DeepSeek

DeepSeek V3.2 Exp

Generally availableOpen weights164K contextReasoning

Summary

DeepSeek V3.2 Exp is an open-weight language model from DeepSeek, released on 29 Sept 2025. It accepts text and generates text, with a 164K-token context window and up to 147K output tokens per response. As of 18 Nov 2026, the lowest listed API price is $0.27 per million input tokens and $0.41 per million output tokens (OpenRouter), across 1 provider we track.

DeepSeek-V3.2-Exp is an experimental large language model released by DeepSeek as an intermediate step between V3.1 and future architectures. It introduces DeepSeek Sparse Attention (DSA), a fine-grained sparse attention mechanism...

Last verified 18 Nov 2026. Every figure on this page links to its source.

Specifications

DeveloperDeepSeek
Release date29 Sept 2025 [OpenRouter]
StatusGenerally available
WeightsOpen weights [Hugging Face]
Licensemit [Hugging Face]
Parameters685B [Hugging Face]
Context window164K (163,840 tokens) [OpenRouter]
Max output147K tokens [OpenRouter]
Knowledge cutoff31 Jul 2025 [OpenRouter]
InputText [OpenRouter]
OutputText [OpenRouter]
Reasoning modeYes [OpenRouter]
Tool callingYes [OpenRouter]
Structured outputYes [OpenRouter]
Hugging Facedeepseek-ai/DeepSeek-V3.2-Exp [Hugging Face]

DeepSeek V3.2 Exp API pricing by provider

USD per million tokens. Sorted by input price.

ProviderInputOutputCached inputContextSource
OpenRouter
deepseek/deepseek-v3.2-exp
$0.27$0.41–164KOpenRouter, 18 Nov 2026

Running DeepSeek V3.2 Exp locally

Estimated memory, from the parameter count and the architecture in config.json on Hugging Face. Includes about 0.8 GB of runtime overhead.

QuantizationWeightsTotal, 8K contextTotal, 32K context
Q4_K_M (4-bit)416 GB417 GB419 GB
Q8_0 (8-bit)728 GB730 GB731 GB
FP16 / BF16 (16-bit)1371 GB1372 GB1374 GB

At 4-bit with an 8K conversation, DeepSeek V3.2 Exp 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 Exp?

DeepSeek V3.2 Exp has a context window of 163,840 tokens (164K) and can generate up to 147,456 tokens in a single response, according to OpenRouter.

How much does the DeepSeek V3.2 Exp API cost?

DeepSeek V3.2 Exp is listed from $0.27 per million input tokens and $0.41 per million output tokens (OpenRouter). Prices last verified 18 Nov 2026.

Is DeepSeek V3.2 Exp open source?

DeepSeek V3.2 Exp is an open-weight model: its weights are published on Hugging Face as deepseek-ai/DeepSeek-V3.2-Exp under the mit license. Check the license terms before commercial use.

When was DeepSeek V3.2 Exp released?

DeepSeek released DeepSeek V3.2 Exp on 29 Sept 2025, according to OpenRouter.

How much VRAM does DeepSeek V3.2 Exp 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.

ModelLabReleasedContextInput $/1MOutput $/1MWeights
Qwen3 VL 30B A3B InstructAlibaba (Qwen)4 Oct 2025131K$0.15$0.60Open
Cydonia 24B V4.1Thedrummer27 Sept 2025131K$0.30$0.50Open
GPT OSS Safeguard 120BOpenAI29 Oct 2025128K$0.15$0.60Open
Nvidia Nemotron Nano 12B V2 VLNVIDIA28 Oct 2025131K$0.20$0.60Open
Gemma Sea Lion V4 27B ItAisingapore23 Sept 2025128K$0.35$0.56Open
Qwen3 VL 8B InstructAlibaba (Qwen)14 Oct 2025262K$0.25$0.75Open

Change history

No changes detected since we started tracking this model. We re-check every source daily.

Sources