DeepSeek V3.2
Summary
DeepSeek V3.2 is an open-weight language model from DeepSeek, released on 1 Dec 2025. It accepts text and generates text, with a 164K-token context window and up to 82K output tokens per response. As of 18 Nov 2026, the lowest listed API price is $0.26 per million input tokens and $0.38 per million output tokens (DeepInfra), across 7 providers we track.
Hybrid-reasoning DeepSeek model with thinking and non-thinking modes, sparse attention, and tool-use
Last verified 18 Nov 2026. Every figure on this page links to its source.
Specifications
| Developer | DeepSeek |
|---|---|
| Release date | 1 Dec 2025 [models.dev] |
| Status | Generally available [models.dev] |
| Weights | Open weights [Hugging Face] |
| License | mit [Hugging Face] |
| Parameters | 685B [Hugging Face] |
| Context window | 164K (163,840 tokens) [models.dev] |
| Max output | 82K tokens [models.dev] |
| Knowledge cutoff | Jul 2024 [models.dev] |
| Input | Text [models.dev] |
| Output | Text [models.dev] |
| Reasoning mode | Yes [models.dev] |
| Tool calling | Yes [models.dev] |
| Structured output | Yes [models.dev] |
| Hugging Face | deepseek-ai/DeepSeek-V3.2 [Hugging Face] |
DeepSeek V3.2 API pricing by provider
USD per million tokens. Sorted by input price.
| Provider | Input | Output | Cached input | Context | Source |
|---|---|---|---|---|---|
| DeepInfra deepseek-ai/DeepSeek-V3.2 | $0.26 | $0.38 | $0.13 | 164K | models.dev, 18 Nov 2026 |
| OpenRouter deepseek/deepseek-v3.2 | $0.26 | $0.38 | $0.13 | 164K | OpenRouter, 18 Nov 2026 |
| SiliconFlow deepseek-ai/DeepSeek-V3.2 | $0.27 | $0.42 | $0.14 | 164K | models.dev, 18 Nov 2026 |
| Hugging Face Inference Providers deepseek-ai/DeepSeek-V3.2 | $0.28 | $0.40 | – | 164K | models.dev, 18 Nov 2026 |
| Azure AI Foundry deepseek-v3.2 | $0.58 | $1.68 | – | 128K | models.dev, 18 Nov 2026 |
| Amazon Bedrock deepseek.v3.2 | $0.62 | $1.85 | – | 164K | models.dev, 18 Nov 2026 |
| Vercel AI Gateway deepseek/deepseek-v3.2-thinking | $0.62 | $1.85 | – | 128K | models.dev, 18 Nov 2026 |
| Vercel AI Gateway deepseek/deepseek-v3.2 | $0.62 | $1.85 | – | 128K | models.dev, 18 Nov 2026 |
Running DeepSeek V3.2 locally
Estimated memory, from the parameter count and the architecture in config.json on Hugging Face. Includes about 0.8 GB of runtime overhead.
| Quantization | Weights | Total, 8K context | Total, 32K context |
|---|---|---|---|
| Q4_K_M (4-bit) | 415 GB | 417 GB | 419 GB |
| Q8_0 (8-bit) | 728 GB | 730 GB | 731 GB |
| FP16 / BF16 (16-bit) | 1371 GB | 1372 GB | 1374 GB |
At 4-bit with an 8K conversation, DeepSeek V3.2 needs about 417 GB, more than any single consumer graphics card; it needs a unified-memory machine, a multi-GPU setup or a split between graphics card and system RAM. Check your own hardware, or see Best computer for local AI on our sister site AIPCs.
Frequently asked questions
What is the context window of DeepSeek V3.2?
DeepSeek V3.2 has a context window of 163,840 tokens (164K) and can generate up to 81,920 tokens in a single response, according to models.dev.
How much does the DeepSeek V3.2 API cost?
DeepSeek V3.2 is listed from $0.26 per million input tokens and $0.38 per million output tokens (DeepInfra). Prices last verified 18 Nov 2026.
Is DeepSeek V3.2 open source?
DeepSeek V3.2 is an open-weight model: its weights are published on Hugging Face as deepseek-ai/DeepSeek-V3.2 under the mit license. Check the license terms before commercial use.
