DeepSeek V3.2 Exp
Sep 2025
Older versionReplace withDeepSeek V4 Pro 0813Newer version
MiniMax M3.1 FlashCheaper, as good
GPT-6 LunaBetter
MiMo v2.6 ProOpen weights
DeepSeek V4 Pro 0813Newer versionMiniMax M3.1 FlashCheaper, as goodGPT-6 LunaBetterMiMo v2.6 ProOpen weightsDeepSeek V3.2 ExpCompare inon
Benchmarks of DeepSeek V3.2 Exp40
BenchmarkByScoreRank
Overall1
Knowledge4
Math7
Science2
Coding3
Writing16
Documents1
Legacy · saturated or no longer reported6
No benchmark matches
Rankingin non-thinking mode?
Retired models are left out of the current rankings.
Alternatives?
ModelScoreCostPer taskWhat changes
CheaperClaude Haiku 5.5max×3.1-37%$0.21Remove sampling parameters+5BetterGPT-6 Lunalow+55%-37%$0.005Function calling only with none · Chat Completions+3OpenMiMo v2.6 Prothinking×3.3+73%$0.13Pass back reasoning_content+3CheaperMiniMax M3.1 Flashthinking×3.1-17%$0.22Complete model response must be appended
CheaperGLM 5.3 FlashmaxOpen×3-25%$0.28Defaults to max · Self-hosted+4ModelScoreCostPer taskWhat changes
Claude Haiku 5.5max×3.1-37%$0.21Remove sampling parameters+5MiniMax M3.1 Flashthinking×3.1-17%$0.22Complete model response must be appendedGLM 5.3 FlashmaxOpen×3-25%$0.28Defaults to max · Self-hosted+4
Qwen3.8 FlashthinkingOpen×2.9-27%$0.37max_tokens range · QwenCloud+5GPT-6 Lunamax×2.7-37%$0.068Function calling only with none · Chat Completions+3ModelScoreCostPer taskWhat changes
GPT-6 Lunalow+55%-37%$0.005Function calling only with none · Chat Completions+3GPT-5.6 Lunalow+51%+43%$0.01MiMo v2.5thinkingOpen+57%-44%$0.019Claude Haiku 5.5low×2.1-37%$0.024Remove sampling parameters+5GPT-5 Minihigh+21%+118%$0.053ModelScoreCostPer taskWhat changes
MiMo v2.6 ProthinkingOpen×3.3+73%$0.13Pass back reasoning_content+3DeepSeek V4.1 Flash 1017maxOpen×3.2+67%$0.26Pass back reasoning_content · Chat Completions+4GLM 5.3maxOpen×3.2+583%$2.01Reasoning cannot be disabled+3
Step 5 PreviewthinkingOpen×3.2+352%$1.03web_search not available · Chat Completions+2
Kimi K3maxOpen×3.1+1,805%$2.00Sampling parameters are fixed+3ModelScoreCostPer taskWhat changes
DeepSeek V4.1 Flash 1017maxOpen×3.2—$0.26Pass back reasoning_content · Chat Completions+4DeepSeek V4 Pro 0813non-thinkingOpen+47%—$0.36Pass back reasoning_content · Chat Completions+4DeepSeek V3.2 SpecialethinkingOpensame——DeepSeek V3.1 Terminusnon-thinkingOpensame—$0.27 / $0.95DeepSeek V3.1non-thinkingOpensame—$0.25 / $0.95R1 0528thinkingOpensame—$0.39 / $1.63DeepSeek V3 0324non-thinkingOpen-30%—$0.021DeepSeek V3non-thinkingOpen-39%—$0.02All DeepSeek models →Run it yourself
Fits onhardware needs not verified yet
Licencenot verified
Benchmarks
all categoriesChat & Knowledge
AA Omniscience Index-48.2
GPQA Diamond73.8
Coding
Agents & Tool Use
τ²-bench · AA33.9
Writing & Language
FAQ
How much does DeepSeek V3.2 Exp cost?›
$0.28 per 1M input tokens and $0.42 per 1M output tokens.
Is DeepSeek V3.2 Exp free to use?›
There is no free API; the cheapest paid price is $0.28 per 1M input tokens. The weights are open, so you can run it yourself.
How can I use DeepSeek V3.2 Exp?›
Through the DeepSeek API.
What is the context window of DeepSeek V3.2 Exp?›
128K tokens.
When was DeepSeek V3.2 Exp released? Is there a newer version?›
Released 2025-09-29. It has been superseded by DeepSeek V3.2.
Is DeepSeek V3.2 Exp open source?›
Yes, the weights are open.
Can I run DeepSeek V3.2 Exp locally?›
The weights are open and can be self-hosted; hardware needs are not confirmed yet.
What is DeepSeek V3.2 Exp?›
DeepSeek V3.2 Exp is an open-weights language model from DeepSeek, released Sep 2025.
What are the best alternatives to DeepSeek V3.2 Exp?›
Claude Haiku 5.5 (cheaper), GPT-6 Luna (better), MiMo v2.6 Pro (open weights).