AREX Base — Hardware Requirements & GPU Compatibility
ChatFunctionsReasoning
AREX Base is a 122.6B-parameter open language model from BAAI. It supports a context window of up to 262,144 tokens. At BF16 it needs about 245.50 GB of VRAM — see which GPUs and Macs can run it below.
AREX Base runs on a 256 GB+ Mac or a datacenter GPU
Best local pick → Mac Studio (M3 Ultra, 256GB) · BF16 · 245.6 GB · ~2 tok/s
Macs from 256 GB (BF16 · ~2 tok/s)
Apache 2.0
Not set up for it?
Reference GPU · RTX 4090 24 GB
Estimates, ±20% · how we estimate
Run AREX Base
One recommended way. Every other method, OS and quant is in the step-by-step guide.Each way opens a step-by-step guide for your OS and quant.
Recommended
vllm serve BAAI/AREX-Base --max-model-len 16384 --tensor-parallel-size 8vLLM repo with its own quantisation
Download: Weights on Hugging Face
Ways to runOther ways
each opens step-by-stepRun in cloud
Runs on B300 (288 GB) (42 GB headroom) · BF16
- Generation speed
- ~21 tok/s
- generation speed
- Cost per 1M output tokens
- $90.93
- per 1M output tokens
How Much VRAM Does AREX Base Need?
Select a quantization level; the listed hardware adheres to it.
| Quantization | VRAM | +32K | Runs on | |
|---|---|---|---|---|
| 245.5 GB | 245.7 GB | 16-bit · bfloat — Brain floating point 16 — preferred for training | Runs on Instinct MI325X 256 GB |
Selected BF16
Brain floating point 16245.6GBat 16K context
- Weights
- 245.1 GB
- KV cache
- 0.2 GB
- Overhead
- 0.3 GB
Server-class: ~245.6 GB at 16K context · one Instinct MI325X 256 GB
Or .
Which Hardware Can Run AREX Base?
BF16 · 245.5 GBAREX Base (BF16) requires 245.5 GB of VRAM to load the model weights. For comfortable inference with headroom for KV cache and system overhead, 320+ GB is recommended. Using the full 262K context window can add up to 2.3 GB, bringing total usage to 247.8 GB. No consumer GPU has enough memory. 2 devices with unified memory can run AREX Base, including Mac Studio (M3 Ultra, 512GB), Mac Studio (M3 Ultra, 256GB).
Needs server-class hardware
AREX Base runs on a 256 GB+ Mac or a datacenter GPU · BF16 needs ~245.6 GB
| Hardware | Quant | Memory | Speed | Context | Hardware page | |
|---|---|---|---|---|---|---|
| Mac 256 GB · Mac Studio (M3 Ultra, 256GB) | BF16 | 245.6 GB | ~2 tok/s | 256K ctx | Mac Studio → | BF16 · 245.6 GB · ~2 tok/s · 256K ctx |
| Instinct MI325X 256 GB · one GPU | BF16 | 245.6 GB | ~15 tok/s | 256K ctx | Instinct MI325X → | BF16 · 245.6 GB · ~15 tok/s · 256K ctx |
| 4× H100 SXM 80 GB · 320 GB | BF16 | 245.6 GB | — | 256K ctx | H100 SXM → | BF16 · 245.6 GB · 256K ctx |
| 2× H200 SXM 141 GB · 282 GB | BF16 | 245.6 GB | — | 256K ctx | H200 SXM → | BF16 · 245.6 GB · 256K ctx |
| 2× B200 192 GB · 384 GB | BF16 | 245.6 GB | — | 256K ctx | B200 → | BF16 · 245.6 GB · 256K ctx |
Also: Instinct MI325X · Instinct MI350X · Instinct MI355X · B300 · Mac Studio (M3 Ultra, 256GB) · Mac Studio (M3 Ultra, 512GB) · NVIDIA DGX A100 640GB · NVIDIA DGX H100
By VRAM tier · typical GPUs and Macs at BF16
Consumer GPUs (up to 32 GB) and Macs up to 192 GB: too small for this model (needs ≥ 245.6 GB)8 GB tier →12 GB tier →16 GB tier →24 GB tier →Macs →
Similar models
6| Model | Score | VRAM | tok/s | Why it is similar |
|---|---|---|---|---|
| Qwen3.5 122B A10BAlibaba | 109.4 | 75.5 GB | —needs > 24 GB | same size |
| GPT OSS 120BOpenAI | 97.5 | 70.8 GB | —needs > 24 GB | same size |
| Ornith 1.5 122Bornith-ai | — | 75.5 GB | —needs > 24 GB | same size |
| NVIDIA Nemotron 3.5 Super VL 120B A12B BF16NVIDIA | — | 87.7 GB | —needs > 24 GB | same size |
| Ling 3.0 Flash VLInclusion AI | — | 76 GB | —needs > 24 GB | same size |
| Leanstral 1.5 119B A6BMistral AI | — | 84 GB | —needs > 24 GB | same size |
Specifications
- Publisher
- BAAI
- Parameters
- 122.6B
- Architecture
- Qwen3_5MoeForConditionalGeneration
- Context Length
- 262,144 tokens
- Vocabulary Size
- 248,320
- Release Date
- 2026-07-23
- License
- Apache 2.0
Frequently Asked Questions
How much VRAM does AREX Base need?
AREX Base requires 245.5 GB of VRAM at BF16. Full 262K context adds up to 2.3 GB (247.8 GB total).
VRAM = Weights + KV Cache + Overhead
Weights = 122.6B × 16 bits ÷ 8 = 245.1 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 ≈ 2.7 GB (at full 262K context)
VRAM usage by quantization
Can NVIDIA GeForce RTX 5090 run AREX Base?
No — AREX Base requires at least 245.5 GB at BF16, which exceeds the NVIDIA GeForce RTX 5090's 32 GB of VRAM.
Can I run AREX Base on a Mac?
Yes — Mac Studio (M3 Ultra, 256GB) and 1 other Mac can run AREX Base. Apple Silicon uses unified memory, so the model shares RAM with the system. At BF16 you need at least 245.5 GB of usable unified memory (RAM minus macOS overhead).
Can I run AREX Base locally?
Yes — AREX Base can run locally on consumer hardware. Popular tools include Ollama, LM Studio, and llama.cpp.
How fast is AREX Base?
At BF16, AREX Base can reach ~20 tok/s on AMD Instinct MI350X. Speed depends mainly on GPU memory bandwidth. Real-world results typically within ±20%.
tok/s = (bandwidth GB/s ÷ model GB) × efficiency
Example: NVIDIA B300 → 8000 ÷ 245.5 × 0.65 = ~21 tok/s
Estimated speed at BF16 (245.5 GB)
Real-world results typically within ±20%. Speed depends on batch size, quantization kernel, and software stack.
What's the download size of AREX Base?
At BF16, the download is about 245.13 GB.
Which GPUs can run AREX Base?
No single consumer GPU has enough VRAM to run AREX Base at BF16 (245.5 GB). Multi-GPU or professional hardware is required.
Which devices can run AREX Base?
2 devices with unified memory can run AREX Base at BF16 (245.5 GB), including Mac Studio (M3 Ultra, 256GB), Mac Studio (M3 Ultra, 512GB). Apple Silicon Macs use unified memory shared between CPU and GPU, making them well-suited for local LLM inference.