AA10.7
τ²-bench · AA23.4
Writing & Language
Vision & Documents
FAQ
How much does Devstral Small 2 cost?›
$0 per 1M input tokens and $0 per 1M output tokens (cheapest host: Mistral).
Is Devstral Small 2 free to use?›
Yes, there is a free API option.
How can I use Devstral Small 2?›
Through the API of Mistral.
What is the context window of Devstral Small 2?›
256K tokens, with up to 256K output tokens.
When was Devstral Small 2 released? Is there a newer version?›
Released 2025-12-09. It has been superseded by Mistral Medium 3.5.
Is Devstral Small 2 open source?›
Yes, the weights are open (Apache-2.0 licence).
Can I run Devstral Small 2 locally?›
The weights are open and can be self-hosted; hardware needs are not confirmed yet.
Can Devstral Small 2 read images?›
It accepts text and images as input.
What is Devstral Small 2?›
Devstral Small 2 is an open-weights language model from Mistral, released Dec 2025, the successor to Devstral Small (Jul '25).
What are the best alternatives to Devstral Small 2?›
GPT-6 Luna (better), Qwen3.8 27B (open weights).