FitMyLLM
DeepSeek/Mixture of Experts

DeepSeekDeepSeek V4 Flash 0731

DeepSeek V4 Flash 0731 — the July re-post-train of V4 Flash, same 284B / 13B-active architecture with large gains on agentic and coding work. The checkpoint weighs 304B because it ships a DSpark speculative-decoding draft module on top of the base model.

chatcodingreasoningmultilingualmathagentictool_use
304.18B
Parameters (13B active)
1024K
Context length
9
Benchmarks
17
Quantizations
1.9M
HF downloads
Architecture
MoE
Released
2026-07-31
Layers
43
KV Heads
1
Head Dim
512
Family
deepseek

Quantization Options

Context length:
QuantBitsVRAM @ 16KQuality
IQ2_XXS2.38
91.1 GB
91.0 + 0.1 KV
low
IQ2_M2.93
112.0 GB
111.9 + 0.1 KV
low
Q2_K3.16
120.7 GB
120.6 + 0.1 KV
low
IQ3_XXS3.25
124.1 GB
124.1 + 0.1 KV
low
IQ3_XS3.5
133.6 GB
133.6 + 0.1 KV
low
Q3_K_S3.64
139.0 GB
138.9 + 0.1 KV
low
IQ3_M3.76
143.5 GB
143.5 + 0.1 KV
low
Q3_K_M4
152.7 GB
152.6 + 0.1 KV
low
Q3_K_L4.3
164.1 GB
164.0 + 0.1 KV
moderate
IQ4_XS4.46
170.1 GB
170.1 + 0.1 KV
moderate
Q4_K_S4.67
178.1 GB
178.1 + 0.1 KV
moderate
Q4_K_M4.89
186.5 GB
186.4 + 0.1 KV
good
Q5_K_S5.57
212.4 GB
212.3 + 0.1 KV
good
Q5_K_M5.7
217.3 GB
217.2 + 0.1 KV
good
Q6_K6.56
250.0 GB
249.9 + 0.1 KV
excellent
Q8_08.5
323.8 GB
323.7 + 0.1 KV
lossless
FP1616
608.9 GB
608.8 + 0.1 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 ~91 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 Flash 0731 now

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

Community Ratings

Loading ratings...

Benchmarks (9)

Run this model

Easiest way to get started·Beginners
Docs ↗
curl -fsSL https://ollama.com/install.sh | sh
$ollama run hf.co/unsloth/DeepSeek-V4-Flash-0731-GGUF:Q4_K_M

Not in Ollama's library: Ollama pulls the GGUF from Hugging Face (unsloth/DeepSeek-V4-Flash-0731-GGUF). If Q4_K_M is not in that repo, pick a file listed there.

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

GPUs that can run this model

At Q4_K_M quantization. Sorted by minimum VRAM.

AMD Instinct MI300X
192 GB VRAM • 5300 GB/s
AMD
$15000
Apple M2 Ultra (192GB)
192 GB VRAM • 800 GB/s
APPLE
$5499
Apple M3 Ultra (192GB)
192 GB VRAM • 800 GB/s
APPLE
$6999
Apple M4 Ultra (192GB)
192 GB VRAM • 1092 GB/s
APPLE
$7499
AMD Radeon Instinct MI300A
192 GB VRAM • 10300 GB/s
AMD
$12000
AMD Radeon Instinct MI300X
192 GB VRAM • 10300 GB/s
AMD
$15000
AMD Radeon Instinct MI308X
192 GB VRAM • 10300 GB/s
AMD
$12000
Apple M5 Ultra (192GB)
192 GB VRAM • 1228 GB/s
APPLE
AMD Radeon Instinct MI325X
288 GB VRAM • 10300 GB/s
AMD
$20000
AMD Radeon Instinct MI350X
288 GB VRAM • 8190 GB/s
AMD
$25000
AMD Radeon Instinct MI355X
288 GB VRAM • 8190 GB/s
AMD
$30000
Apple M4 Ultra (384GB)
384 GB VRAM • 1092 GB/s
APPLE
$9999
Apple M5 Ultra (384GB)
384 GB VRAM • 1228 GB/s
APPLE

Find the best GPU for DeepSeek V4 Flash 0731

Build Hardware for DeepSeek V4 Flash 0731
Spec sheet

DeepSeek V4 Flash 0731 — 304.18B MoE.

Specifications
Parameters
304.18B (13B active)
Architecture
Mixture of Experts
Context Length
1024K tokens
Capabilities
chat, coding, reasoning, multilingual, math, agentic, tool_use
Release Date
2026-07-31
Provider
DeepSeek
Family
deepseek
VRAM requirements
QuantBPWVRAMQuality
IQ2_XXS2.3891.0 GB65%
IQ2_M2.93111.9 GB75%
Q2_K3.16120.6 GB78%
IQ3_XXS3.25124.1 GB82%
IQ3_XS3.5133.6 GB84%
Q3_K_S3.64138.9 GB85%
IQ3_M3.76143.5 GB86%
Q3_K_M4152.6 GB88%
Q3_K_L4.3164.0 GB90%
IQ4_XS4.46170.1 GB92%
Q4_K_S4.67178.1 GB93%
Q4_K_M4.89186.4 GB94%
Q5_K_S5.57212.3 GB96%
Q5_K_M5.7217.2 GB96%
Q6_K6.56249.9 GB97%
Q8_08.5323.7 GB100%
FP1616608.8 GB100%
Benchmark scores
GPQA Diamond90.8
HLE38.6
AA Intelligence51.8
AA Coding69.1
SciCode49.9
AA Long Context74.3
Terminal-Bench 2.178.7
Terminal-Bench 412.1
EQ-Bench Creative Writing v31441
Compatible GPUs
13 @ Q4_K_M