AI21 Jamba Mini 1.5 — Hardware Requirements & GPU Compatibility
Chat
10.0K downloads
268 likesSpecifications
- Publisher
- AI21 Labs
- Parameters
- 51.6B
- Release Date
- 2026-02-02
- License
- Other
Get Started
HuggingFace
How Much VRAM Does AI21 Jamba Mini 1.5 Need?
Select a quantization to see compatible GPUs below.
| Quantization | Bits | VRAM | + Context | File Size | Quality |
|---|---|---|---|---|---|
| BF16 | 16.00 | 113.5 GB | — | 103.14 GB | Brain floating point 16 — preferred for training |
Which GPUs Can Run AI21 Jamba Mini 1.5?
BF16 · 113.5 GBAI21 Jamba Mini 1.5 (BF16) requires 113.5 GB of VRAM to load the model weights. For comfortable inference with headroom for KV cache and system overhead, 148+ GB is recommended. No single GPU has enough memory — multi-GPU or cluster setups are needed.
Which Devices Can Run AI21 Jamba Mini 1.5?
BF16 · 113.5 GB5 devices with unified memory can run AI21 Jamba Mini 1.5, including NVIDIA DGX H100, NVIDIA DGX A100 640GB, Mac Studio M4 Max (128 GB).
Runs great
— Plenty of headroom
Decent
— Enough memory, may be tightRelated Models
Frequently Asked Questions
- How much VRAM does AI21 Jamba Mini 1.5 need?
AI21 Jamba Mini 1.5 requires 113.5 GB of VRAM at BF16.
VRAM = Weights + KV Cache + Overhead
Weights = 51.6B × 16 bits ÷ 8 = 103.1 GB
KV Cache + Overhead ≈ 10.4 GB (at 2K context + ~0.3 GB framework)
VRAM usage by quantization
BF16113.5 GB- Can NVIDIA GeForce RTX 5090 run AI21 Jamba Mini 1.5?
No — AI21 Jamba Mini 1.5 requires at least 113.5 GB at BF16, which exceeds the NVIDIA GeForce RTX 5090's 32 GB of VRAM.
- Can I run AI21 Jamba Mini 1.5 on a Mac?
AI21 Jamba Mini 1.5 requires at least 113.5 GB at BF16, which exceeds the unified memory of most consumer Macs. You would need a Mac Studio or Mac Pro with a high-memory configuration.
- Can I run AI21 Jamba Mini 1.5 locally?
Yes — AI21 Jamba Mini 1.5 can run locally on consumer hardware. At BF16 quantization it needs 113.5 GB of VRAM. Popular tools include Ollama, LM Studio, and llama.cpp.
- How fast is AI21 Jamba Mini 1.5?
At BF16, AI21 Jamba Mini 1.5 can reach ~26 tok/s on AMD Instinct MI300X. Speed depends mainly on GPU memory bandwidth. Real-world results typically within ±20%.
tok/s = (bandwidth GB/s ÷ model GB) × efficiency
Example: AMD Instinct MI300X → 5300 ÷ 113.5 × 0.55 = ~26 tok/s
Estimated speed at BF16 (113.5 GB)
AMD Instinct MI300X~26 tok/sAMD Instinct MI250X~16 tok/sReal-world results typically within ±20%. Speed depends on batch size, quantization kernel, and software stack.
- What's the download size of AI21 Jamba Mini 1.5?
At BF16, the download is about 103.14 GB.
- Which GPUs can run AI21 Jamba Mini 1.5?
No single consumer GPU has enough VRAM to run AI21 Jamba Mini 1.5 at BF16 (113.5 GB). Multi-GPU or professional hardware is required.
- Which devices can run AI21 Jamba Mini 1.5?
5 devices with unified memory can run AI21 Jamba Mini 1.5 at BF16 (113.5 GB), including Mac Pro M2 Ultra (192 GB), Mac Studio M2 Ultra (192 GB), Mac Studio M4 Max (128 GB), NVIDIA DGX A100 640GB. Apple Silicon Macs use unified memory shared between CPU and GPU, making them well-suited for local LLM inference.