NOUMENOpen app
NOUMEN

DeepSeek · Open weights

DeepSeek V4.1 Flash TEE

Added Sep 12, 2026Open weights

Compare models

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

Loading models…

Use in the app

DeepSeek V4.1 Flash's October 17, 2026 checkpoint (1017) supports text and image input, reasoning, tool calling, and structured output with a 1M-token context window. This route runs through Tinfoil attested inference inside a Trusted Execution Environment. Tinfoil lists this model as experimental, with availability subject to change.

  • Reasoning
  • Tools
  • Vision
  • Structured output
  • Parallel tools
Input
Text, Images
Output
Text
Reasoning effort
none · low · medium · high · max
Discovery tier
Open weights

Pricing

USD per 1M tokens
Input
$0.65
Output
$1.45
Cache read
$0.13

Reference prices as published by the model provider. API example

Specifications

Context window
1M
Max output
384K
Average output (7 days)
571 tokens
Longer than 53% of 465 models

Side by side

DeepSeek V4.1 Flash TEE 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 V4.1 Flash TEEDeepSeekThis modelClaude Sonnet 4.6AnthropicRemove
ProviderDeepSeekAnthropic
Discovery tierOpen weightsFrontier
AddedSep 12, 2026Feb 17, 2026
Limits
Context window1.05MBest1M
Max output384KBest128K
Average output571 tokens401 tokens
Pricing, USD per 1M tokens
Input$0.65Best$3.00
Output$1.45Best$15.00
Cache read$0.13Best$0.30
Performance
Throughput239.9 tok/sBest79.9 tok/s
Time to first token3.0 s1.6 sBest
Request success (24h)99.9%Best99.6%
Benchmark indexes
Intelligence Index24.724.7
Capabilities
ReasoningYesYes
ToolsYesYes
VisionYesYes
Structured outputYesYes
PDF inputNoYes
Parallel toolsYesNo
Reasoning effortnone, low, medium, high, maxlow, medium, high, xhigh, max

Public evaluations

Benchmarks

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

  • DeepSeek V4.1 Flash TEE
  • Claude Sonnet 4.6

Intelligence Index

  • DeepSeek V4.1 Flash TEE24.7
  • Claude Sonnet 4.624.7

DeepSeek V4.1 Flash TEE: Better than 60% of 326 scored models

3.8Catalog models by score55.1

Agentic work

  • AutomationBench-AA

    Workflow automation, penalized for guardrail violations

    • DeepSeek V4.1 Flash TEE: 47.8%
    • Claude Sonnet 4.6: —

  • AutomationBench-AA tasks completed

    Workflows finished end to end without a guardrail violation

    • DeepSeek V4.1 Flash TEE: 14.2%
    • Claude Sonnet 4.6: —

  • AA-Briefcasev1.1

    Agentic knowledge work, rated head to head

    • DeepSeek V4.1 Flash TEE: 1,118 Elo
    • Claude Sonnet 4.6: —

  • GDPval-AAv2.1

    Economically valuable occupational tasks, rated head to head

    • DeepSeek V4.1 Flash TEE: 1,340 Elo
    • Claude Sonnet 4.6: —

Document reasoning

  • GDP.pdf

    Professional reasoning over PDF documents, all-pass rate

    • DeepSeek V4.1 Flash TEE: 3.8%
    • Claude Sonnet 4.6: —

  • AA-LCRv1.1

    Reasoning across long documents

    • DeepSeek V4.1 Flash TEE: 55.3%
    • Claude Sonnet 4.6: 68.3%Top

Reasoning and knowledge

  • Humanity’s Last Exam

    Expert-level questions across academic fields

    • DeepSeek V4.1 Flash TEE: 10.8%
    • Claude Sonnet 4.6: 13.3%Top

  • GPQA Diamond

    Graduate-level science questions

    • DeepSeek V4.1 Flash TEE: —
    • Claude Sonnet 4.6: 79.9%

  • IFBench

    Following precise output instructions

    • DeepSeek V4.1 Flash TEE: —
    • Claude Sonnet 4.6: 41.2%

Coding

  • Terminal-Benchv4.0

    Practical software engineering in a terminal

    • DeepSeek V4.1 Flash TEE: 5.6%
    • Claude Sonnet 4.6: —

  • SciCode

    Scientific computing in Python

    • DeepSeek V4.1 Flash TEE: 35.5%
    • Claude Sonnet 4.6: —

Legacy evaluations

  • Terminal-Bench Hard

    Earlier terminal task suite

    • DeepSeek V4.1 Flash TEE: —
    • Claude Sonnet 4.6: 46.2%

  • τ²-Bench Telecom

    Tool use in a customer-support simulation

    • DeepSeek V4.1 Flash TEE: —
    • Claude Sonnet 4.6: 79.5%

  • AA-LCR (unversioned)

    Long-context reasoning, earlier grading

    • DeepSeek V4.1 Flash TEE: 55.3%
    • Claude Sonnet 4.6: 68.3%Top

Scores from Artificial Analysis, updated Oct 26, 2026. 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
239.9 tok/s
Faster than 88% of 250 models
Time to first token
3.0 s
Quicker than 57% of 250 models
Prompt cache hits
58.8%
About $0.34 per 1M input tokens with caching
Request success (24h)
99.9%
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 97.6%
  • 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": "TEE/deepseek-v4.1-flash",
    "reasoning_effort": "medium",
    "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="TEE/deepseek-v4.1-flash",
    messages=[{"role": "user", "content": "Explain this stack trace."}],
)
print(reply.choices[0].message.content)

DeepSeek

Other DeepSeek models, starting with the same family.