NOUMENOpen app
NOUMEN

DeepSeek · Open weights

DeepSeek V4 Flash 0423 (Thinking)

Added Oct 7, 2026Open weights

Compare models

Choose up to 3 models to compare.0 of 3 chosen

Loading models…

Use in the app

The original 0423 checkpoint of DeepSeek V4 Flash, with a 1M-token context window, text input, tool calling, structured output, and optional reasoning.

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

Pricing

USD per 1M tokens
Input
$0.14
Output
$0.28
Cache read
$0.028

Reference prices as published by the model provider. API example

Specifications

Context window
1M
Max output
—
Average output (7 days)
1.1K tokens
Longer than 72% of 465 models

Side by side

Compare models

Put DeepSeek V4 Flash 0423 (Thinking) next to up to three other models. The comparison stays in the link, so you can share it.

Compare models

Choose up to 3 models to compare.0 of 3 chosen

Loading models…

TryLing 3.0 Flash VLSolar Mini 4Claude Sonnet 4.6All 3

Public evaluations

Benchmarks

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

Intelligence Index

24.4

Tested at high reasoning effort

Better than 59% of 326 scored models

3.8Catalog models by score55.1

Indexes

  • Coding Index

    52.0

    Composite of the coding evaluations

    Better than 50% of 168 scored models

Agentic work

  • AutomationBench-AA

    28.5%

    Workflow automation, penalized for guardrail violations

    Better than 49% of 173 scored models

  • AutomationBench-AA tasks completed

    8.7%

    Workflows finished end to end without a guardrail violation

    Better than 50% of 173 scored models

  • AA-Briefcasev1.1

    1,041 Elo

    Agentic knowledge work, rated head to head

    Better than 47% of 166 scored models

  • GDPval-AAv2.1

    1,000 Elo

    Economically valuable occupational tasks, rated head to head

    Better than 35% of 187 scored models

Document reasoning

  • GDP.pdf

    10.4%

    Professional reasoning over PDF documents, all-pass rate

    Better than 43% of 171 scored models

  • AA-LCRv1.1

    72.0%

    Reasoning across long documents

    Better than 53% of 309 scored models

Reasoning and knowledge

  • Humanity’s Last Exam

    30.3%

    Expert-level questions across academic fields

    Better than 64% of 323 scored models

  • GPQA Diamond

    86.7%

    Graduate-level science questions

    Better than 67% of 291 scored models

  • IFBench

    73.5%

    Following precise output instructions

    Better than 83% of 215 scored models

Coding

  • Terminal-Benchv4.0

    3.0%

    Practical software engineering in a terminal

    Better than 50% of 171 scored models

  • SciCode

    40.2%

    Scientific computing in Python

    Better than 22% of 174 scored models

Legacy evaluations

  • Terminal-Bench Hard

    38.6%

    Earlier terminal task suite

    Better than 75% of 212 scored models

  • τ²-Bench Telecom

    95.6%

    Tool use in a customer-support simulation

    Better than 87% of 214 scored models

  • AA-LCR (unversioned)

    72.0%

    Long-context reasoning, earlier grading

    Better than 53% of 309 scored models

Scores from Artificial Analysis, updated Oct 26, 2026. DeepSeek V4 Flash 0423 (Thinking) was tested at high 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
47.5 tok/s
Faster than 35% of 250 models
Time to first token
1.6 s
Quicker than 86% of 250 models
Prompt cache hits
57.2%
About $0.076 per 1M input tokens with caching
Request success (24h)
95.8%
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 80.0%
  • 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-flash-0423: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-0423:thinking",
    messages=[{"role": "user", "content": "Explain this stack trace."}],
)
print(reply.choices[0].message.content)

DeepSeek

Other DeepSeek models, starting with the same family.