DeepSeek logo

DeepSeek-V4-Flash-Max

DeepSeek

DeepSeek-V4-Flash-Max is the highest-effort variant of the DeepSeek-V4-Flash line, released in April 2026. It tops DeepSeek's competitive programming results and delivers near-perfect scores on CodeForces while retaining an efficient MoE architecture. It is served first-party by DeepSeek's own API.

Key Specifications

Parameters
284.0B
Context
1.0M
Release Date
April 23, 2026
Average Score
72.8%

Timeline

Key dates in the model's history
Announcement
April 23, 2026
Last Update
August 27, 2026
Today
September 11, 2026

Technical Specifications

Parameters
284.0B
Training Tokens
-
Knowledge Cutoff
-
Family
-
Capabilities
MultimodalZeroEval

Pricing & Availability

Input (per 1M tokens)
$0.14
Output (per 1M tokens)
$0.28
Max Input Tokens
1.0M
Max Output Tokens
393.2K
Supported Features
Function CallingStructured OutputCode ExecutionWeb SearchBatch InferenceFine-tuning

Benchmark Results

Model performance metrics across various tests and benchmarks

Programming

Programming skills tests
SWE-Bench Verified
Self-reported
79.0%

Reasoning

Logical reasoning and analysis
GPQA
Diamond (Pass@1)Self-reported
88.1%

Other Tests

Specialized benchmarks
CodeForces
RatingSelf-reported
100.0%
HMMT Feb 26
Pass@1Self-reported
95.0%
LiveCodeBench
Pass@1Self-reported
92.0%
IMO-AnswerBench
Pass@1Self-reported
88.0%
BrowseComp
Pass@1Self-reported
73.2%
CorpusQA 1M
AccuracySelf-reported
60.5%
CSimpleQA
Pass@1Self-reported
78.9%
Humanity's Last Exam
Pass@1 (w/ tools)Self-reported
45.1%
MathArena Apex
Shortlist (Pass@1)Self-reported
85.7%
MCP Atlas
Pass@1Self-reported
69.0%
MMLU-Pro
MaxSelf-reported
86.2%
MRCR 1M
MMRSelf-reported
78.7%
SimpleQA
Verified (Pass@1)Self-reported
34.1%
SWE-bench Multilingual
Self-reported
73.3%
SWE-Bench Pro
Self-reported
52.6%
Terminal-Bench 2.0
Self-reported
56.9%
Toolathlon
Pass@1Self-reported
47.8%

License & Metadata

License
mit
Announcement Date
April 23, 2026
Last Updated
August 27, 2026

Similar Models

All Models

Recommendations are based on similarity of characteristics: developer organization, multimodality, parameter size, and benchmark performance. Choose a model to compare or go to the full catalog to browse all available AI models.