TokenAssemble

DeepSeek · open-weights model

DeepSeek-V4-Flash

DeepSeek-V4-Flash is a 284B-parameter model (MoE, 13B active) by DeepSeek. Its smallest quant (UD-Q2_K_XL) needs at least 98.0GB of GPU-addressable memory at 8K context; it loads cleanly at a quality quant on 2 of the 50 machines we track.

Parameters
284B total · 13B active
Layers
43
Max context
1024K
Attention
HYBRID
License
mit

Config from deepseek-ai/DeepSeek-V4-Flash

Quants and what they need

Measured file sizes; “needs” = weights + KV cache at 8K context + overhead.

QuantFile sizeNeeds ≥
UD-Q8_K_XL161.9 GB163.1 GB
UD-Q4_K_XL155.1 GB156.3 GB
UD-Q3_K_XL129.4 GB130.7 GB
UD-Q2_K_XL96.8 GB98.0 GB

What runs DeepSeek-V4-Flash

Every machine we track, graded at its best clean-fit quant.

HardwareBest-quant verdict · llama.cpp · 8K context
Intel Arc A770 16 GBWon't fitbeta
Intel Arc B580Won't fitbeta
AMD Radeon PRO W7900Tight · C · UD-Q2_K_XLinteractive · offloadbeta
NVIDIA GeForce RTX 3060 12 GBWon't fitbeta
NVIDIA GeForce RTX 3070Won't fitbeta
NVIDIA GeForce RTX 3070 TiWon't fitbeta
NVIDIA GeForce RTX 3080 10 GBWon't fitbeta
NVIDIA GeForce RTX 3080 12 GBWon't fitbeta
NVIDIA GeForce RTX 3080 TiWon't fitbeta
NVIDIA GeForce RTX 3090Won't fitbeta
NVIDIA GeForce RTX 3090 TiWon't fitbeta
NVIDIA GeForce RTX 4060Won't fitbeta
NVIDIA GeForce RTX 4060 Ti 16 GBWon't fitbeta
NVIDIA GeForce RTX 4060 Ti 8 GBWon't fitbeta
NVIDIA GeForce RTX 4070Won't fitbeta
NVIDIA GeForce RTX 4070 SUPERWon't fitbeta
NVIDIA GeForce RTX 4070 TiWon't fitbeta
NVIDIA GeForce RTX 4070 Ti SUPERWon't fitbeta
NVIDIA GeForce RTX 4080Won't fitbeta
NVIDIA GeForce RTX 4080 SUPERWon't fitbeta
NVIDIA GeForce RTX 4090Won't fitbeta
NVIDIA RTX 5000 Ada GenerationWon't fitbeta
NVIDIA GeForce RTX 5060 Ti 16 GBWon't fitbeta
NVIDIA GeForce RTX 5070Won't fitbeta
NVIDIA GeForce RTX 5070 TiWon't fitbeta
NVIDIA GeForce RTX 5080Won't fitbeta
NVIDIA GeForce RTX 5090Won't fitbeta
NVIDIA RTX 6000 Ada GenerationTight · C · UD-Q2_K_XLinteractive · offloadbeta
NVIDIA RTX A6000Tight · C · UD-Q2_K_XLinteractive · offloadbeta
AMD Radeon RX 6800Won't fitbeta
AMD Radeon RX 6900 XTWon't fitbeta
AMD Radeon RX 7800 XTWon't fitbeta
AMD Radeon RX 7900 XTWon't fitbeta
AMD Radeon RX 7900 XTXWon't fitbeta
AMD Radeon RX 9070Won't fitbeta
AMD Radeon RX 9070 XTWon't fitbeta
ASUS Ascent GX10Fits · B · UD-Q2_K_XLinteractivebeta
Beelink GTR9 Pro Ryzen AI Max+ 395 128GBWon't fitbeta
NVIDIA DGX SparkFits · B · UD-Q2_K_XLinteractivebeta
Framework Desktop Ryzen AI Max+ 395 128GBWon't fitbeta
Framework Desktop Ryzen AI Max+ 395 64GBWon't fitbeta
GMKtec EVO-X2 Ryzen AI Max+ 395 128GBWon't fitbeta
HP Z2 Mini G1a Ryzen AI Max+ PRO 395 128GBWon't fitbeta
Mac mini M4 16GBWon't fitbeta
Mac mini M4 24GBWon't fitbeta
Mac mini M4 Pro 24GBWon't fitbeta
Mac mini M4 Pro 48GBWon't fitbeta
Mac Studio M3 Ultra 96GBWon't fitbeta
Mac Studio M4 Max 36GBWon't fitbeta
Mac Studio M4 Max 64GBWon't fitbeta

Verdicts computed from published specs and measured GGUF sizes — how the math works.