download dots

DeepSeek · tested in Taskade Genesis

DeepSeek V4.1 Flash in Taskade

DeepSeek V4.1 Flash from DeepSeek runs on every plan, including Free. We last tested it on Oct 8, 2026 in TSK-1, our app-building test.

Status
Available
Plan
Every plan, including Free
Your own key
No
Last test
Context window
1Mtokens
Max output
384Ktokens
Released
Knowledge cutoff
Where it runs
14services

TSK-1 scorecard for the DeepSeek family

TSK Score

DeepSeek family, across real apps made in Taskade Genesis

TSK Score 75 out of 100
DeepSeek family on the four TSK-1 axesInterfaceTaskMemoryAdapt
Interfaceready to share
Leading
Taskcompletes your task
Strong
Memorykeeps your data
Strong
Adapthandles follow-up edits
Emerging

The richest app per shape among the economical models.

More

It built a six-page sales CRM whose scoring automation ran, a four-page match tracker, and a four-page cookbook with both automations live. On a family intake form it saved every answer and both photos.

DeepSeek on TSK-1

Published scores

Benchmark scores that DeepSeek and test sites published for DeepSeek V4.1 Flash. Taskade did not re-run them. Lab-reported marks a score from DeepSeek itself.

Published benchmark scores for DeepSeek V4.1 Flash, with sources
BenchmarkScore
Terminal-Benchpass@1 · v2.1 · reasoning_effort=100 · DeepSeek Harness minimal modeHugging FaceLab-reported90.6%
GPQA Diamondpass@1 · reasoning_effort=100Hugging FaceLab-reported90.9%
Humanity's Last Exampass@1 · reasoning_effort=100 · full setHugging FaceLab-reported36.8%
More published scores (31)
Published benchmark scores for DeepSeek V4.1 Flash, with sources: More published scores (31)
BenchmarkScore
MathArena Apexpass@1 · reasoning_effort=100Hugging FaceLab-reported65.6%
Humanity's Last Exampass@1 · reasoning_effort=100 · text-only subsetHugging FaceLab-reported39.1%
Codeforcesrating · reasoning_effort=100Hugging FaceLab-reported3,471
Terminal-Benchpass@1 · v3.0 · reasoning_effort=100 · DeepSeek Harness minimal modeHugging FaceLab-reported30%
Terminal-Benchpass@1 · v4.0 · reasoning_effort=100 · DeepSeek Harness minimal modeHugging FaceLab-reported31.2%
DeepSWEresolved · v1.1 · reasoning_effort=100 · mini-swe-agentHugging FaceLab-reported74.2%
ProgramBenchalmost@1 · reasoning_effort=100 · DeepSeek Harness minimal modeHugging FaceLab-reported20.3
NL2Reporeasoning_effort=100 · DeepSeek Harness minimal modeHugging FaceLab-reported64
CyberGympass@1 · reasoning_effort=100Hugging FaceLab-reported88.1%
ExploitGympass@1 · reasoning_effort=100Hugging FaceLab-reported15.3%
AutomationBenchpass@1 · reasoning_effort=100Hugging FaceLab-reported54.8%
Agents' Last Exampass@1 · reasoning_effort=100Hugging FaceLab-reported31.8%
SEC-Bench Propass@1 · reasoning_effort=100 · Claude CodeHugging FaceLab-reported62.8%
Humanity's Last Exampass@1 · reasoning_effort=100; with toolsHugging FaceLab-reported63.9%
Chartographypass@1 · reasoning_effort=100; with tools · Claude CodeHugging FaceLab-reported78.9%
BabyVisionpass@1 · reasoning_effort=100; with tools · Claude CodeHugging FaceLab-reported89.6%
ZeroBenchpass@5 · reasoning_effort=100; with tools · Claude Code · mainHugging FaceLab-reported49%
DeepSWEresolved · v1.1 · reasoning_effort=100 · Claude CodeHugging FaceLab-reported69.8%
Terminal-Benchpass@1 · v2.1 · reasoning_effort=100 · Claude CodeHugging FaceLab-reported88%
DeepSWEresolved · v1.1 · reasoning_effort=100 · CodexHugging FaceLab-reported65.6%
Terminal-Benchpass@1 · v2.1 · reasoning_effort=100 · CodexHugging FaceLab-reported84.1%
DeepSWEresolved · v1.1 · reasoning_effort=100 · OpenCodeHugging FaceLab-reported65.5%
Terminal-Benchpass@1 · v2.1 · reasoning_effort=100 · OpenCodeHugging FaceLab-reported85%
DeepSWEresolved · v1.1 · reasoning_effort=100 · PiHugging FaceLab-reported66.2%
Terminal-Benchpass@1 · v2.1 · reasoning_effort=100 · PiHugging FaceLab-reported86.1%
Terminal-Benchpass@1 · v2.1 · reasoning_effort=100 · mini-swe-agentHugging FaceLab-reported90.3%
DeepSWEresolved · v1.1 · reasoning_effort=100 · DeepSeek Harness minimal modeHugging FaceLab-reported72.6%
DeepSWEresolved · v1.1 · reasoning_effort=100 · DeepSeek Harness standard modeHugging FaceLab-reported70.5%
Terminal-Benchpass@1 · v2.1 · reasoning_effort=100 · DeepSeek Harness standard modeHugging FaceLab-reported85.8%
DeepSWEresolved · v1.1 · reasoning_effort=100 · DeepSeek Harness PTC modeHugging FaceLab-reported67.6%
Terminal-Benchpass@1 · v2.1 · reasoning_effort=100 · DeepSeek Harness PTC modeHugging FaceLab-reported85.8%
How DeepSeek builds apps in TSK-1, our own test

