DeepSeek/Mixture of Experts

DeepSeek V4.1 Flash

Name: DeepSeek V4.1 Flash
Author: DeepSeek

DeepSeek V4.1 Flash — a new natively multimodal DeepSeek MoE trained from scratch: 552B backbone with 16B active per decoded token, 1M context and a KV cache about a quarter of V4 Flash's. The checkpoint weighs 763B because it also carries 196B of Engram lookup tables and a speculative-decoding draft module.

chatcodingreasoningmultilingualvisionmathagentictool_use

763.21B

Parameters (16B active)

1024K

Context length

Benchmarks

Quantizations

2.1M

HF downloads

Architecture

MoE

Released

2026-09-10

Layers

KV Heads

Head Dim

512

Family

deepseek

Quantization Options

Context length:

Quant	Bits	VRAM @ 16K	Quality
IQ2_XXS	2.38	227.6 GB 227.5 + 0.0 KV	low
IQ2_M	2.93	280.1 GB 280.0 + 0.0 KV	low
Q2_K	3.16	302.0 GB 302.0 + 0.0 KV	low
IQ3_XXS	3.25	310.6 GB 310.5 + 0.0 KV	low
IQ3_XS	3.5	334.4 GB 334.4 + 0.0 KV	low
Q3_K_S	3.64	347.8 GB 347.7 + 0.0 KV	low
IQ3_M	3.76	359.2 GB 359.2 + 0.0 KV	low
Q3_K_M	4	382.1 GB 382.1 + 0.0 KV	low
Q3_K_L	4.3	410.8 GB 410.7 + 0.0 KV	moderate
IQ4_XS	4.46	426.0 GB 426.0 + 0.0 KV	moderate
Q4_K_S	4.67	446.0 GB 446.0 + 0.0 KV	moderate
Q4_K_M	4.89	467.0 GB 467.0 + 0.0 KV	good
Q5_K_S	5.57	531.9 GB 531.9 + 0.0 KV	good
Q5_K_M	5.7	544.3 GB 544.3 + 0.0 KV	good
Q6_K	6.56	626.4 GB 626.3 + 0.0 KV	excellent
Q8_0	8.5	811.4 GB 811.4 + 0.0 KV	lossless
FP16	16	1526.9 GB 1526.9 + 0.0 KV	lossless

Select your GPU above to see speed estimates and compatibility for each quantization.

Too big for a single GPU — plan a multi-GPU deployment

Even the lightest quant needs ~228 GB. Size GPUs, replicas, TCO and scaling for a production setup. Open in Enterprise →

▸ READY TO RUN THIS?RENT BY THE HOUR

RENT A GPU AND RUN DEEPSEEK V4.1 FLASH NOW

Rent on Vast.ai →Or RunPod →

Spin up an A100 / H100 / 4090 in ~60s. Pay by the second. Cancel anytime.

Community Ratings

Loading ratings...

Benchmarks (6)

EQ-Bench Creative Writing1540

GPQA Diamond90.9

AA Long Context84.0

SciCode51.9

HLE39.2

Terminal-Bench 426.8

Run this model

▸Easiest way to get started·Beginners

DOCS ↗

curl -fsSL https://ollama.com/install.sh | sh

$ollama run deepseek:763b-q4_K_M

Tag may need adjustment — check ollama.com/library/deepseek for available tags.

▸ SETUP GUIDE

Auto-setup with fitmyllm CLI

Detects your GPU, recommends the best model, downloads it, and starts chatting — zero config. Benchmarks your speed and contributes anonymous data to improve predictions.

pip install fitmyllmthen run fitmyllmLearn more

Auto-detect GPULive tok/s in chatSpeed benchmarks9 inference engines

HuggingFace GGUF Downloads Build Hardware

Find the best GPU for DeepSeek V4.1 Flash

Build Hardware for DeepSeek V4.1 Flash

QUANT	BPW	VRAM	QUALITY
IQ2_XXS	2.38	227.5 GB	65%
IQ2_M	2.93	280.0 GB	75%
Q2_K	3.16	302.0 GB	78%
IQ3_XXS	3.25	310.5 GB	82%
IQ3_XS	3.5	334.4 GB	84%
Q3_K_S	3.64	347.7 GB	85%
IQ3_M	3.76	359.2 GB	86%
Q3_K_M	4	382.1 GB	88%
Q3_K_L	4.3	410.7 GB	90%
IQ4_XS	4.46	426.0 GB	92%
Q4_K_S	4.67	446.0 GB	93%
Q4_K_M	4.89	467.0 GB	94%
Q5_K_S	5.57	531.9 GB	96%
Q5_K_M	5.7	544.3 GB	96%
Q6_K	6.56	626.3 GB	97%
Q8_0	8.5	811.4 GB	100%
FP16	16	1526.9 GB	100%