DeepSeek V4 Flash

DeepSeek·Operational·Released 2026-07-31text
Context1,000k
Max output-
Input / 1M-
Output / 1M-
Cached in-
Benchmarks28

Benchmarks

Cited public sources · rank is exact-benchmark, apples-to-apples

Chess Puzzles (Epoch AI run)#29/9233% ↗
DeepSWE v1.1 (Datacurve)#20/2853.3% ↗
LiveBench Agentic Coding (LiveBench)#41/5646.8% ↗
LiveBench Coding (LiveBench)#42/5675% ↗
LiveBench Data Analysis (LiveBench)#10/5679.3% ↗
LiveBench Instruction Following (LiveBench)#37/5665.5% ↗
LiveBench Language (LiveBench)#32/5679.2% ↗
LiveBench Mathematics (LiveBench)#41/5686.8% ↗
LiveBench Reasoning (LiveBench)#28/5686.6% ↗
Mystery Game Puzzles (Epoch AI run)#19/6834% ↗
Terminal-Bench 4.0 (Vals AI)#22/4018.7% ↗
WeirdML (Håvard Tveit Ihle)#25/9263% ↗
LMArena Elo#72/1411436 ↗
SimpleBench (SimpleBench)#30/7361.1% ↗
Agents' Last Exam#5/525.2% ↗
AutomationBench Public25.1% ↗
Cybergym76.7% ↗
DeepSWE#2/354.4% ↗
NL2Repo#3/654.2% ↗
Terminal Bench 2.182.7% ↗
Toolathlon-Verified#4/470.3% ↗
Toolathlon-Verified (HKUST)#9/2370.7% ↗

DeepSeek V4 Flash FAQ

What is DeepSeek V4 Flash?
DeepSeek V4 Flash is a large language model from DeepSeek, released on July 31, 2026. It accepts text input.
When was DeepSeek V4 Flash released?
DeepSeek released DeepSeek V4 Flash on July 31, 2026.
What is DeepSeek V4 Flash's context window?
DeepSeek V4 Flash has a 1M-token context window.
How does DeepSeek V4 Flash score on benchmarks?
DeepSeek V4 Flash has 28 cited benchmark scores, including Chess Puzzles (Epoch AI run) 33%, DeepSWE v1.1 (Datacurve) 53.3% and FrontierMath Tier 4 v2 (Epoch AI run) 24.4%. Each score links to its source.

Access DeepSeek V4 Flash and every other model through one endpoint with automatic failover: Respan gateway.