Scores collected by models.dev (MIT). Each source links the page that published the score.

Versions tested

Open a version to read its findings.

DeepSeek V4.1 FlashTested Oct 202612 findings

The richest app per shape among the economical models. It built a six-page sales CRM whose scoring automation ran, a four-page match tracker, and a four-page cookbook with both automations live. On a family intake form it saved every answer and both photos.

Sep 19, 2026

Built the richest sales CRM of the test, with six pages and a scoring automation that scored all 7 sample leads.

Sep 19, 2026

Built a four-page match tracker and said plainly that its weekly recap was still switched off.

Sep 19, 2026

Typed all 32 client sign-up questions and built a 48-field submissions table, then ran out of time before finishing.

Sep 19, 2026

Built a three-page fleet inspection app for 42 vans where the brief named 40.

Sep 19, 2026

Its recipe box came out as a printed cookbook with 13 recipes, four pages, and both automations live.

Sep 22, 2026

Built a three-page match tracker whose coaching automation ran on the first match logged and wrote a note back into that match about six minutes later.

Sep 30, 2026

Built a two-page family intake form that saved every answer and both photos, and its alert automation ran twice. The build ran past the 15-minute test limit.

Sep 30, 2026

Added the new teacher question and put the change live.

Oct 8, 2026

Built the family intake form in 7 minutes 25 seconds so it saved every answer and both photos, and its new-family automation ran. Three sample families showed on the teacher page as if they were real.

Oct 8, 2026

Added the new teacher question as an optional field and kept it off the printed facesheet, and said so.

Oct 8, 2026

Built a five-page Dota 2 match tracker with labeled sample matches, but every hero showed as unknown and no match could be saved, while its summary said the save had been checked.

Oct 8, 2026

Rebuilt the 32-question client sign-up form with all 32 questions word for word and a scoring rubric, but the form saved each answer one question off, so every submission was rejected.

See every DeepSeek result

Model facts

The token limits are the most the model accepts. Taskade can run it with lower limits.

Last updated
Input types
Text, Images
Output types
Text
Reasoning
Yes
Tool calling
Yes
File attachments
Yes
Structured output
Yes
Open weights
Yes
Reasoning effort
NoneLowHighMax

Published specs from the model maker, via models.dev (MIT).

Where it runs

Services outside Taskade that serve DeepSeek V4.1 Flash, from models.dev. Each link opens in a new tab.

How to choose DeepSeek V4.1 Flash

  1. In the builder chat, open the model menu and choose More models. Then pick the model.
  2. In the agent editor, pick the model under All Models.
  3. If your model menu does not show the model, Taskade uses TSK-Auto instead.

DeepSeek in the docs

Where DeepSeek ranks

TSK Score out of 100.

Compare DeepSeek

Other DeepSeek models

FAQ

Which Taskade plans include DeepSeek V4.1 Flash?

DeepSeek V4.1 Flash runs in Taskade on every plan, including Free. If your model menu does not show it, Taskade uses TSK-Auto instead.

Has Taskade tested DeepSeek V4.1 Flash?

Yes. We last tested DeepSeek V4.1 Flash on Oct 8, 2026 in the TSK-1 Benchmark. The richest app per shape among the economical models. It built a six-page sales CRM whose scoring automation ran, a four-page match tracker, and a four-page cookbook with both automations live. On a family intake form it saved every answer and both photos.

How do I choose DeepSeek V4.1 Flash in Taskade?

In the builder chat, open the model menu and choose More models. Then pick DeepSeek V4.1 Flash. In the agent editor, pick it under All Models.

Who makes DeepSeek V4.1 Flash?

DeepSeek makes DeepSeek V4.1 Flash. Taskade runs it on every plan, including Free, beside frontier models from other top AI labs.