DeepSeek

DeepSeek V4 Flash 0731

by DeepSeektext->text28 endpoints
Neura Intelligence Index
89.1/ 100
Rank
#15
Confidence
low
Benchmarks
1(3% coverage)

Overview

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....

Capabilities

Text generation
Tool use / Function calling
Structured output
Extended reasoning
Prompt caching
Long context
Top-P sampling
Stop sequences
Deterministic seed
Log probabilities

Modalities

Input
Text
Output
Text

Technical Specifications

Context Window
1.3M tokens
Max Output
943.7K tokens
Tokenizer
DeepSeek
Uptime (24h)
10000.0%
Parameters
304000000000B
License
mitOpen Weights
Country
🇨🇳CN
Release Date
July 31, 20261mo

Supported Parameters

frequency_penaltyinclude_reasoninglogit_biaslogprobsmax_tokensmin_pparallel_tool_callspresence_penaltyreasoningreasoning_effortrepetition_penaltyresponse_formatseedstopstructured_outputstemperaturetool_choicetoolstop_atop_ktop_logprobstop_p

Benchmark Performance

agent

BenchmarkScoreSource
Toolathlon
70.3
verified

Performance Over Time

Pricing

Per 1M tokens
Input$0.04
Output$0.08
Cache read$0.02
Added
July 31, 2026
Last synced 9/19/2026