bf16 Open on Hugging Face
Parameters
6.6B
Weights (est.)
≈ 26.4 GB
Mac memory (est.)
32 GB+
Size is estimated from the parameter count and quantisation. Memory allows a fifth extra for context plus 2 GB for the system; long contexts need more.
Run it
With mlx-lm on an Apple silicon Mac:
pip install mlx-lmmlx_lm.generate --model mlx-community/DeepSeek-V4-Flash-MTP-bf16 --prompt "Hello"
Details
- Base model
- deepseek-ai/DeepSeek-V4-Flash (quantized) · DeepSeek V4 Flash on the radar →
- Vendor
- DeepSeek
- Task
- Text generation
- Quantisation
- bf16
- Licence
- mit
- Downloads
- 130 in the last 30 days · 9.8K all time · 1 likes
- Published
- 2026-05-27 · updated 2026-06-01 (5mo ago)
- Languages
- en
- Tags
- deepseek_v4_mtp, mlx-vlm, deepseek-v4, deepseek-v4-flash, speculative-decoding, draft-model, bf16
Other MLX builds of deepseek-ai/DeepSeek-V4-Flash
| Build | Quantisation | Weights | Mac memory | Downloads |
|---|---|---|---|---|
| DeepSeek-V4-Flash-MTP-bf16 (this one) | bf16 | ≈ 26.4 GB | 32 GB+ | 130 |
| DeepSeek-V4-Flash | 8-bit | ≈ 53.9 GB | 64 GB+ | 673 |
| DeepSeek-V4-Flash-2bit-M-DQ | 2-bit | ≈ 89.1 GB | 128 GB+ | 129 |
| deepseek-ai-DeepSeek-V4-Flash-3bit | 3-bit | ≈ 125 GB | 192 GB+ | 102 |
| DeepSeek-V4-Flash-3bit-DQ | 3-bit | ≈ 125 GB | 192 GB+ | 102 |
| deepseek-ai-DeepSeek-V4-Flash-4bit | 4-bit | ≈ 160 GB | 192 GB+ | 252 |
| deepseek-ai-DeepSeek-V4-Flash-6bit | 6-bit | ≈ 231 GB | 512 GB+ | 132 |
| deepseek-ai-DeepSeek-V4-Flash-8bit | 8-bit | ≈ 302 GB | 512 GB+ | 642 |