When was DeepSeek V3.2 released?
DeepSeek released DeepSeek V3.2 on 1 Dec 2025, according to models.dev.
Which providers offer DeepSeek V3.2?
We track DeepSeek V3.2 on 7 providers: DeepInfra, OpenRouter, SiliconFlow, Hugging Face Inference Providers, Azure AI Foundry, Amazon Bedrock, Vercel AI Gateway.
How much VRAM does DeepSeek V3.2 need?
About 417 GB at 4-bit (Q4_K_M) with an 8K-token conversation, and about 730 GB at 8-bit. That is more than any single consumer graphics card holds. These are estimates from the model's published parameter count and architecture.
Similar models from other labs
Comparable on output type, weights, release date, price and context window. Prices are the lowest listed, USD per million tokens.
| Model | Lab | Released | Context | Input $/1M | Output $/1M | Weights |
|---|---|---|---|---|---|---|
| GPT OSS Safeguard 120B | OpenAI | 29 Oct 2025 | 128K | $0.15 | $0.60 | Open |
| Nvidia Nemotron Nano 12B V2 VL | NVIDIA | 28 Oct 2025 | 131K | $0.20 | $0.60 | Open |
| Qwen3 VL 30B A3B Instruct | Alibaba (Qwen) | 4 Oct 2025 | 131K | $0.15 | $0.60 | Open |
| Jamba Mini | AI21 Labs | 8 Jan 2026 | 256K | $0.20 | $0.40 | Open |
| GLM-4.7-Flash | Zhipu AI (Z.ai) | 19 Jan 2026 | 200K | $0.06 | $0.40 | Open |
| GLM-4.7-FlashX | Zhipu AI (Z.ai) | 19 Jan 2026 | 200K | $0.06 | $0.40 | Open |
Change history
- Name changed from DeepSeek-V3.2 to DeepSeek V3.2
- Max output changed from 66K to 82K
- Context window changed from 128K to 164K
- Max output changed from 128K to 66K
- Name changed from deepseek-ai/DeepSeek-V3.2 to DeepSeek-V3.2
- Max output changed from 164K to 128K
- Context window changed from 164K to 128K
- Name changed from DeepSeek-V3.2 to deepseek-ai/DeepSeek-V3.2
- Context window changed from 164K to 164K
- Max output changed from 66K to 164K
- Max output changed from 82K to 66K
- Name changed from DeepSeek V3.2 to DeepSeek-V3.2
- Name changed from DeepSeek-V3.2 to DeepSeek V3.2
- Max output changed from 64K to 82K
- Max output changed from 128K to 64K
- Context window changed from 128K to 164K
- Tool calling changed from no to yes
- Name changed from DeepSeek V3.2 to DeepSeek-V3.2
- Max output changed from 8K to 128K
- Reasoning mode changed from no to yes
- Reasoning mode changed from yes to no
- Tool calling changed from yes to no
- Name changed from DeepSeek-V3.2 to DeepSeek V3.2
- Max output changed from 64K to 8K
- Context window changed from 164K to 128K
Sources
- https://openrouter.ai/deepseek/deepseek-v3.2
- https://github.com/sst/models.dev/blob/dev/providers/amazon-bedrock/models/deepseek.v3.2.toml
- https://huggingface.co/deepseek-ai/DeepSeek-V3.2
- https://github.com/sst/models.dev/blob/dev/providers/deepinfra/models/deepseek-ai/DeepSeek-V3.2.toml
- https://github.com/sst/models.dev/blob/dev/providers/siliconflow/models/deepseek-ai/DeepSeek-V3.2.toml
- https://github.com/sst/models.dev/blob/dev/providers/huggingface/models/deepseek-ai/DeepSeek-V3.2.toml
- https://github.com/sst/models.dev/blob/dev/providers/azure/models/deepseek-v3.2.toml
- https://github.com/sst/models.dev/blob/dev/providers/vercel/models/deepseek/deepseek-v3.2-thinking.toml
- https://github.com/sst/models.dev/blob/dev/providers/vercel/models/deepseek/deepseek-v3.2.toml