Model introduction
DeepSeek V4-Pro is the Pro model in DeepSeek's V4 family. DeepSeek's update notes confirm access through both OpenAI Chat Completions and Anthropic interfaces.
Model capabilities
Specifications come from the cited model material. Catalog tags help identify access features; use the source documentation for exact limits.
Verified specifications
- Official model parameter
- deepseek-v4-pro
- Verified interfaces
- OpenAI Chat Completions; Anthropic
- Official announcement date
- 2026-04-24
Sources: DeepSeek models and pricing documentation
Independent capability evaluation
Only public measurements matched to the exact model ID and reasoning profile are shown; nearby variants are not substituted.
Intelligence index
43
Cohort rank #5 / 97
Output speed
67.8 tok/s
Cohort rank #23 / 97
First-token latency
1.64s
Reasoning, high effort
Ranks near the top of its large open-weight cohort with low first-token latency and a strong capability-cost balance.
Evaluation profile: Reasoning, high effort. Artificial Analysis · Methodology
Model pricing and access
Prices are read directly from the MoleAPI console catalog and shown by current billing group and context tier.
Your final charge follows the account group shown in the console.
Live pricing source
Pricing table
USD / 1M tokensStandardx1defaultInput $0.67 · Output $1.34
Input
$0.67 / 1M tokens
Output
$1.34 / 1M tokens
Cache read
$0.02 / 1M tokens
Cache write
$0.83 / 1M tokens
Discountx0.8discountInput $0.536 · Output $1.072
Input
$0.536 / 1M tokens
Output
$1.072 / 1M tokens
Cache read
$0.016 / 1M tokens
Cache write
$0.664 / 1M tokens
Relayx0.3relayInput $0.201 · Output $0.402
Input
$0.201 / 1M tokens
Output
$0.402 / 1M tokens
Cache read
$0.006 / 1M tokens
Cache write
$0.249 / 1M tokens
Temporary (availability may vary)x0.1tempInput $0.067 · Output $0.134
Input
$0.067 / 1M tokens
Output
$0.134 / 1M tokens
Cache read
$0.002 / 1M tokens
Cache write
$0.083 / 1M tokens
Access protocols
Available billing groups: Standard, Discount, Relay, Temporary (availability may vary)
- openai
- POST
- /v1/chat/completions
- openai-response
- POST
- /v1/responses
- anthropic
- POST
- /v1/messages
- gemini
- POST
- /v1beta/models/{model}:generateContent
- openai-response-compact
- POST
- /v1/responses/compact
- openai-alpha-search
- POST
- /v1/alpha/search
Complete model guide
Prepared from vendor sources and independently reviewed. · 2026-07-28
DeepSeek-V4-Pro model guide
DeepSeek-V4-Pro is DeepSeek’s preview flagship model for agentic coding, complex reasoning, world knowledge, mathematics, STEM, and coding workloads. It is an open-weight model with selectable thinking and non-thinking modes. [4][1]
Positioning and core capabilities
DeepSeek positions V4-Pro as the higher-capability model in its V4 Preview family, with agentic capabilities and reasoning-oriented performance. [4] The model was announced as part of the live, open-sourced DeepSeek-V4 Preview on April 24, 2026. [4]
V4-Pro supports text input and text output, structured JSON output, tool calls, chat-prefix completion, and fill-in-the-middle completion. [1] Thinking mode is the default, while non-thinking mode is also available. [1]
Capabilities and specs
DeepSeek-V4-Pro combines a long context window with a large maximum output allowance for extended reasoning and generation tasks. [1]
| Specification | Value |
|---|---|
| Total parameters | 1.6T [4] |
| Active parameters | 49B [4] |
| Context window | 1M tokens [1] |
| Maximum output | 384K tokens [1] |
| Modes | Thinking and non-thinking; thinking is the default [1] |
| Structured output | JSON output [1] |
| Tool use | Supported [1] |
| Completion features | Chat-prefix completion and fill-in-the-middle completion, both beta features [1] |
DeepSeek documents compatibility with OpenAI-format and Anthropic-format API interfaces for V4-Pro. [1]
Use cases
DeepSeek-V4-Pro is positioned for coding agents, complex software tasks, mathematical reasoning, STEM problem solving, and knowledge-intensive work. [4] Its 1M-token context window also supports workloads that require large source collections, lengthy documents, or extended conversational state. [1]
Tool calling and structured JSON output support agentic workflows that combine model responses with external systems or structured application pipelines. [1] Chat-prefix and fill-in-the-middle completion features also support code-completion-oriented generation flows. [1]
Material limitations
Fill-in-the-middle completion is available only in non-thinking mode. [1] Chat-prefix completion and fill-in-the-middle completion are documented as beta features. [1]
DeepSeek applies a default account-level concurrency limit of 500 simultaneous V4-Pro requests; requests exceeding that limit receive HTTP 429 responses. [5]
Sources
Code examples
These examples use MoleAPI's Responses API endpoint and run after you replace the API key.
curl https://api.moleapi.com/v1/responses \
-H "Authorization: Bearer $MOLEAPI_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "deepseek-v4-pro",
"input": "Explain this problem step by step."
}'Frequently asked questions
How is deepseek-v4-pro priced?
This page reads prices from the MoleAPI console API and updates with model prices, context tiers, and account groups.
How can I access deepseek-v4-pro?
Protocols and endpoints come from the supported_endpoint_types field in the MoleAPI model catalog.
How do I switch an existing project to deepseek-v4-pro?
Keep the MoleAPI API address and key, replace the model parameter with the model ID on this page, then check protocol-specific parameter differences.
Where do the deepseek-v4-pro model details come from?
Capabilities and limitations are checked against DeepSeek and the other cited pages. Pricing and available protocols come only from the MoleAPI console.