FITMYLLM · JULY 28, 2026 · OPEN METHODOLOGY · COMMUNITY BENCHMARKS
LIVE · UPDATED DAILY
Quantization Options Select your GPU for speed estimates Quant Bits VRAM @ 4K Quality IQ3_XXS 3.25 14.3 GB
low IQ3_XS 3.5 15.4 GB
low Q3_K_S 3.64 16.0 GB
low IQ3_M 3.76 16.5 GB
low Q3_K_M 4 17.5 GB
low Q3_K_L 4.3 18.8 GB
moderate IQ4_XS 4.46 19.4 GB
moderate Q4_K_S 4.67 20.3 GB
moderate Q4_K_M 4.89 21.3 GB
good Q5_K_S 5.57 24.2 GB
good Q5_K_M 5.7 24.7 GB
good Q6_K 6.56 28.4 GB
excellent Q8_0 8.5 36.6 GB
lossless FP16 16 68.5 GB
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 LLAVA-1.6 YI 34B 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 IQ3_XXS — 14.3 GB VRAM IQ3_XS — 15.4 GB VRAM Q3_K_S — 16.0 GB VRAM IQ3_M — 16.5 GB VRAM Q3_K_M — 17.5 GB VRAM Q3_K_L — 18.8 GB VRAM IQ4_XS — 19.4 GB VRAM Q4_K_S — 20.3 GB VRAM Q4_K_M — 21.3 GB VRAM Q5_K_S — 24.2 GB VRAM Q5_K_M — 24.7 GB VRAM Q6_K — 28.4 GB VRAM Q8_0 — 36.6 GB VRAM FP16 — 68.5 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 llava:34b-v1.6-q4_K_MCOPY
Downloads and runs automatically. Add --verbose for speed stats.
▸ 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 LLaVA-1.6 Yi 34B
Build Hardware for LLaVA-1.6 Yi 34B LLaVA-1.6 Yi 34B — most capable LLaVA variant. Detailed image analysis.
Read full model card ▸ COLOPHON FITMYLLM · INDEPENDENT · DATA-DRIVEN
FITMYLLM · EST. 2025 · © 2026
RECOMMENDATIONS FROM PUBLISHED MATH, CORRECTED BY THE COMMUNITY — 30.
▸ SPEC SHEET
LLaVA-1.6 Yi 34B — 34B Dense. ▸ SPECIFICATIONS
PARAMETERS 34B
ARCHITECTURE Dense Transformer
CONTEXT LENGTH 4K tokens
CAPABILITIES vision, chat
RELEASE DATE 2024-01-30
PROVIDER LLaVA Team
FAMILY other ▸ VRAM REQUIREMENTS
QUANT BPW VRAM QUALITY IQ3_XXS 3.25 14.3 GB 82% IQ3_XS 3.5 15.4 GB 84% Q3_K_S 3.64 16.0 GB 85% IQ3_M 3.76 16.5 GB 86% Q3_K_M 4 17.5 GB 88% Q3_K_L 4.3 18.8 GB 90% IQ4_XS 4.46 19.4 GB 92% Q4_K_S 4.67 20.3 GB 93% Q4_K_M 4.89 21.3 GB 94% Q5_K_S 5.57 24.2 GB 96% Q5_K_M 5.7 24.7 GB 96% Q6_K 6.56 28.4 GB 97% Q8_0 8.5 36.6 GB 100% FP16 16 68.5 GB 100%
§ 01 BENCHMARK SCORES
MMLU-PRO 58.0
MATH 16.5
IFEval 68.0
BBH 58.0
MMMU 46.7
GPQA 35.0
MUSR 18.0
MMBench 79.3
§ 02 RUN COMMAND
Run LLaVA-1.6 Yi 34B locally with Ollama — needs 21.3 GB VRAM at Q4_K_M:
$ ollama run llava:34b
§ 03 COMPATIBLE GPUs
30 @ Q4_K_M Feedback