Skip to content
RankShark

DeepSeek V3.2

DeepSeek-V3.2 is a large language model designed to harmonize high computational efficiency with strong reasoning and agentic tool-use perfo

ModelOpen sourceby DeepSeeksince 2025
43.2

SharkScore

#83 in Models

Why it scores 43.2

DeepSeek V3.2 has a SharkScore of 43.2/100, ranked #83 of 608 in Models (updated 2026-11-19).

Each part compares DeepSeek V3.2 with other listings in Models. Evidence covers 87% of the score's weight.

Adoption
50 · 30%
User reviews
no data · 25%
Maintenance
2 · 20%
Trust & transparency
80 · 15%
Momentum
no data · 10%

Paying never changes this score. How ranking works · method v2026-11-18.2

RankShark review of DeepSeek V3.2

Automated evaluation by RankShark · 18 Nov 2026 · method v1.0 · How we review

We were not paid for this review.

As of 2026-11-18, DeepSeek V3.2 is a text model with a 163,840-token context window and OpenRouter rates of $0.26 per million input tokens and $0.38 per million output tokens. [E2]

DeepSeek-V3.2 is a text-in/text-out language model. Its model card describes reasoning and tool-use capabilities, and provides instructions for local deployment. [E1][E2]

Best for

  • People looking for a model whose vendor describes it as designed for reasoning and agentic tool use. [E1] · model card
  • Applications that need a long context window; the OpenRouter listing reports 163,840 tokens. [E2] · OpenRouter
  • People who want to explore local deployment; the model card points to a repository with local-running information. [E1] · model card

Not for

  • API workflows that send messages with the developer role; the model card says the official API does not accept that role. [E1] · model card

Key features

  • Sparse attention. The vendor says DeepSeek Sparse Attention is designed for long-context scenarios and reduces computational complexity while preserving model performance. [E1] · model card
  • Tool-use chat format. The vendor says the chat template includes revised tool calling and a “thinking with tools” capability. [E1] · model card
  • Encoding examples. The vendor provides Python scripts and test cases for encoding OpenAI-compatible messages and parsing model output. [E1] · model card
  • Local deployment. The model card points to the DeepSeek-V3.2-Exp repository for information about running the model locally. [E1] · model card
  • License. The model card states that the repository and model weights are licensed under MIT. [E1] · model card

Pros

  • The OpenRouter listing gives per-token rates and reports a 163,840-token context window. [E2] · OpenRouter
  • The vendor provides code and test cases for the model’s chat encoding and output parsing. [E1] · model card
  • The model card states that the model weights and repository are MIT-licensed. [E1] · model card

Cons

  • The release does not include a Jinja-format chat template. [E1] · model card
  • The included output parser handles well-formatted strings only, according to the vendor. [E1] · model card
  • The official API does not accept messages assigned the developer role, according to the model card. [E1] · model card
  • No free plan or free trial stated on the pages we checked.
  • Whether customer data is used for model training is not stated on the pages we checked.

Pricing

PlanPriceLimits
OpenRouter token usage$0.26 per million input tokens; $0.38 per million output tokensContext window: 163,840 tokens. [E2]

Checked 18 Nov 2026

Free plan
Not stated (checked 18 Nov 2026)
Free trial
Not stated (checked 18 Nov 2026)
Trains on customer data
Not stated (checked 18 Nov 2026)
Data retention
Not stated (checked 18 Nov 2026)
Data regions
Not stated (checked 18 Nov 2026)
Data processing agreement
Not stated (checked 18 Nov 2026)
AI models used
DeepSeek-V3.2 · model card, OpenRouter
API
The OpenRouter listing includes chat completions, completions, messages, and responses endpoints. [E2] · OpenRouter
SDKs
Not stated (checked 18 Nov 2026)
Webhooks
Not stated (checked 18 Nov 2026)
Platforms
The model is listed on Hugging Face; the model card includes a Transformers example and points to a repository for local deployment information. [E1] · model card
Security claims
No security certification stated (checked 18 Nov 2026)

What we measured

Context window (OpenRouter listing) · now
163,840 tokens
Price per million tokens (OpenRouter listing) · now
$0.26 in, $0.38 out
Package downloads · 30 days
9,558,338
How we tested · sources
  • Hourly uptime and response-time checks of the product's URL.
  • Repository, release and download data from GitHub, npm, PyPI and Hugging Face where they exist.
  • We did not sign up, pay, or use the product as a customer.
  • 18 Nov 2026: Plans or prices changed on the pricing page.
  • 18 Nov 2026: Plans or prices changed on the pricing page.

Something wrong? Report an error with a public link that shows it; corrections re-run the review.

Top alternatives to DeepSeek V3.2

See all
  • 67.1

    Qwen3.8 27B is an open-weight dense vision-language model from Qwen.

    ModelOpen source
  • 65.5

    Qwen3.6-35B-A3B is an open-weight multimodal model from Alibaba Cloud with 35 billion total parameters and 3 billion active parameters per t

    ModelOpen source
  • 64.7

    Qwen3.8 Flash is a multimodal reasoning model from Alibaba.

    ModelOpen source
  • 64.3

    Qwen3.6 27B is a dense 27-billion-parameter language model from the Qwen Team at Alibaba, released in April 2026.

    ModelOpen source