NOUMENOpen app
NOUMEN

DeepSeek · Open weights

DeepSeek V4 Flash 0731 (Thinking)

Added Aug 1, 2026Open weights

Compare models

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

Loading models…

Use in the app

DeepSeek V4 Flash 0731 Thinking enables reasoning by default on the re-post-trained Mixture-of-Experts model with a 1M-token context window, built for coding, reasoning, and agent workflows.

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

Pricing

USD per 1M tokens
Input
$0.05
Output
$0.16
Cache read
$0.013

Reference prices as published by the model provider. API example

Specifications

Context window
1M
Max output
131.1K
Average output (7 days)
755 tokens
Longer than 63% of 465 models

Side by side

DeepSeek V4 Flash 0731 (Thinking) 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 Flash 0731 (Thinking)DeepSeekThis modelMuse Spark 1.1MetaRemove
ProviderDeepSeekMeta
Discovery tierOpen weightsBalanced
AddedAug 1, 2026Jul 9, 2026
Limits
Context window1.05MBest1M
Max output131.07KBest65.54K
Average output755 tokens460 tokens
Pricing, USD per 1M tokens
Input$0.05Best$1.25
Output$0.16Best$4.25
Cache read$0.013Best$0.15
Performance
Throughput147.4 tok/s—
Time to first token5.7 s—
Request success (24h)99.8%—
Benchmark indexes
Intelligence Index34.3Best33.7
Coding Index69.171.3Best
Capabilities
ReasoningYesYes
ToolsYesNo
VisionNoYes
Structured outputYesNo
Audio inputNoYes
Video inputNoYes
Reasoning effortnone, low, high, maxminimal, low, medium, high, xhigh

Public evaluations

Benchmarks

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

  • DeepSeek V4 Flash 0731 (Thinking)
  • Muse Spark 1.1

Intelligence Index

  • DeepSeek V4 Flash 0731 (Thinking)34.3
  • Muse Spark 1.133.7

DeepSeek V4 Flash 0731 (Thinking) tested at max reasoning effort

DeepSeek V4 Flash 0731 (Thinking): Better than 78% of 326 scored models

3.8Catalog models by score55.1

Indexes

  • Coding Index

    Composite of the coding evaluations

    • DeepSeek V4 Flash 0731 (Thinking): 69.1
    • Muse Spark 1.1: 71.3Top

Agentic work

  • AutomationBench-AA

    Workflow automation, penalized for guardrail violations

    • DeepSeek V4 Flash 0731 (Thinking): 54.0%Top
    • Muse Spark 1.1: 38.8%

  • AutomationBench-AA tasks completed

    Workflows finished end to end without a guardrail violation

    • DeepSeek V4 Flash 0731 (Thinking): 23.0%Top
    • Muse Spark 1.1: 13.7%

  • AA-Briefcasev1.1

    Agentic knowledge work, rated head to head

    • DeepSeek V4 Flash 0731 (Thinking): 1,257 EloTop
    • Muse Spark 1.1: 850 Elo

  • GDPval-AAv2.1

    Economically valuable occupational tasks, rated head to head

    • DeepSeek V4 Flash 0731 (Thinking): 1,433 EloTop
    • Muse Spark 1.1: 1,211 Elo

Document reasoning

  • GDP.pdf

    Professional reasoning over PDF documents, all-pass rate

    • DeepSeek V4 Flash 0731 (Thinking): 11.0%
    • Muse Spark 1.1: 14.4%Top

  • AA-LCRv1.1

    Reasoning across long documents

    • DeepSeek V4 Flash 0731 (Thinking): 79.7%Top
    • Muse Spark 1.1: 77.7%

  • MLCR-AA

    Long-context reasoning over medical records

    • DeepSeek V4 Flash 0731 (Thinking): —
    • Muse Spark 1.1: 7.2%

Reasoning and knowledge

  • Humanity’s Last Exam

    Expert-level questions across academic fields

    • DeepSeek V4 Flash 0731 (Thinking): 38.6%
    • Muse Spark 1.1: 46.2%Top

  • GPQA Diamond

    Graduate-level science questions

    • DeepSeek V4 Flash 0731 (Thinking): 90.8%Top
    • Muse Spark 1.1: 89.8%

Coding

  • Terminal-Benchv4.0

    Practical software engineering in a terminal

    • DeepSeek V4 Flash 0731 (Thinking): 12.1%Top
    • Muse Spark 1.1: 6.1%

  • SciCode

    Scientific computing in Python

    • DeepSeek V4 Flash 0731 (Thinking): 50.3%
    • Muse Spark 1.1: 58.8%Top

Legacy evaluations

  • AA-LCR (unversioned)

    Long-context reasoning, earlier grading

    • DeepSeek V4 Flash 0731 (Thinking): 79.7%Top
    • Muse Spark 1.1: 77.7%

Scores from Artificial Analysis, updated Oct 26, 2026. DeepSeek V4 Flash 0731 (Thinking) 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
147.4 tok/s
Faster than 75% of 250 models
Time to first token
5.7 s
Quicker than 28% of 250 models
Prompt cache hits
74.1%
About $0.0226 per 1M input tokens with caching
Request success (24h)
99.8%
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 97.3%
  • 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-flash-0731:thinking",
    "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-flash-0731:thinking",
    messages=[{"role": "user", "content": "Explain this stack trace."}],
)
print(reply.choices[0].message.content)

DeepSeek

Other DeepSeek models, starting with the same family.