What is DeepSeek-V4-Flash?
DeepSeek-V4-Flash is a 284B parameter Mixture-of-Experts model from DeepSeek, published as a preview of the V4 series alongside the 1.6T DeepSeek-V4-Pro. It activates 13B parameters per token and supports a context of one million tokens. It is the smaller of the two V4 models and the one within reach of a big workstation: the smallest GGUF build is 82.6 GB. The weights went up on Hugging Face on April 22, 2026 under the MIT license.
| Specification | DeepSeek-V4-Flash |
|---|---|
| Total parameters | 284B |
| Activated parameters | 13B per token |
| Architecture | MoE, hybrid attention (CSA + HCA) |
| Context window | 1M tokens |
| Modalities | Text |
| Reasoning modes | Non-think, Think High, Think Max |
| Pre-training data | More than 32T tokens |
| Precision | FP4 + FP8 mixed (MoE experts in FP4, most other weights FP8) |
| Release date | April 22, 2026 |
| License | MIT |
The million-token context comes from a hybrid attention design that pairs Compressed Sparse Attention with Heavily Compressed Attention. DeepSeek quotes the larger V4-Pro at 27% of the single-token inference FLOPs and 10% of the KV cache of DeepSeek-V3.2 at a 1M-token context, and both V4 models share the architecture. Training used the Muon optimizer, Manifold-Constrained Hyper-Connections to strengthen conventional residual connections, and more than 32T tokens of pre-training data. Post-training ran in two stages: domain experts trained separately with SFT and RL, then merged into a single model through on-policy distillation. One practical note: the repo ships Python encoding scripts instead of a Jinja chat template, which matters if you wire the model up yourself but not if an app handles the template for you.
DeepSeek-V4-Flash benchmarks
DeepSeek published launch numbers on the model card comparing V4-Flash across its three reasoning modes, with the 1.6T V4-Pro in Think Max mode as the reference:
| Benchmark | V4-Flash Non-Think | V4-Flash High | V4-Flash Max | V4-Pro Max |
|---|---|---|---|---|
MMLU-Pro Academic knowledge | 83.0 | 86.4 | 86.2 | 87.5 |
GPQA Diamond Expert science | 71.2 | 87.4 | 88.1 | 90.1 |
LiveCodeBench Competitive coding | 55.2 | 88.4 | 91.6 | 93.5 |
HMMT Feb 2026 Olympiad math | 40.8 | 91.9 | 94.8 | 95.2 |
MRCR 1M Long context | 37.5 | 76.9 | 78.7 | 83.5 |
SWE-bench Verified Software engineering | 73.7 | 78.6 | 79.0 | 80.6 |
Terminal Bench 2.0 Terminal agents | 49.1 | 56.6 | 56.9 | 67.9 |
Think Max brings the 13B-active Flash within a point or two of the flagship on math and code: 94.8 against 95.2 on HMMT, 91.6 against 93.5 on LiveCodeBench. The gap is wider on agentic work, where Pro Max leads Terminal Bench 2.0 by 11 points. The Non-think column shows how much of the score comes from thinking: 40.8 to 94.8 on HMMT.
DeepSeek-V4-Flash hardware requirements
The system requirement to check is memory. DeepSeek ships the original weights in mixed FP4 + FP8; the GGUF builds below come from unsloth/DeepSeek-V4-Flash-GGUF, each split into shards of up to 50 GB, listed here as totals.
| Memory | Build to pick | File size |
|---|---|---|
| 96 GB | UD-IQ1_S | 82.6 GB |
| 128 GB | UD-IQ3_XXS | 103.0 GB |
| 192 GB | UD-Q4_K_XL | 155.1 GB |
| 256 GB and up | UD-Q8_K_XL | 161.9 GB |
When two builds both fit, take the larger one. Only 13B of the 284B parameters are active per token, so generation speed holds up better than the file sizes suggest once the weights are in memory. Budget headroom for context on top of the weights: DeepSeek recommends at least 384K tokens of context for Think Max mode. If the format is new to you, start with what GGUF is.
How to run DeepSeek-V4-Flash in Atomic Chat
Atomic Chat is a free local app for macOS, Windows and Linux. It includes a Hugging Face model browser and a built-in chat, with no manual llama.cpp build required.
- Download Atomic Chat for your platform and open it.
- Search for DeepSeek-V4-Flash in the model browser and open Download Options.
- Pick the build that fits the memory you have, then start a chat.
For the rest of the lineup, see every DeepSeek model you can run locally, or the 1.6T flagship DeepSeek-V4-Pro.
DeepSeek-V4-Flash license
DeepSeek-V4-Flash is released under the MIT license, one of the most permissive licenses in open source. It permits commercial use, modification and redistribution with no royalties, so you can build products on top of the model and run it on your own hardware without a usage fee.
