readygpu

AREX 2

27.4B parameters · 256k context · Apache License 2.0 · released 2026-09-29

agenticreasoningcodingvisionchat

In short

AREX 2 is an open model from BAAI with 27.4B parameters. At Q4_K_M, AREX 2 needs about 17 GB of memory at 8k context and 26 GB at 128k, counting the weights, the KV cache and runtime buffers. The cheapest hardware in the UK that runs AREX 2 fully in memory at 128k context is the NVIDIA RTX PRO 4500 Blackwell, £4,587 at eBay, at about 19 tok/s (estimated).

Updated . Speeds are estimates; how they are worked out.

Cheapest way to run it in the UK today

NVIDIA RTX PRO 4500 Blackwell, new at eBay

Runs AREX 2 at Q5_K_M fully in memory at 128k context, about 19 tok/sest.

£4,587

In stock, checked

What to run it on in the UK

Fully in memory at 128k context, priced at retailers today.

All hardware that runs it

Everything we price, best value first. Two cards split the model by layers.

HardwareQuantFitsDecodePricetok/s per 1,000
Intel Arc Pro B70Q5_K_Mfits10 tok/s£1,4227.2
AMD Radeon AI PRO R9700Q5_K_Mfits12 tok/s£1,6107.5
GMKtec EVO-X2 Ryzen AI Max+ 395 128GBQ8_0fits4 tok/s£3,3991.1
Mac mini M4 Pro 64GBQ8_0fits5 tok/s£3,3991.4
Mac Studio M4 Max 64GBQ8_0fits10 tok/s£3,3992.8
NVIDIA RTX A6000Q8_0fits13 tok/s£3,9993.2
NVIDIA RTX PRO 4500 BlackwellQ5_K_Mfits19 tok/s£4,5874.1
NVIDIA GeForce RTX 5090Q5_K_Mfits38 tok/s£4,8377.8
NVIDIA DGX Spark 128GBQ8_0fits5 tok/s£4,9990.9
Mac Studio M3 Ultra 96GBQ8_0fits14 tok/s£5,1492.8
ASUS Ascent GX10 128GBQ8_0fits5 tok/s£5,2990.8
Lenovo ThinkStation PGX GB10 128GBQ8_0fits5 tok/s£5,9440.8
NVIDIA RTX PRO 5000 Blackwell 48GBQ8_0fits22 tok/s£7,5003.0
Mac Studio M3 Ultra 256GBQ8_0fits14 tok/s£8,3491.7
NVIDIA RTX 6000 Ada GenerationQ8_0fits16 tok/s£11,2831.4
NVIDIA RTX PRO 6000 BlackwellQ8_0fits30 tok/s£14,4992.0
Mac Studio M3 Ultra 512GBQ8_0fits14 tok/s£17,4950.8
Mac Studio M4 Max 128GBQ8_0fits10 tok/s––
Framework Desktop Ryzen AI Max+ 395 128GBQ8_0fits4 tok/s––
Mac mini M4 Pro 48GBQ6_Kfits6 tok/s––
Framework Desktop Ryzen AI Max+ 395 64GBQ8_0fits4 tok/s––
NVIDIA RTX PRO 5000 Blackwell 72GBQ8_0fits22 tok/s––
AMD Radeon PRO W7900Q8_0fits13 tok/s––
Beelink GTR9 Pro Ryzen AI Max+ 395 128GBQ8_0fits4 tok/s––
NVIDIA GeForce RTX 5050Q4_K_Moffload3 tok/s£3606.9
NVIDIA GeForce RTX 3060 12GBQ4_K_Moffload3 tok/s£3957.6
AMD Radeon RX 7600 XTQ4_K_Moffload3 tok/s£4008.3
NVIDIA GeForce RTX 5060Q4_K_Moffload3 tok/s£4006.5
Intel Arc A770 16GBQ4_K_Moffload4 tok/s£4059.4
NVIDIA GeForce RTX 4060 Ti 8GBQ4_K_Moffload3 tok/s£4076.1
NVIDIA GeForce RTX 5060 Ti 8GBQ4_K_Moffload3 tok/s£4405.9
NVIDIA GeForce RTX 3080 10GBQ4_K_Moffload3 tok/s£4995.8
Intel Arc B580Q4_K_Moffload3 tok/s£5145.8
AMD Radeon RX 9060 XT 16GBQ4_K_Moffload3 tok/s£5406.3
NVIDIA GeForce RTX 4060Q4_K_Moffload3 tok/s£5504.5
NVIDIA Tesla P40Q4_K_Moffload7 tok/s£6969.3
NVIDIA GeForce RTX 5060 Ti 16GBQ4_K_Moffload4 tok/s£7005.6
AMD Radeon RX 9070Q4_K_Moffload4 tok/s£7005.7
NVIDIA RTX PRO 2000 BlackwellQ4_K_Moffload4 tok/s£7025.0
NVIDIA GeForce RTX 5070Q4_K_Moffload3 tok/s£7404.3
AMD Radeon RX 9070 XTQ4_K_Moffload4 tok/s£7655.2
NVIDIA GeForce RTX 4070Q4_K_Moffload3 tok/s£7863.9
AMD Radeon RX 7800 XTQ4_K_Moffload4 tok/s£8005.0
AMD Radeon RX 7900 XTQ4_K_Moffload6 tok/s£8297.0
NVIDIA GeForce RTX 4070 SuperQ4_K_Moffload3 tok/s£9543.2
NVIDIA GeForce RTX 5070 TiQ4_K_Moffload4 tok/s£9754.4
NVIDIA GeForce RTX 4060 Ti 16GBQ4_K_Moffload4 tok/s£9773.6
NVIDIA GeForce RTX 4070 Ti SuperQ4_K_Moffload4 tok/s£1,3293.2
NVIDIA GeForce RTX 3090Q4_K_Moffload12 tok/s£1,3308.9
NVIDIA GeForce RTX 5080Q4_K_Moffload4 tok/s£1,3723.2
AMD Radeon RX 7900 XTXQ4_K_Moffload11 tok/s£1,3997.6
NVIDIA GeForce RTX 4080 SuperQ4_K_Moffload4 tok/s£1,4302.9
NVIDIA GeForce RTX 3090 TiQ4_K_Moffload12 tok/s£2,1995.5
NVIDIA RTX PRO 4000 BlackwellQ4_K_Moffload10 tok/s£2,3534.2
NVIDIA GeForce RTX 4090Q4_K_Moffload12 tok/s£5,0822.4
Intel Arc Pro B60Q4_K_Moffload7 tok/s––

