AI Model Release Tracker

DeepSeek-V4-FlashOpen Weight

Released

DeepSeek-V4-Flash is an AI model released by DeepSeek on Friday, Apr 24 2026, the same day as DeepSeek-V4-Pro. It is an open-weight model — the trained weights are available to download and run. It is a 284B parameterThe learned values that shape how a model responds. B means billion and T means trillion. For mixture-of-experts models, the total count is larger than the number active for each token. Parameter count alone does not measure quality. model with a 1M token context windowHow much information the model can work with in one request, measured in tokens — pieces of words or other data. This includes your question, attachments, conversation and reply. Limits can vary by model variant and provider.. Benchmark results (shown below) cover BullshitBench v2, LiveCodeBench, and GPQA Diamond.

Available from

StreamLake—$0.0826$0.16521Mfp8
Wafer—$0.70$0.171Mmxfp4
DeepInfra—$0.09$0.181Mfp8
GMICloud—$0.091$0.1821Mfp8
Venice—$0.0966$0.19251M—
DigitalOcean—$0.098$0.1961M—
SiliconFlow—$0.13$0.281Mfp8
AtlasCloud—$0.14$0.281Mfp4
Baidu—$0.14$0.281Mfp8
Parasail—$0.14$0.281Mfp8
Alibabaus-east-1$0.15$0.301Mfp8
Relace—$0.05$0.451Mfp4
Mancer 2—$0.19$0.501Mfp8
Cloudflare—$0.44$1.32400K—
USD per 1M tokensEvery price here is for one million tokens. A token is roughly three-quarters of a word, so a million tokens is about 750,000 words of text.Checked October 21, 2026Third-party rates from openrouter.ai, not the lab’s own list price.

Benchmarks

Coding

LiveCodeBenchCompetitive coding — Coding problems published so recently the AI can't have seen them in training — a contamination-free test of raw programming skill. Higher is better.via BenchLM
91.6%
#3 of 23

Reasoning & science

GPQA DiamondScience — Graduate-level science questions in biology, physics, and chemistry — hard enough that subject-matter PhDs score around 65%. Higher is better.
88.1%
#23 of 66

Robustness

BullshitBench v2Nonsense detection — Given a confidently-worded but nonsensical prompt, does the AI spot that it makes no sense and push back — instead of playing along and inventing an answer? The score is how often it clearly called out the nonsense. Higher is better.
18%
#76 of 97

Source: BenchLM, retrieved 11 July 2026. Other sources are identified on the linked benchmark pages.

Compare DeepSeek-V4-Flash with

DeepSeek-V4-Flash

Suggested comparisons

DeepSeek-V4-FlashvsDeepSeek-V4.1-Flash-1017DeepSeek-V4-FlashvsGPT-6.1 SolDeepSeek-V4-FlashvsClaude Haiku 5.5DeepSeek-V4-FlashvsGemini 4 ArgonDeepSeek-V4-FlashvsMuse Spark OpenDeepSeek-V4-FlashvsGrok 4.7DeepSeek-V4-FlashvsMistral Large 4DeepSeek-V4-FlashvsKimi K3DeepSeek-V4-FlashvsGLM-5.3-FlashDeepSeek-V4-FlashvsQwen3.8-Max-0902DeepSeek-V4-FlashvsNemotron 3.5 Super VL

About

DeepSeek V4 Flash was introduced in the April 24, 2026 V4 preview alongside the larger V4 Pro. Flash used a mixture-of-experts design with 284 billion total parameters and about 13 billion active per token. DeepSeek positioned it as the faster, less expensive V4 option, with a one-million-token context window and thinking and non-thinking modes.

The V4 generation combined token-level compression with sparse attention to address the memory and computation costs of long contexts. Flash reached the assistant and API with downloadable weights, giving developers hosted and self-managed deployment paths. This entry records the original April preview checkpoint.

Frequently asked questions

DeepSeek-V4-Flash was released by DeepSeek on Friday, Apr 24 2026.

All DeepSeek releases

24 tracked