What is DeepSeek V4 Pro?
DeepSeek's 1.6T-parameter open MoE model for coding and agents, with a 1M-token context.
Who is DeepSeek V4 Pro best for?
DeepSeek V4 Pro is best for developers who want an open model for coding agents.
Hard coding problems
In its Think Max mode DeepSeek reports 93.5 on LiveCodeBench and a 3206 Codeforces rating, the top scores in its own comparison table.
Very long documents and codebases
It reads up to 1M tokens, and at that length it needs about a tenth of the KV cache that DeepSeek-V3.2 used.
Teams that host their own models
The weights are MIT-licensed on Hugging Face and ModelScope, with 1.6T total and 49B active parameters per token.
Batch work outside peak hours
API prices drop by half outside DeepSeek's weekday peak windows, including all weekend.
How much does DeepSeek V4 Pro cost?
DeepSeek V4 Pro is free to download under the MIT license, and DeepSeek's API charges $1.32 per million input tokens and $3.96 per million output tokens at peak hours, half that off-peak.
- Download from Hugging Face or ModelScope
- 1.6T parameters, 49B active per token
- Base and instruct versions
- Cached input $0.022 per million tokens
- All hours outside peak, including weekends
- Model name deepseek-v4-pro
- Cached input $0.044 per million tokens
- Weekdays 01:00 to 04:00 and 06:00 to 10:00 UTC
- Up to 384K output tokens
Off-peak API rates are half the peak rates, and Chinese public holidays count as off-peak. The API serves the general-release version of V4 Pro, supports thinking and non-thinking modes, takes text only (no image input) and allows 500 concurrent requests. Charges come out of a prepaid balance. Prices are in US dollars. Full pricing on api-docs.deepseek.com. We checked these prices on .
How do readers rate DeepSeek V4 Pro?
No one has rated DeepSeek V4 Pro yet. Be the first.
Sign in to rate. One rating per account.
Is DeepSeek V4 Pro open source?
Yes. DeepSeek V4 Pro's code is public under the MIT licence, so you can run it on your own server.
What are the best DeepSeek V4 Pro alternatives?
DeepSeek V4 Pro is an open source alternative to Kimi K2.6, Muse Spark Open, Claude Opus 5.5 and 2 more. MiMo-V2.6-Pro does the same job and is open source too.
Moonshot AI's open-weights model for long coding jobs and agent swarms of up to 300 sub-agents.
Meta's 816B open-weight model for coding and long agent tasks, free to download and run yourself.
Anthropic's default model for coding and agents, at $4 and $20 per million tokens.
Google's advanced reasoning model for coding, research and agents.
OpenAI's agentic model for coding and professional work, with a Pro tier.
Xiaomi's 1T open omnimodal model for coding and agents, with a 1M context.
Questions about DeepSeek V4 Pro
What is the difference between DeepSeek V4 Pro and V4 Flash?
V4 Pro is the larger model, with 1.6T parameters and 49B active per token, against 284B and 13B for V4 Flash. Both read 1M tokens. DeepSeek says Flash can reason about as well given more thinking time, but trails Pro on knowledge questions and the most complex agent work.
Who is DeepSeek V4 Pro for?
DeepSeek V4 Pro is best for developers who want an open model for coding agents.
Is DeepSeek V4 Pro free?
Yes. DeepSeek V4 Pro has a free Open weights plan: download from Hugging Face or ModelScope.
How much does DeepSeek V4 Pro cost?
DeepSeek V4 Pro's paid plans are API, off-peak at $0.66 / $1.98 per million input / output tokens and API, peak at $1.32 / $3.96 per million input / output tokens.
Is DeepSeek V4 Pro open source?
Yes. DeepSeek V4 Pro's code is public under the MIT licence, so you can run it on your own server.
What are the best DeepSeek V4 Pro alternatives?
The closest alternatives on Recaply are Kimi K2.6, Muse Spark Open, Claude Opus 5.5, Gemini 3.1 Pro, GPT-5.5 and MiMo-V2.6-Pro.
Where can I try DeepSeek V4 Pro?
On its official website, huggingface.co.







