FITMYLLM · NOVEMBER 4, 2026 · OPEN METHODOLOGY · COMMUNITY BENCHMARKS
LIVE · UPDATED DAILY
304.65B
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 91.2 GB
91.1 + 0.1 KV
low IQ2_M 2.93 112.1 GB
112.1 + 0.1 KV
low Q2_K 3.16 120.9 GB
120.8 + 0.1 KV
low IQ3_XXS 3.25 124.3 GB
124.3 + 0.1 KV
low IQ3_XS 3.5 133.9 GB
133.8 + 0.1 KV
low Q3_K_S 3.64 139.2 GB
139.1 + 0.1 KV
low IQ3_M 3.76 143.8 GB
143.7 + 0.1 KV
low Q3_K_M 4 152.9 GB
152.8 + 0.1 KV
low Q3_K_L 4.3 164.3 GB
164.2 + 0.1 KV
moderate IQ4_XS 4.46 170.4 GB
170.3 + 0.1 KV
moderate Q4_K_S 4.67 178.4 GB
178.3 + 0.1 KV
moderate Q4_K_M 4.89 186.8 GB
186.7 + 0.1 KV
good Q5_K_S 5.57 212.7 GB
212.6 + 0.1 KV
good Q5_K_M 5.7 217.6 GB
217.6 + 0.1 KV
good Q6_K 6.56 250.4 GB
250.3 + 0.1 KV
excellent Q8_0 8.5 324.3 GB
324.2 + 0.1 KV
lossless FP16 16 609.9 GB
609.8 + 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 VISION EXP 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 — 91.1 GB VRAM IQ2_M — 112.1 GB VRAM Q2_K — 120.8 GB VRAM IQ3_XXS — 124.3 GB VRAM IQ3_XS — 133.8 GB VRAM Q3_K_S — 139.1 GB VRAM IQ3_M — 143.7 GB VRAM Q3_K_M — 152.8 GB VRAM Q3_K_L — 164.2 GB VRAM IQ4_XS — 170.3 GB VRAM Q4_K_S — 178.3 GB VRAM Q4_K_M — 186.7 GB VRAM Q5_K_S — 212.6 GB VRAM Q5_K_M — 217.6 GB VRAM Q6_K — 250.3 GB VRAM Q8_0 — 324.2 GB VRAM FP16 — 609.8 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:305b-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 Vision Exp
Build Hardware for DeepSeek V4 Flash Vision Exp ▸ COLOPHON FITMYLLM · INDEPENDENT · DATA-DRIVEN
FITMYLLM · EST. 2025 · © 2026
RECOMMENDATIONS FROM PUBLISHED MATH, CORRECTED BY THE COMMUNITY — 30.
▸ SPEC SHEET
DeepSeek V4 Flash Vision Exp — 304.65B MoE. ▸ SPECIFICATIONS
PARAMETERS 304.65B (13B active)
ARCHITECTURE Mixture of Experts
CONTEXT LENGTH 1024K tokens
CAPABILITIES chat, coding, reasoning, multilingual, vision, math, agentic, tool_use
RELEASE DATE 2026-08-31
PROVIDER DeepSeek
FAMILY deepseek ▸ VRAM REQUIREMENTS
QUANT BPW VRAM QUALITY IQ2_XXS 2.38 91.1 GB 65% IQ2_M 2.93 112.1 GB 75% Q2_K 3.16 120.8 GB 78% IQ3_XXS 3.25 124.3 GB 82% IQ3_XS 3.5 133.8 GB 84% Q3_K_S 3.64 139.1 GB 85% IQ3_M 3.76 143.7 GB 86% Q3_K_M 4 152.8 GB 88% Q3_K_L 4.3 164.2 GB 90% IQ4_XS 4.46 170.3 GB 92% Q4_K_S 4.67 178.3 GB 93% Q4_K_M 4.89 186.7 GB 94% Q5_K_S 5.57 212.6 GB 96% Q5_K_M 5.7 217.6 GB 96% Q6_K 6.56 250.3 GB 97% Q8_0 8.5 324.2 GB 100% FP16 16 609.8 GB 100%
§ 01 BENCHMARK SCORES
GPQA Diamond 91.3
HLE 34.5
aa_scicode 49.7
aa_lcr 81.3
aa_terminal_bench_v2 74.2
aa_terminal_bench_v4 12.1
§ 03 COMPATIBLE GPUs
13 @ Q4_K_M Feedback 