DeepSeek · tested in Taskade Genesis
DeepSeek V4 Flash in Taskade
DeepSeek V4 Flash from DeepSeek is no longer in the Taskade model menu. Taskade now runs a request for it on TSK-Instant. Its test results stay on this page. We last tested it on Aug 25, 2026 in TSK-1, our app-building test.
- Status
- Replaced
- Last test
- Context window
- 1Mtokens
- Max output
- 384Ktokens
- Released
- Knowledge cutoff
- Where it runs
- 5services
TSK-1 scorecard for the DeepSeek family
TSK Score
DeepSeek family, across real apps made in Taskade Genesis
- Interfaceready to share
- Leading
- Taskcompletes your task
- Strong
- Memorykeeps your data
- Strong
- Adapthandles follow-up edits
- Emerging
The cheapest model won the design test.
MoreLess
It was the only one that looked right in both light and dark, ran without a single error, and laid out cleanly on a phone, for a fraction of what the others cost. On a real customer brief it was cheapest, fastest and cleanest at once. It has since left the Taskade model picker.
Published scores
Benchmark scores that DeepSeek and test sites published for DeepSeek V4 Flash. Taskade did not re-run them. Lab-reported marks a score from DeepSeek itself.
| Benchmark | Score | Source |
|---|---|---|
| SWE-Bench VerifiedresolvedHugging FaceLab-reported | 79% | Hugging FaceLab-reported |
| SWE-Bench Proresolved · preview checkpoint; max effortHugging FaceLab-reported | 52.6% | Hugging FaceLab-reported |
| Terminal-Benchaccuracy · v2.0 · preview checkpoint; max effortHugging FaceLab-reported | 56.9% | Hugging FaceLab-reported |
| LiveCodeBenchpass@1 · preview checkpoint; max effortHugging FaceLab-reported | 91.6% | Hugging FaceLab-reported |
| BrowseComppass@1 · preview checkpoint; max effortHugging FaceLab-reported | 73.2% | Hugging FaceLab-reported |
| MCP Atlaspass@1 · preview checkpoint; max effort · publicHugging FaceLab-reported | 69% | Hugging FaceLab-reported |
| Toolathlonpass@1 · preview checkpoint; max effortHugging FaceLab-reported | 47.8% | Hugging FaceLab-reported |
| GPQA Diamondpass@1 · preview checkpoint; max effortHugging FaceLab-reported | 88.1% | Hugging FaceLab-reported |
More published scores (14)
| Benchmark | Score | Source |
|---|---|---|
| MMLU-ProEM · preview checkpoint; max effortHugging FaceLab-reported | 86.2 | Hugging FaceLab-reported |
| SimpleQA-Verifiedpass@1 · preview checkpoint; max effortHugging FaceLab-reported | 34.1% | Hugging FaceLab-reported |
| Chinese SimpleQApass@1 · preview checkpoint; max effortHugging FaceLab-reported | 78.9% | Hugging FaceLab-reported |
| Humanity's Last Exampass@1 · preview checkpoint; max effort; without toolsHugging FaceLab-reported | 34.8% | Hugging FaceLab-reported |
| Codeforcesrating · preview checkpoint; max effortHugging FaceLab-reported | 3,052 | Hugging FaceLab-reported |
| HMMTpass@1 · preview checkpoint; max effort · February 2026Hugging FaceLab-reported | 94.8% | Hugging FaceLab-reported |
| IMOAnswerBenchpass@1 · preview checkpoint; max effortHugging FaceLab-reported | 88.4% | Hugging FaceLab-reported |
| MathArena Apexpass@1 · preview checkpoint; max effortHugging FaceLab-reported | 33% | Hugging FaceLab-reported |
| MathArena Apex Shortlistpass@1 · preview checkpoint; max effortHugging FaceLab-reported | 85.7% | Hugging FaceLab-reported |
| MRCRMMR · preview checkpoint; max effort · 1M contextHugging FaceLab-reported | 78.7 | Hugging FaceLab-reported |
| CorpusQAaccuracy · preview checkpoint; max effort · 1M contextHugging FaceLab-reported | 60.5% | Hugging FaceLab-reported |
| SWE-Bench Multilingualresolved · preview checkpoint; max effortHugging FaceLab-reported | 73.3% | Hugging FaceLab-reported |
| Humanity's Last Exampass@1 · preview checkpoint; max effort; with toolsHugging FaceLab-reported | 45.1% | Hugging FaceLab-reported |
| GDPval-AAElo · preview checkpoint; max effortHugging FaceLab-reported | 1,395 | Hugging FaceLab-reported |
LMArena scores
| Benchmark | Score | Source |
|---|---|---|
| Text arena±4 · 51,809 votes · Rank 98 of 416 · LMArena | 1,439 | LMArena |
| WebDev arena±7 · 10,106 votes · Rank 85 of 145 · LMArena | 1,431 | LMArena |
Scores from the LMArena leaderboard dataset (CC BY 4.0). People vote between two answers, and the votes set the score. Taskade picked the Overall category and the default setting of each model.
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 FlashTested Aug 20265 findings
The cheapest model won the design test. It was the only one that looked right in both light and dark, ran without a single error, and laid out cleanly on a phone, for a fraction of what the others cost. On a real customer brief it was cheapest, fastest and cleanest at once. It has since left the Taskade model picker.
Won on design: the only model that looked right in both light and dark, ran without a single error, and laid out cleanly on a 390px phone screen, at a fraction of the field's cost.
Won on design again, and this time the finished workspace saved real data with light and dark themes both working throughout.
Cheapest, fastest and cleanest run on the 32-question client sign-up form: 12 minutes 36 seconds.
Same test, same prompt: better value than Luna, at half the cost of the build.
Finished the tracker and made both follow-up changes cleanly, but its long-form build ran out of time.
Model facts
The token limits are the most the model accepts. Taskade can run it with lower limits.
- Last updated
- Input types
- Text
- Output types
- Text
- Reasoning
- Yes
- Tool calling
- Yes
- File attachments
- No
- Structured output
- Yes
- Open weights
- Yes
- Reasoning effort
- NoneLowMediumHigh
Published specs from the model maker, via models.dev (MIT).
Where it runs
Services outside Taskade that serve DeepSeek V4 Flash, from models.dev. Each link opens in a new tab.
What to pick instead
- Leave TSK-Auto on, and Taskade picks a current model for each request.
- To pick a model yourself, choose one from the model menu.
- To compare, open the models page and read the test results of the current models.
Where DeepSeek ranks
TSK Score out of 100.
Compare DeepSeek
Other DeepSeek models
FAQ
Can I still pick DeepSeek V4 Flash in Taskade?
No. DeepSeek V4 Flash is no longer in the Taskade model menu. TSK-Auto picks a current model for each request. The models page lists the models you can pick now.
Has Taskade tested DeepSeek V4 Flash?
Yes. We last tested DeepSeek V4 Flash on Aug 25, 2026 in the TSK-1 Benchmark. The cheapest model won the design test. It was the only one that looked right in both light and dark, ran without a single error, and laid out cleanly on a phone, for a fraction of what the others cost. On a real customer brief it was cheapest, fastest and cleanest at once. It has since left the Taskade model picker.
Who makes DeepSeek V4 Flash?
DeepSeek makes DeepSeek V4 Flash. Taskade runs frontier models from top AI labs.