# Arc Pro B60 > OnPrem World · https://onpremworld.com/en/hardware/intel-arc-pro-b60 · Last updated: 2026-10-14 · CC BY 4.0 24 GB of VRAM at 456 GB/s; Intel set no MSRP, and its Project Battlematrix platform scales to eight cards (192 GB) for multi-GPU inference on Linux. ## Specifications | | | |---|---| | Memory | 24 GB GDDR6 | | Memory bandwidth | 456 GB/s | | Power | 200 W | | Launch | 2025-06 | | Architecture | Xe2 (Battlemage) | | Software | oneAPI (SYCL) / Vulkan | ## What it can run (24 GB, 8K) Estimates, not benchmarks. See the methodology. | Model | Parameters | Best quant | Total | tok/s (est.) | |---|---|---|---|---| | Qwen3.5-35B-A3B | 35B | Q4_K_M | 21.4 GiB | 176 | | Qwen3.6-35B-A3B | 35B | Q4_K_M | 21.4 GiB | 176 | | Qwen3-32B | 32.8B | Q4_K_M | 21.9 GiB | 16 | | DeepSeek-R1-Distill-Qwen-32B | 32.76B | Q4_K_M | 21.9 GiB | 16 | | Gemma 4 31B IT | 30.7B | Q5_K_M | 22.9 GiB | 15 | | Qwen3-30B-A3B (2507) | 30.5B | Q5_K_M | 22.5 GiB | 136 | | Qwen3-Coder-30B-A3B-Instruct | 30.5B | Q5_K_M | 22.5 GiB | 136 | | GLM-4.7-Flash (30B-A3B) | 30B | Q5_K_M | 21.8 GiB | 150 | | Qwen3.5-27B | 27B | Q6_K | 22.7 GiB | 14 | | Qwen3.6-27B | 27B | Q6_K | 22.7 GiB | 14 | | Qwen3.8-27B | 27B | Q6_K | 22.7 GiB | 14 | | Gemma 3 27B IT | 27B | Q6_K | 23.3 GiB | 14 | | Gemma 4 26B A4B IT | 25.2B | Q6_K | 21.1 GiB | 102 | | Mistral Small 3.2 24B Instruct (2506) | 24B | Q6_K | 21.1 GiB | 16 | | Magistral Small 1.2 (2509) | 24B | Q6_K | 21.1 GiB | 16 | | Devstral Small 2 24B Instruct (2512) | 24B | Q6_K | 21.1 GiB | 16 | | gpt-oss-20b | 21B | Q8_0 | 22.5 GiB | 83 | | Phi-4 (14B) | 14B | Q8_0 | 16.6 GiB | 21 | | Gemma 3 12B IT | 12B | Q8_0 | 13.8 GiB | 25 | | Gemma 4 12B Unified IT | 11.95B | Q8_0 | 13.3 GiB | 25 | | Qwen3.5-9B | 9B | FP16 | 18.4 GiB | 18 | | Qwen3-8B | 8.2B | FP16 | 17.7 GiB | 19 | | DeepSeek-R1-0528-Qwen3-8B | 8.19B | FP16 | 17.6 GiB | 19 | | Llama 3.1 8B Instruct | 8B | FP16 | 17.1 GiB | 20 | | Gemma 4 E4B IT | 8B | FP16 | 16.4 GiB | 20 | | Llama 3.2 3B Instruct | 3.21B | FP16 | 7.7 GiB | 50 | | Kimi K2.6 | 1000B | Doesn't fit | – | – | | DeepSeek-R1-0528 (671B) | 671B | Doesn't fit | – | – | | DeepSeek-V4-Flash | 284B | Doesn't fit | – | – | | Qwen3.8-Flash-Next | 180B | Doesn't fit | – | – | | Qwen3.5-122B-A10B | 122B | Doesn't fit | – | – | | Mistral Small 4 119B (2603) | 119B | Doesn't fit | – | – | | gpt-oss-120b | 117B | Doesn't fit | – | – | | Llama 4 Scout 17B-16E Instruct | 109B | Doesn't fit | – | – | | GLM-4.5-Air | 106B | Doesn't fit | – | – | | Qwen3-Coder-Next (80B-A3B) | 80B | Doesn't fit | – | – | | Llama 3.3 70B Instruct | 70B | Doesn't fit | – | – | ## Prices Latest price at each shop, as listed on the day we checked. | Region | Shop | Listing | Price | Per GB | Checked | |---|---|---|---|---|---| | US | Newegg (SmartJoy Shop) | ASRock Intel Arc Pro B60 Creator 24GB Graphics Card, Workstation GPU, 24GB GDDR6 192-bit, PCIe 5.0, Blower (Newegg marketplace seller) | [US$917.99](https://www.newegg.com/p/3C6-037X-003N8?Item=9SIC8E6M6A0811) | US$38/GB | 2026-10-14 | US value: about 65.8 tok/s on Llama 3.1 8B Instruct (Q4_K_M, 8K), or 7.2 tok/s per US$100. All prices: https://onpremworld.com/en/prices.md ## Frequently asked questions ### Which LLMs can the Arc Pro B60 run? With 24 GB and an 8K context, the largest models that fit include: Qwen3.5-35B-A3B, Qwen3.6-35B-A3B, Qwen3-32B, DeepSeek-R1-Distill-Qwen-32B, Gemma 4 31B IT. See the full table below. ### How fast is the Arc Pro B60 for local LLMs? Generation speed is mostly set by memory bandwidth (456 GB/s). For Llama 3.1 8B Instruct at FP16, expect roughly 20 tokens/s; bigger models are slower. ### Can the Arc Pro B60 run 70B models? Llama 3.3 70B Instruct does not fit on the Arc Pro B60 (24 GB), even at Q4_K_M with an 8K context. ## Sources - https://www.intel.com/content/www/us/en/products/sku/243916/intel-arc-pro-b60-graphics/specifications.html - https://www.silicon.co.uk/press-release/computex-2025-intel-unveils-new-gpus-for-ai-and-workstations - https://download.intel.com/newsroom/2025/client-computing/Intel-Arc-Pro-B60-Data-Sheet.pdf