Megatron MarketsWholesale
Sign inContact Us
Models Model details

DeepSeek

DeepSeek: DeepSeek V4 Flash 0731

deepseek/deepseek-v4-flash-0731

Enquire about this model Request a quote

DeepSeek V4 Flash 0731 is a sparse mixture-of-experts model from DeepSeek, with 13B active parameters out of 284B total. This re-post-trained revision is suited for coding, reasoning, and agent workflows. This is the GA release of DeepSeek V4 Flash.

Input
$0.0045

/ 1M tokens

Output
$1.28

/ 1M tokens

Context window
1.05M

tokens

Limits

Context

1,048,576

tokens

Max output

943,718

90% of the context window

Parameters

304,180,418,494

Model facts

Author

DeepSeek

Tokenizer

DeepSeek

Modality

Text -> Text

License

MIT

Created

Jul 31, 2026

Hugging Face

deepseek-ai/DeepSeek-V4-Flash-0731

Reasoning

Default HighMaxHighLowNone

Publication status

Hugging Face: verifiedLicense: publishedParameter count: published

Modalities

Input · TextOutput · Text

Generation controls

Frequency penaltyInclude reasoningLogit biasLogprobsMax tokensMin pParallel tool callsPresence penaltyReasoningReasoning effortRepetition penaltyResponse formatSeedStopStructured outputsTemperatureTool choiceToolsTop aTop kTop logprobsTop p

Capability scores

41
Agentic index

Better than 73% of models compared

69.1
Coding index

Better than 72% of models compared

34.3
Intelligence index

Better than 77% of models compared

Creative tasks

Category performance

3D
Top 43%

1209 Elo

48.7% win

UI components
Top 44%

1235 Elo

46.8% win

Websites
Top 41%

1239 Elo

46.7% win

Code
Top 43%

1231 Elo

46.4% win

SVG
Top 45%

1183 Elo

45.1% win

Games
Top 43%

1209 Elo

44.6% win

Data visualization
Top 57%

1180 Elo

41.1% win

ASCII art
Top 93%

1093 Elo

33.3% win