NOUMENOpen app
NOUMEN

DeepSeek · Open weights

DeepSeek V4 Pro

Added Apr 24, 2026Open weights

Compare models

Choose up to 3 models to compare.1 of 3 chosen

Loading models…

Use in the app

DeepSeek V4 Pro is a large-scale Mixture-of-Experts model from DeepSeek with a 1M-token context window, built for advanced reasoning, coding, long-horizon agent workflows, knowledge, math, and software engineering tasks.

  • Reasoning
  • Tools
  • Structured output
Input
Text
Output
Text
Reasoning effort
none · high · max
Discovery tier
Open weights

Pricing

USD per 1M tokens
Input
$0.70
Output
$1.40
Cache read
$0.07

Reference prices as published by the model provider. API example

Specifications

Context window
1M
Max output
384K
Average output (7 days)
416 tokens
Longer than 42% of 468 models

Side by side

DeepSeek V4 Pro compared

Specifications, prices, speed and benchmark scores next to each other. Benchmarks below show every compared model too.

Compare models

Choose up to 3 models to compare.1 of 3 chosen

Loading models…

DeepSeek V4 ProDeepSeekThis modelMiMo V2.6 FlashxiaomiRemove
ProviderDeepSeekxiaomi
Discovery tierOpen weightsOpen weights
AddedApr 24, 2026Sep 21, 2026
Limits
Context window1.05M1.05M
Max output384KBest131.07K
Average output416 tokens818 tokens
Pricing, USD per 1M tokens
Input$0.70$0.14Best
Output$1.40$0.28Best
Cache read$0.07$0.0028Best
Cache write—$0.00
Performance
Throughput49.9 tok/s105.6 tok/sBest
Time to first token2.9 s2.0 sBest
Request success (24h)99.7%Best99.3%
Benchmark indexes
Intelligence Index36.037.9Best
Coding Index68.8—
Capabilities
ReasoningYesYes
ToolsYesYes
VisionNoYes
Structured outputYesYes
Audio inputNoYes
Video inputNoYes
Reasoning effortnone, high, maxnone, high

Public evaluations

Benchmarks

Independent evaluation results, with how each score ranks among the models in the Noumen catalog.

  • DeepSeek V4 Pro
  • MiMo V2.6 Flash

Intelligence Index

  • DeepSeek V4 Pro36.0
  • MiMo V2.6 Flash37.9

DeepSeek V4 Pro tested at max reasoning effort

DeepSeek V4 Pro: Better than 79% of 326 scored models

3.8Catalog models by score55.1

Indexes

  • Coding Index

    Composite of the coding evaluations

    • DeepSeek V4 Pro: 68.8
    • MiMo V2.6 Flash: —

Agentic work

  • AutomationBench-AA

    Workflow automation, penalized for guardrail violations

    • DeepSeek V4 Pro: 56.7%
    • MiMo V2.6 Flash: 64.1%Top

  • AutomationBench-AA tasks completed

    Workflows finished end to end without a guardrail violation

    • DeepSeek V4 Pro: 25.9%
    • MiMo V2.6 Flash: 31.5%Top

  • AA-Briefcasev1.1

    Agentic knowledge work, rated head to head

    • DeepSeek V4 Pro: 1,256 Elo
    • MiMo V2.6 Flash: 1,488 EloTop

  • GDPval-AAv2.1

    Economically valuable occupational tasks, rated head to head

    • DeepSeek V4 Pro: 1,449 Elo
    • MiMo V2.6 Flash: 1,609 EloTop

Document reasoning

  • GDP.pdf

    Professional reasoning over PDF documents, all-pass rate

    • DeepSeek V4 Pro: 11.4%Top
    • MiMo V2.6 Flash: 9.0%

  • AA-LCRv1.1

    Reasoning across long documents

    • DeepSeek V4 Pro: 80.3%Top
    • MiMo V2.6 Flash: 74.3%

  • MLCR-AA

    Long-context reasoning over medical records

    • DeepSeek V4 Pro: —
    • MiMo V2.6 Flash: 11.7%

Reasoning and knowledge

  • Humanity’s Last Exam

    Expert-level questions across academic fields

    • DeepSeek V4 Pro: 41.0%Top
    • MiMo V2.6 Flash: 35.1%

  • GPQA Diamond

    Graduate-level science questions

    • DeepSeek V4 Pro: 92.8%
    • MiMo V2.6 Flash: —

Coding

  • Terminal-Benchv4.0

    Practical software engineering in a terminal

    • DeepSeek V4 Pro: 14.1%
    • MiMo V2.6 Flash: 22.7%Top

  • SciCode

    Scientific computing in Python

    • DeepSeek V4 Pro: 51.0%
    • MiMo V2.6 Flash: 51.3%Top

Legacy evaluations

  • AA-LCR (unversioned)

    Long-context reasoning, earlier grading

    • DeepSeek V4 Pro: 80.3%Top
    • MiMo V2.6 Flash: 74.3%

Scores from Artificial Analysis, updated Oct 26, 2026. DeepSeek V4 Pro was tested at max reasoning effort; compare scores measured at the same effort.Artificial Analysis

Live routing data

Performance and reliability

Recent averages across requests routed for this model. Your own speed depends on prompt length, reasoning effort and load.

Throughput
49.9 tok/s
Faster than 36% of 253 models
Time to first token
2.9 s
Quicker than 59% of 253 models
Prompt cache hits
64.5%
About $0.29 per 1M input tokens with caching
Request success (24h)
99.7%
Reliable

Cache hits are the share of input tokens recently served from a prompt cache across all requests. They are not a forecast of your own hit rate.

Hourly request success · last 72 hoursLowest hour 99.0%
  • Healthy (98% or more)
  • Degraded (90–98%)
  • Down (under 90%)
  • No requests

Developers

Use it through the API

Noumen’s API is OpenAI-compatible: any Chat Completions client works with a base URL and a key.

Read the quickstart
curl https://www.noumen.xyz/api/v1/chat/completions \
  -H "Authorization: Bearer $NOUMEN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "deepseek/deepseek-v4-pro",
    "reasoning_effort": "none",
    "messages": [{ "role": "user", "content": "Explain this stack trace." }]
  }'
import os
from openai import OpenAI

client = OpenAI(base_url="https://www.noumen.xyz/api/v1", api_key=os.environ["NOUMEN_API_KEY"])
reply = client.chat.completions.create(
    model="deepseek/deepseek-v4-pro",
    messages=[{"role": "user", "content": "Explain this stack trace."}],
)
print(reply.choices[0].message.content)

DeepSeek

Other DeepSeek models, starting with the same family.