FITMYLLM · AUGUST 17, 2026 · OPEN METHODOLOGY · COMMUNITY BENCHMARKS
LIVE · UPDATED DAILY
51.6B
Parameters (12B active)
Quantization Options Select your GPU for speed estimates Context length: 4K 8K 16K 32K 64K 128K 256K
Quant Bits VRAM @ 16K Quality IQ3_XXS 3.25 23.0 GB
21.5 + 1.5 KV
low IQ3_XS 3.5 24.6 GB
23.1 + 1.5 KV
low Q3_K_S 3.64 25.5 GB
24.0 + 1.5 KV
low IQ3_M 3.76 26.2 GB
24.7 + 1.5 KV
low Q3_K_M 4 27.8 GB
26.3 + 1.5 KV
low Q3_K_L 4.3 29.7 GB
28.2 + 1.5 KV
moderate IQ4_XS 4.46 30.8 GB
29.3 + 1.5 KV
moderate Q4_K_S 4.67 32.1 GB
30.6 + 1.5 KV
moderate Q4_K_M 4.89 33.5 GB
32.0 + 1.5 KV
good Q5_K_S 5.57 37.9 GB
36.4 + 1.5 KV
good Q5_K_M 5.7 38.8 GB
37.3 + 1.5 KV
good Q6_K 6.56 44.3 GB
42.8 + 1.5 KV
excellent Q8_0 8.5 56.8 GB
55.3 + 1.5 KV
lossless FP16 16 105.2 GB
103.7 + 1.5 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 JAMBA 1.5 MINI 52B 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 — 21.5 GB VRAM IQ3_XS — 23.1 GB VRAM Q3_K_S — 24.0 GB VRAM IQ3_M — 24.7 GB VRAM Q3_K_M — 26.3 GB VRAM Q3_K_L — 28.2 GB VRAM IQ4_XS — 29.3 GB VRAM Q4_K_S — 30.6 GB VRAM Q4_K_M — 32.0 GB VRAM Q5_K_S — 36.4 GB VRAM Q5_K_M — 37.3 GB VRAM Q6_K — 42.8 GB VRAM Q8_0 — 55.3 GB VRAM FP16 — 103.7 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 jamba:52b-q4_K_MCOPY
Tag may need adjustment — check ollama.com/library/jamba 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 Jamba 1.5 Mini 52B
Build Hardware for Jamba 1.5 Mini 52B AI21's Jamba 1.5 Mini — hybrid SSM-Transformer MoE. Fast with 256K context window.
Read full model card ▸ COLOPHON FITMYLLM · INDEPENDENT · DATA-DRIVEN
FITMYLLM · EST. 2025 · © 2026
RECOMMENDATIONS FROM PUBLISHED MATH, CORRECTED BY THE COMMUNITY — 30.
▸ SPEC SHEET
Jamba 1.5 Mini 52B — 51.6B MoE. ▸ SPECIFICATIONS
PARAMETERS 51.6B (12B active)
ARCHITECTURE Mixture of Experts
CONTEXT LENGTH 256K tokens
CAPABILITIES chat, reasoning
RELEASE DATE 2024-08-22
PROVIDER AI21 Labs
FAMILY jamba ▸ VRAM REQUIREMENTS
QUANT BPW VRAM QUALITY IQ3_XXS 3.25 21.5 GB 82% IQ3_XS 3.5 23.1 GB 84% Q3_K_S 3.64 24.0 GB 85% IQ3_M 3.76 24.7 GB 86% Q3_K_M 4 26.3 GB 88% Q3_K_L 4.3 28.2 GB 90% IQ4_XS 4.46 29.3 GB 92% Q4_K_S 4.67 30.6 GB 93% Q4_K_M 4.89 32.0 GB 94% Q5_K_S 5.57 36.4 GB 96% Q5_K_M 5.7 37.3 GB 96% Q6_K 6.56 42.8 GB 97% Q8_0 8.5 55.3 GB 100% FP16 16 103.7 GB 100%
§ 01 BENCHMARK SCORES
HumanEval 64.0
MMLU-PRO 42.0
MATH 57.0
IFEval 64.0
BBH 10.7
GPQA 2.5
MUSR 3.7
GPQA Diamond 30.2
LiveCodeBench 6.2
AIME 1.0
MATH-500 35.7
HLE 5.1
AA Intelligence 8.0
aa_scicode 8.0
§ 03 COMPATIBLE GPUs
30 @ Q4_K_M Feedback