DeepSeek model product
DeepSeek-V3.2
2 in non-thinking mode.
Updated Aug 12, 2026. Default version: DeepSeek-V3.2-Speciale
LLMBoard score58.1DeepSeek-V3.2-Speciale
Coverage60%8 benchmark families
Context window131.1KTokens
Official input price$0.14DeepSeek API
DeepSeek-V3.2 Specifications
Technical details for the model's default version.
- Version
- DeepSeek-V3.2-Speciale
- Released
- Dec 1, 2025
- Knowledge cutoff
- Unknown
- Parameters
- 685B
- Context window
- 131.1K
- Max output
- 131.1K
- Inputs
- text
- Outputs
- text
- Open weights
- No
- License
- MIT
DeepSeek-V3.2 Capability Profile
This profile uses the model's current scored version. Arena ratings and prices are shown separately.
DeepSeek-V3.2 Benchmark Results
Benchmark scores for DeepSeek-V3.2-Speciale.
| CodeForces | 90.0% | 03 | 16 | 86.7% | C | |
| HMMT 2025 | 99.2% | 03 | 33 | 93.8% | C | |
| t2-bench | 80.3% | 10 | 23 | 59.1% | C | |
| AIME 2025 | 96.0% | 20 | 114 | 83.2% | C | |
| Toolathlon | 35.2% | 30 | 31 | 3.3% | C | |
| Terminal-Bench 2.0 | 46.4% | 40 | 49 | 18.8% | C | |
| Humanity's Last Exam | 30.6% | 43 | 93 | 54.4% | C | |
| SWE-Bench Verified | 73.1% | 49 | 105 | 53.9% | C |
DeepSeek-V3.2 Arena Results
Preference and agent-evaluation results for the default version.
No Arena results
The default version does not have a matching Arena result yet.
DeepSeek-V3.2 Runtime Performance
Provider-specific output speed and catalog latency for DeepSeek-V3.2-Speciale. Runtime does not affect the capability score.
| DeepSeek | 30 tok/s | 1 s | 131.1K | 131.1K |
Output Speed is generated output tokens received per second. Catalog latency is reported separately from observed provider TTFT.
DeepSeek-V3.2 Pricing
Official vendor API pricing appears first, followed by individual provider offers.
- Official API
- $0.14 input, $0.28 output per 1M
- Official provider
- DeepSeek
- Lowest third-party
- From $0.58 input, $1.68 output per 1M via Azure Cognitive Services
- Tracked offerings
- 3
| Azure Cognitive Services | deepseek-v3.2-speciale | global | $0.58 | $1.68 | 128K | |
| Azure | deepseek-v3.2-speciale | global | $0.58 | $1.68 | 128K |
Official prices use only the vendor's configured official Provider and positive standard USD PAYG rates. Third-party offers remain explicitly labeled.
DeepSeek-V3.2 Versions
Available versions of this model. The score column identifies the version used in the overall ranking.
| DeepSeek-V3.2 (Non-thinking) | N/A | 685B | 1M | 384K | Yes | MIT | |
| DeepSeek-V3.2 (Thinking) | N/A | 685B | 1M | 384K | Yes | MIT | |
| DeepSeek-V3.2 | N/A | 685B | 128K | 64K | Yes | MIT | |
| DeepSeek-V3.2-Speciale | 58.1 | 685B | 131.1K | 131.1K | No | MIT | |
| DeepSeek-V3.2-Exp | N/A | 685B | 163.8K | 65.5K | No | MIT |
What is DeepSeek-V3.2?
Key information about DeepSeek-V3.2 and its available data.
2 in non-thinking mode. A powerful language model with 685B parameters using DeepSeek Sparse Attention (DSA) for efficient long-context processing. Supports JSON output, tool calls, and chat prefix completion. This is the standard chat mode without extended reasoning.
Data as of 2026-08-11.
DeepSeek-V3.2 vs nearby models
Open a comparison with the three ranked models immediately above and below this model.
Models similar to DeepSeek-V3.2
Recommendations prioritize the same model type and family, then the closest LLMBoard score.
DE
DeepSeek-R1
DeepSeek
DE
DeepSeek-V4-Pro
DeepSeek
DE
DeepSeek-V3.1
DeepSeek
DE
DeepSeek-V4-Flash
DeepSeek
DE
DeepSeek-V3
DeepSeek
DE
DeepSeek-R1-Distill-Qwen
DeepSeek
FAQ
Common questions about DeepSeek-V3.2.
When was DeepSeek-V3.2 released?
DeepSeek-V3.2's default version was released on Dec 1, 2025.
How much does DeepSeek-V3.2 cost?
DeepSeek-V3.2's official API price is $0.14 per million input tokens and $0.28 per million output tokens via DeepSeek. The lowest tracked third-party offer starts at $0.58 input and $1.68 output via Azure Cognitive Services.
Who created DeepSeek-V3.2?
DeepSeek-V3.2 was created by DeepSeek.
What is the context window for DeepSeek-V3.2?
The default version has a 131.1K token context window.
Is DeepSeek-V3.2 open weight?
No. The default version is not marked as having publicly available weights.
How many API providers offer DeepSeek-V3.2?
3 provider offerings are linked to the default version.
What models should I compare DeepSeek-V3.2 with?
Nearby ranked alternatives include ERNIE 5.0, Muse Glimmer 30B, MiniMax M2.7.