FITMYLLM · JULY 28, 2026 · OPEN METHODOLOGY · COMMUNITY BENCHMARKS
LIVE · UPDATED DAILY
Quantization Options Select your GPU for speed estimates Context length: 4K 8K 16K 32K 64K 128K
Quant Bits VRAM @ 16K Quality Q3_K_M 4 8.4 GB
6.5 + 1.9 KV
low Q3_K_L 4.3 8.8 GB
6.9 + 1.9 KV
moderate IQ4_XS 4.46 9.1 GB
7.2 + 1.9 KV
moderate Q4_K_S 4.67 9.4 GB
7.5 + 1.9 KV
moderate Q4_K_M 4.89 9.7 GB
7.8 + 1.9 KV
good Q5_K_S 5.57 10.7 GB
8.8 + 1.9 KV
good Q5_K_M 5.7 10.9 GB
9.0 + 1.9 KV
good Q6_K 6.56 12.2 GB
10.3 + 1.9 KV
excellent Q8_0 8.5 15.1 GB
13.2 + 1.9 KV
lossless FP16 16 26.4 GB
24.5 + 1.9 KV
lossless
Select your GPU above to see speed estimates and compatibility for each quantization.
Deploying for a team or in production? Size GPUs, cost & scaling in Enterprise → ▸ READY TO RUN THIS? RENT BY THE HOUR
RENT A GPU AND RUN PIXTRAL 12B 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 Q3_K_M — 6.5 GB VRAM Q3_K_L — 6.9 GB VRAM IQ4_XS — 7.2 GB VRAM Q4_K_S — 7.5 GB VRAM Q4_K_M — 7.8 GB VRAM Q5_K_S — 8.8 GB VRAM Q5_K_M — 9.0 GB VRAM Q6_K — 10.3 GB VRAM Q8_0 — 13.2 GB VRAM FP16 — 24.5 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 mistral:12b-q4_K_MCOPY
Tag may need adjustment — check ollama.com/library/mistral 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 Pixtral 12B
Build Hardware for Pixtral 12B Pixtral 12B — Mistral's first vision model. Understands images and code.
Read full model card ▸ COLOPHON FITMYLLM · INDEPENDENT · DATA-DRIVEN
FITMYLLM · EST. 2025 · © 2026
RECOMMENDATIONS FROM PUBLISHED MATH, CORRECTED BY THE COMMUNITY — 30.
▸ SPEC SHEET
Pixtral 12B — 12B Dense. ▸ SPECIFICATIONS
PARAMETERS 12B
ARCHITECTURE Dense Transformer
CONTEXT LENGTH 128K tokens
CAPABILITIES vision, chat, coding
RELEASE DATE 2024-09-17
PROVIDER Mistral AI
FAMILY mistral ▸ VRAM REQUIREMENTS
QUANT BPW VRAM QUALITY Q3_K_M 4 6.5 GB 88% Q3_K_L 4.3 6.9 GB 90% IQ4_XS 4.46 7.2 GB 92% Q4_K_S 4.67 7.5 GB 93% Q4_K_M 4.89 7.8 GB 94% Q5_K_S 5.57 8.8 GB 96% Q5_K_M 5.7 9.0 GB 96% Q6_K 6.56 10.3 GB 97% Q8_0 8.5 13.2 GB 100% FP16 16 24.5 GB 100%
§ 01 BENCHMARK SCORES
HumanEval 52.0
MMLU-PRO 58.0
MATH 42.0
IFEval 72.0
BBH 58.0
MMMU 52.5
GPQA 38.0
MUSR 18.0
MMBench 78.0
§ 02 RUN COMMAND
Run Pixtral 12B locally with Ollama — needs 7.8 GB VRAM at Q4_K_M:
$ ollama run pixtral:12b
§ 03 COMPATIBLE GPUs
30 @ Q4_K_M Feedback 