2 other cards and machines we price cannot hold it at this context.

Memory needed

Weights, KV cache and runtime buffers at each context length (KV cache at fp16).

QuantFile@8k@32k@64k@128k
Q4_K_Mgguf16.0 GB17.4 GB19.1 GB21.2 GB25.6 GB
Q5_K_Mgguf19.3 GB20.7 GB22.3 GB24.5 GB28.9 GB
Q6_Kgguf22.0 GB23.4 GB25.0 GB27.2 GB31.6 GB
Q8_0gguf26.7 GB28.1 GB29.7 GB31.9 GB36.3 GB

Run it

Copy a command for the runtime you use.

  • Q4_K_M
    • Ollamaollama run hf.co/bartowski/BAAI_AREX-2-GGUF:Q4_K_M
    • LM Studiolms get bartowski/BAAI_AREX-2-GGUF@q4_k_m
    • llama.cppllama-server -hf bartowski/BAAI_AREX-2-GGUF:Q4_K_M
  • Q5_K_M
    • Ollamaollama run hf.co/bartowski/BAAI_AREX-2-GGUF:Q5_K_M
    • LM Studiolms get bartowski/BAAI_AREX-2-GGUF@q5_k_m
    • llama.cppllama-server -hf bartowski/BAAI_AREX-2-GGUF:Q5_K_M
  • Q6_K
    • Ollamaollama run hf.co/bartowski/BAAI_AREX-2-GGUF:Q6_K
    • LM Studiolms get bartowski/BAAI_AREX-2-GGUF@q6_k
    • llama.cppllama-server -hf bartowski/BAAI_AREX-2-GGUF:Q6_K
  • Q8_0
    • Ollamaollama run hf.co/bartowski/BAAI_AREX-2-GGUF:Q8_0
    • LM Studiolms get bartowski/BAAI_AREX-2-GGUF@q8_0
    • llama.cppllama-server -hf bartowski/BAAI_AREX-2-GGUF:Q8_0

Notes

Dense 27B multimodal model fine-tuned from Qwen3.8-27B; requires a recent Transformers release with Qwen3.8 support. The card says to follow the Qwen base model's terms and notices.

Questions

How much VRAM does AREX 2 need?
At Q4_K_M, AREX 2 needs about 17 GB of memory at 8k context and 26 GB at 128k, counting the weights, the KV cache and runtime buffers.
What is the best GPU for running AREX 2 locally?
The cheapest hardware in the UK that runs AREX 2 fully in memory at 128k context is the NVIDIA RTX PRO 4500 Blackwell, £4,587 at eBay, at about 19 tok/s (estimated). The best value is the NVIDIA GeForce RTX 5090 (about 38 tok/s for £4,837). The fastest we price is the NVIDIA GeForce RTX 5090, at about 38 tok/s.
What is the cheapest way to run AREX 2 in the UK today?
The NVIDIA RTX PRO 4500 Blackwell, £4,587 new at eBay, in stock: it runs AREX 2 at Q5_K_M fully in memory at 128k context, at about 19 tok/s (estimated).
Can I run AREX 2 on a 16 GB GPU?
Only partly: at 128k context it does not fit in 16 GB, so part of it runs from system RAM and it slows to about 3 tok/s on the AMD Radeon RX 7600 XT (estimated).
Can I run AREX 2 on a 24 GB GPU?
Only partly: at 128k context it does not fit in 24 GB, so part of it runs from system RAM and it slows to about 7 tok/s on the NVIDIA Tesla P40 (estimated).
How do I run AREX 2 locally?
With Ollama: ollama run hf.co/bartowski/BAAI_AREX-2-GGUF:Q4_K_M (the Q4_K_M build). LM Studio and llama.cpp commands are on this page.
What licence is AREX 2 released under?
Apache License 2.0, which allows commercial use.

More hardware for local AI