DeepSeek V4 Pro

DeepSeekMCP Tool CallReasoning

deepseek/deepseek-v4-proOpen Weights

Open MoE flagship with million-token context for coding and long agent runs

Benchmark Pricing$0.66 / 1M in$1.98 / 1M out$0.02 cache read

Context: 1,000,000 tokens•Max Output: 384,000 tokens•Knowledge: 2025-05•Released: 2026-04-24

TECHNICAL SPECIFICATIONS

Context Window1,000,0001M tokens

Max Output384,000384K tokens

LicenseOpen SourceSelf-hostable

Input Modalities

Text

Function / Tool CallingSupported (MCP Ready)

Structured OutputsGuaranteed Schema

MODEL CONTEXT PROTOCOL (MCP) READINESS

✓ DeepSeek V4 Pro natively supports OpenAI / Anthropic format tool calling and is fully compatible with MCP servers and agent loops.

MCP Context Window Recommendations

With a context capacity of 1,000,000 tokens, this model can handle extremely large MCP resource manifests, complete database schemas, and multiple active server tool definitions simultaneously without degradation.

STANDARDIZED BENCHMARKS & EVALS

SWE-Bench Verifiedresolved

80.6

Artificial Analysis Coding Agent Indexaverage pass@1

50.1

SWE-Atlas Codebase QnApass@1

67.8

SWE-Bench Propass@1

Terminal-Benchpass@1

64.7

MMLU-ProEM

87.5

SimpleQA-Verifiedpass@1

57.9

Chinese SimpleQApass@1

84.4

GPQA Diamondpass@1

90.1

Humanity's Last Exampass@1

37.7

LiveCodeBenchpass@1

93.5

Codeforcesrating

3206

HMMTpass@1

95.2

IMOAnswerBenchpass@1

89.8

MathArena Apexpass@1

38.3

MathArena Apex Shortlistpass@1

90.2

MRCRMMR

83.5

CorpusQAaccuracy

Terminal-Benchaccuracy

67.9

SWE-Bench Proresolved

55.4

SWE-Bench Multilingualresolved

76.2

BrowseComppass@1

83.4

Humanity's Last Exampass@1

48.2

MCP Atlaspass@1

73.6

GDPval-AAElo

1554

Toolathlonpass@1

51.8

API CALL INVOCATION EXAMPLE

// Example: Calling DeepSeek V4 Pro via OpenAI SDK
import OpenAI from "openai";

const client = new OpenAI({
  apiKey: process.env.DEEPSEEK_API_KEY,
});

const response = await client.chat.completions.create({
  model: "deepseek/deepseek-v4-pro",
  messages: [{ role: "user", content: "Hello! Can you inspect MCP tools?" }],
});

console.log(response.choices[0].message.content);