LLaDA2.1 Mini — Hardware Requirements & GPU Compatibility
ChatFunctions
LLaDA2.1 Mini is a 16.3B-parameter open language model from Inclusion AI. It supports a context window of up to 32,768 tokens. At BF16 it needs about 32.90 GB of VRAM — see which GPUs and Macs can run it below.
LLaDA2.1 Mini runs on a 48 GB+ Mac or a workstation GPU
Best local pick → Mac Studio M5 Max (48 GB) · BF16 · 33.5 GB · ~93 tok/s
Macs from 48 GB (BF16 · ~93 tok/s)
Apache 2.0
Not set up for it?
Reference GPU · RTX 4090 24 GB
Estimates, ±20% · how we estimate
Run LLaDA2.1 Mini
One recommended way. Every other method, OS and quant is in the step-by-step guide.
Recommended
vllm serve inclusionAI/LLaDA2.1-mini --max-model-len 16384vLLM repo with its own quantisation
Download: Weights on Hugging Face
Other ways
each opens step-by-stepRun in cloud
Fits on A40 (48 GB) (14 GB headroom) · BF16
- Generation speed
- ~133 tok/s
- generation speed
- Cost per 1M output tokens
- $0.73
- per 1M output tokens
How Much VRAM Does LLaDA2.1 Mini Need?
Select a quantization level; the listed hardware adheres to it.
| Quantization | VRAM | +32K | On RTX 4090 (reference) | |
|---|---|---|---|---|
| 32.9 GB | 34.2 GB | 16-bit · bfloat — Brain floating point 16 — preferred for training | Does not fit: needs 33.5 GB |
Selected BF16
Brain floating point 1633.5GBat 16K context
- Weights
- 32.5 GB
- KV cache
- 0.7 GB
- Overhead
- 0.3 GB
Doesn't fit RTX 4090 (reference) · needs 9.5 GB more
Which Hardware Can Run LLaDA2.1 Mini?
BF16 · 32.9 GBLLaDA2.1 Mini (BF16) requires 32.9 GB of VRAM to load the model weights. For comfortable inference with headroom for KV cache and system overhead, 43+ GB is recommended. Using the full 33K context window can add up to 1.3 GB, bringing total usage to 34.1 GB. No consumer GPU has enough memory. 67 devices with unified memory can run LLaDA2.1 Mini, including Mac Studio (M3 Ultra, 256GB), Mac Studio (M3 Ultra, 512GB), Mac Studio M5 Max (48 GB).
48 GB
48 GB tier →fits · 14.5 GB free · up to 32K ctx
2 Mac classes · 1 device family
M5 Max48 GB~93Simulate output on Mac Studio M5 Max (48 GB)M5 Pro48 GB~57Simulate output on Mac Mini M5 Pro (48 GB)Snapdragon X2 Elite Extreme (X2E-96-100)~46Simulate output on Snapdragon X2 Elite Extreme Copilot+ PC
By VRAM tier · typical GPUs and Macs at BF16
Similar models
4| Model | Score | VRAM | tok/s | Why it is similar |
|---|---|---|---|---|
| Kimi VL A3B InstructMoonshot AI | — | 10.7 GB | ~208 | same size |
| Moonlight 16B A3BMoonshot AI | — | 10.1 GB | ~229 | same size |
| DeepSeek v2 LiteDeepSeek | — | 10.2 GB | ~243 | same size |
| Deepseek Moe 16B ChatDeepSeek | — | 11.1 GB | ~231 | same size |
Specifications
- Publisher
- Inclusion AI
- Parameters
- 16.3B
- Active parameters
- 1.4B per token
- Architecture
- LLaDA2MoeModelLM
- Context Length
- 32,768 tokens
- Vocabulary Size
- 157,184
- Release Date
- 2026-02-09
- License
- Apache 2.0
Frequently Asked Questions
How much VRAM does LLaDA2.1 Mini need?
LLaDA2.1 Mini requires 32.9 GB of VRAM at BF16. Full 33K context adds up to 1.3 GB (34.1 GB total).
VRAM = Weights + KV Cache + Overhead
Weights = 16.3B × 16 bits ÷ 8 = 32.5 GB
KV Cache + Overhead ≈ 0.4 GB (at 2K context + ~0.3 GB framework)
Fit ratings and hardware model lists check this model with room for a 16K-token context, which needs a little more memory.
KV Cache + Overhead ≈ 1.6 GB (at full 33K context)
VRAM usage by quantization
Can NVIDIA GeForce RTX 5090 run LLaDA2.1 Mini?
No — LLaDA2.1 Mini requires at least 32.9 GB at BF16, which exceeds the NVIDIA GeForce RTX 5090's 32 GB of VRAM.
Can I run LLaDA2.1 Mini on a Mac?
Yes — Mac Studio M5 Max (48 GB) and 55 other Macs can run LLaDA2.1 Mini. Apple Silicon uses unified memory, so the model shares RAM with the system. At BF16 you need at least 32.9 GB of usable unified memory (RAM minus macOS overhead).
Can I run LLaDA2.1 Mini locally?
Yes — LLaDA2.1 Mini can run locally on consumer hardware. Popular tools include Ollama, LM Studio, and llama.cpp.
How fast is LLaDA2.1 Mini?
At BF16, LLaDA2.1 Mini can reach ~217 tok/s on AMD Instinct MI350X. Speed depends mainly on GPU memory bandwidth. Real-world results typically within ±20%.
tok/s = 1000 ÷ (active GB ÷ (bandwidth GB/s × efficiency) × 1000 + layers × routing ms)
Mixture-of-Experts: just the active experts get read for each token, along with a fixed routing cost per layer.
Example: NVIDIA B200 → 2.9 GB active ÷ (8000 × 0.65) = 0.558 ms, plus 20 layers × 0.055 ms = 1.100 ms, so 1000 ÷ 1.658 ms = ~603 tok/s
Estimated speed at BF16 (32.9 GB)
Real-world results typically within ±20%. Speed depends on batch size, quantization kernel, and software stack.
What's the download size of LLaDA2.1 Mini?
At BF16, the download is about 32.51 GB.
Which GPUs can run LLaDA2.1 Mini?
No single consumer GPU has enough VRAM to run LLaDA2.1 Mini at BF16 (32.9 GB). Multi-GPU or professional hardware is required.
Which devices can run LLaDA2.1 Mini?
67 devices with unified memory can run LLaDA2.1 Mini at BF16 (32.9 GB), including ASUS Ascent GX10, Beelink GTR9 Pro (Ryzen AI Max+ 395, 128 GB), Framework Desktop (Ryzen AI Max+ 395, 128 GB), GMKtec EVO-X2 (Ryzen AI Max+ 395, 128 GB). Apple Silicon Macs use unified memory shared between CPU and GPU, making them well-suited for local LLM inference.