FITMYLLM · SEPTEMBER 6, 2026 · OPEN METHODOLOGY · COMMUNITY BENCHMARKS
LIVE · UPDATED DAILY
304.18B
Parameters (13B active)
Quantization Options Select your GPU for speed estimates Context length: 4K 8K 16K 32K 64K 128K 256K 512K 1M
Quant Bits VRAM @ 16K Quality IQ2_XXS 2.38 92.0 GB
91.0 + 1.0 KV
low IQ2_M 2.93 112.9 GB
111.9 + 1.0 KV
low Q2_K 3.16 121.6 GB
120.6 + 1.0 KV
low IQ3_XXS 3.25 125.1 GB
124.1 + 1.0 KV
low IQ3_XS 3.5 134.6 GB
133.6 + 1.0 KV
low Q3_K_S 3.64 139.9 GB
138.9 + 1.0 KV
low IQ3_M 3.76 144.5 GB
143.5 + 1.0 KV
low Q3_K_M 4 153.6 GB
152.6 + 1.0 KV
low Q3_K_L 4.3 165.0 GB
164.0 + 1.0 KV
moderate IQ4_XS 4.46 171.1 GB
170.1 + 1.0 KV
moderate Q4_K_S 4.67 179.1 GB
178.1 + 1.0 KV
moderate Q4_K_M 4.89 187.4 GB
186.4 + 1.0 KV
good Q5_K_S 5.57 213.3 GB
212.3 + 1.0 KV
good Q5_K_M 5.7 218.2 GB
217.2 + 1.0 KV
good Q6_K 6.56 250.9 GB
249.9 + 1.0 KV
excellent Q8_0 8.5 324.7 GB
323.7 + 1.0 KV
lossless FP16 16 609.9 GB
608.8 + 1.0 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 ~92 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 Chat Coding Reasoning Creative Vision Roleplay Agentic
Loading ratings...
Run this model IQ2_XXS — 91.0 GB VRAM IQ2_M — 111.9 GB VRAM Q2_K — 120.6 GB VRAM IQ3_XXS — 124.1 GB VRAM IQ3_XS — 133.6 GB VRAM Q3_K_S — 138.9 GB VRAM IQ3_M — 143.5 GB VRAM Q3_K_M — 152.6 GB VRAM Q3_K_L — 164.0 GB VRAM IQ4_XS — 170.1 GB VRAM Q4_K_S — 178.1 GB VRAM Q4_K_M — 186.4 GB VRAM Q5_K_S — 212.3 GB VRAM Q5_K_M — 217.2 GB VRAM Q6_K — 249.9 GB VRAM Q8_0 — 323.7 GB VRAM FP16 — 608.8 GB VRAM
Ollama llama.cpp vLLM LM Studio KoboldCpp Jan Docker
▸ Easiest way to get started · Beginners
DOCS ↗ curl -fsSL https://ollama.com/install.sh | shCOPY
$ ollama run deepseek:304b-q4_K_MCOPY
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.
Auto-detect GPU Live tok/s in chat Speed benchmarks 9 inference engines
GPUs that can run this model At Q4_K_M quantization. Sorted by minimum VRAM.
Find the best GPU for DeepSeek V4 Flash 0731
Build Hardware for DeepSeek V4 Flash 0731 ▸ COLOPHON FITMYLLM · INDEPENDENT · DATA-DRIVEN
FITMYLLM · EST. 2025 · © 2026
RECOMMENDATIONS FROM PUBLISHED MATH, CORRECTED BY THE COMMUNITY — 30.
▸ 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
QUANT BPW VRAM QUALITY IQ2_XXS 2.38 91.0 GB 65% IQ2_M 2.93 111.9 GB 75% Q2_K 3.16 120.6 GB 78% IQ3_XXS 3.25 124.1 GB 82% IQ3_XS 3.5 133.6 GB 84% Q3_K_S 3.64 138.9 GB 85% IQ3_M 3.76 143.5 GB 86% Q3_K_M 4 152.6 GB 88% Q3_K_L 4.3 164.0 GB 90% IQ4_XS 4.46 170.1 GB 92% Q4_K_S 4.67 178.1 GB 93% Q4_K_M 4.89 186.4 GB 94% Q5_K_S 5.57 212.3 GB 96% Q5_K_M 5.7 217.2 GB 96% Q6_K 6.56 249.9 GB 97% Q8_0 8.5 323.7 GB 100% FP16 16 608.8 GB 100%
§ 01 BENCHMARK SCORES
GPQA Diamond 90.8
HLE 38.6
AA Intelligence 51.8
AA Coding 69.1
aa_scicode 49.9
aa_lcr 74.3
§ 03 COMPATIBLE GPUs
13 @ Q4_K_M Feedback 