DeepSeek V3.2
deepseek/deepseek-v3.2- DeepSeek
- Family · DeepSeek V3
- Released 1 Dec 2025
- Open weights
- Text → Text
- Tools · JSON
Not on OSMEXO yet
Spec & list price
Context160K
Max output144K
Input / 1M$0.26
Output / 1M$0.381.5× input
Scores · standing
Intelligence21.5#127/489
Coding44.2#81/196
Usage
Tokens / day38.7B▼ 1.5% 7d
Usage rank#55
Requests / day3.3M
Tool calls / req0.02
Per request · 30 days
Avg input6.4K0.4× median
Avg output4560.7× median
Cache hit55%0.8× median
Reasoning share25%0.4× median
Tool-call failures1.1%0.6× median
Consumption
Peak 58.3B 1 OctLow 24.8B 31 OctRank today #55
News & trends
5 mentions this week ▼ 17%this model · videos, posts, articles · 30 days
Trending · 14 days · DeepSeek V3 family
- What Did DeepSeek V3's $5.6 Million Cost Leave Out? #deepseek #nvidia #ai #mixtureofexperts1.5K views
- Best provider for deepseek v3.2?17 upvotes
- Free Claude Code CLI Setup: Use DeepSeek V3 & Gemini API (1-Click Windows Installer)29 views
News · latest from DeepSeek
- Tell HN: Mistrap Pro includes DeepSeek v4.1 Flash 1017docs.mistral.ai · 1 Nov
- Show HN: DeepSeek v4.1 Flash-1017 Safety Evaluationtwitter.com · 26 Oct
- DeepSeek V4 Pro 0813 vs Claude Fable 5 on DeepSWE: Cost, Coding, and RoutingTogether AI · 17 Aug
- DeepSeek-V4 Flash 0731 vs GPT-5.6 Luna on DeepSWE: Cost and CodingTogether AI · 6 Aug
Benchmarks
bar = share of the top score · Benchmark data: Artificial AnalysisSpeed · median across 9 providers, latest day
Output speed#302/33326 tok/s2.6× slower than median
Fastest provider59 tok/sbest of 9 providers
First token#2582.11 s2.1× longer than median
#4 cheapest of 29 with a similar score
Value
Intelligence Index
Blended price per 1M tokens (log)
Family progress
Intelligence Index
- Family releases21.5
Technical details
- Tokenizer
- DeepSeek
- License
- MIT
- Parameters
- 685B
- Moderation
- none
- Downloads · likes
- 6.4M · 1.6K
Request parameters 19
- tools
- tool_choice
- response_format
- structured_outputs
- reasoning
- include_reasoning
- temperature
- top_p
- top_k
- min_p
+9 more
- max_tokens
- stop
- seed
- frequency_penalty
- logit_bias
- logprobs
- presence_penalty
- repetition_penalty
- top_logprobs
Interesting facts
- The most used of 5 DeepSeek V3 models by tokens last month.
- Released 335 days ago, the newest of 5 DeepSeek V3 models.
- 1.5× the output room of similar models (147.5K against a 96.8K median).
- Weekdays average 44.4B tokens a day against 44.8B on weekends.
When to use
Best for
- Coding help and competitive programming
- Math and science reasoning
- Tool-calling agents with structured output
- Open-weight self-hosted assistants
Think twice for
- Image input (text-only)
- Hardest terminal agent tasks
- Very large single responses
Research papers
- Vendster: An Evaluation Arena to Investigate How Forecasting Tool Quality Affects LLM Agent Behaviors in Buyer-Vendor Negotiation28 Oct
- Interpreting LLM Scores in Peer Review: Alignment and Divergence Across Prompts, Domains, and Manuscript Features26 Oct
- From Document Layout to Causal Topology: An End-to-End Architecture for Temporal Causal Reasoning25 Oct
- Multi-Agent Retrieval-Augmented Causality Assessment (MARCA) of Individual Case Safety Reports in Pharmacovigilance23 Oct
- AdaPRA: Process Reward Agents with Adaptive Rubrics for Data Science23 Oct
- HyScale: Scaling Long-Context Training for Learned Sparse Attention23 Oct
- Correlated Failures of AI Approval Monitors21 Oct
- IF-MAP: Influence-Function-Inspired Agent Prioritization in Multi-Agent Systems17 Oct
Same family
DeepSeek V3- 13.7
- 8.5
- 16.6
- 9.7
Alternatives
other vendors · similar capability- 25.3
- 26.5
- 26.3
- 21.8
Want DeepSeek V3.2 on OSMEXO?
Sign in to request it. Requests are reviewed; a request does not guarantee availability.