Qwen3.8 27B is an open-weight dense vision-language model from Qwen.
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
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
| Plan | Price | Limits |
|---|---|---|
| OpenRouter token usage | $0.26 per million input tokens; $0.38 per million output tokens | Context 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.
- model card: https://huggingface.co/deepseek-ai/DeepSeek-V3.2 · fetched 18 Nov 2026
- OpenRouter: https://openrouter.ai/deepseek/deepseek-v3.2 · fetched 18 Nov 2026
- 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- 67.1ModelOpen 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