NOUMENOpen app
NOUMEN

DeepSeek · Open weights

DeepSeek V4 Flash

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 Flash is an efficiency-optimized Mixture-of-Experts model from DeepSeek with a 1M-token context window, built for fast inference, high-throughput workloads, reasoning, coding, and agent workflows.

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

Pricing

USD per 1M tokens
Input
$0.14
Output
$0.28
Cache read
$0.0028

Reference prices as published by the model provider. API example

Specifications

Context window
1M
Max output
384K
Average output (7 days)
137 tokens
Longer than 14% of 468 models

Side by side

DeepSeek V4 Flash 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 FlashDeepSeekThis modelDeepSeek V4 Flash 0423DeepSeekRemove
ProviderDeepSeekDeepSeek
Discovery tierOpen weightsOpen weights
AddedApr 24, 2026Oct 7, 2026
Limits
Context window1.05M1.05M
Max output384K—
Average output137 tokens245 tokens
Pricing, USD per 1M tokens
Input$0.14$0.14
Output$0.28$0.28
Cache read$0.0028Best$0.028
Performance
Throughput74.5 tok/s148.4 tok/sBest
Time to first token2.2 s1.9 sBest
Request success (24h)99.8%Best98.4%
Benchmark indexes
Intelligence Index34.3Best18.9
Coding Index69.1—
Capabilities
ReasoningYesYes
ToolsYesYes
Structured outputYesYes
Reasoning effortnone, high, maxnone, low, high, max

Public evaluations

Benchmarks

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

  • DeepSeek V4 Flash
  • DeepSeek V4 Flash 0423

Intelligence Index

  • DeepSeek V4 Flash34.3
  • DeepSeek V4 Flash 042318.9

DeepSeek V4 Flash tested at max reasoning effort

DeepSeek V4 Flash: Better than 78% of 326 scored models

3.8Catalog models by score55.1

Indexes

  • Coding Index

    Composite of the coding evaluations

    • DeepSeek V4 Flash: 69.1
    • DeepSeek V4 Flash 0423: —

Agentic work

  • AutomationBench-AA

    Workflow automation, penalized for guardrail violations

    • DeepSeek V4 Flash: 54.0%
    • DeepSeek V4 Flash 0423: —

  • AutomationBench-AA tasks completed

    Workflows finished end to end without a guardrail violation

    • DeepSeek V4 Flash: 23.0%
    • DeepSeek V4 Flash 0423: —

  • AA-Briefcasev1.1

    Agentic knowledge work, rated head to head

    • DeepSeek V4 Flash: 1,257 Elo
    • DeepSeek V4 Flash 0423: —

  • GDPval-AAv2.1

    Economically valuable occupational tasks, rated head to head

    • DeepSeek V4 Flash: 1,433 Elo
    • DeepSeek V4 Flash 0423: —

Document reasoning

  • GDP.pdf

    Professional reasoning over PDF documents, all-pass rate

    • DeepSeek V4 Flash: 11.0%
    • DeepSeek V4 Flash 0423: —

  • AA-LCRv1.1

    Reasoning across long documents

    • DeepSeek V4 Flash: 79.7%Top
    • DeepSeek V4 Flash 0423: 41.7%

Reasoning and knowledge

  • Humanity’s Last Exam

    Expert-level questions across academic fields

    • DeepSeek V4 Flash: 38.6%Top
    • DeepSeek V4 Flash 0423: 7.8%

  • GPQA Diamond

    Graduate-level science questions

    • DeepSeek V4 Flash: 90.8%Top
    • DeepSeek V4 Flash 0423: 71.6%

  • IFBench

    Following precise output instructions

    • DeepSeek V4 Flash: —
    • DeepSeek V4 Flash 0423: 47.2%

Coding

  • Terminal-Benchv4.0

    Practical software engineering in a terminal

    • DeepSeek V4 Flash: 12.1%
    • DeepSeek V4 Flash 0423: —

  • SciCode

    Scientific computing in Python

    • DeepSeek V4 Flash: 50.3%
    • DeepSeek V4 Flash 0423: —

Legacy evaluations

  • Terminal-Bench Hard

    Earlier terminal task suite

    • DeepSeek V4 Flash: —
    • DeepSeek V4 Flash 0423: 34.1%

  • τ²-Bench Telecom

    Tool use in a customer-support simulation

    • DeepSeek V4 Flash: —
    • DeepSeek V4 Flash 0423: 94.4%

  • AA-LCR (unversioned)

    Long-context reasoning, earlier grading

    • DeepSeek V4 Flash: 79.7%Top
    • DeepSeek V4 Flash 0423: 41.7%

Scores from Artificial Analysis, updated Oct 26, 2026. DeepSeek V4 Flash 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
74.5 tok/s
Faster than 49% of 254 models
Time to first token
2.2 s
Quicker than 71% of 254 models
Prompt cache hits
85.9%
About $0.0222 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 98.2%
  • 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",
    "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",
    messages=[{"role": "user", "content": "Explain this stack trace."}],
)
print(reply.choices[0].message.content)

DeepSeek

Other DeepSeek models, starting with the same family.