# Intel Arc Pro B70 for local LLMs

> 32GB of new VRAM for well under NVIDIA or R9700 prices, held back by Intel's maturing software.

The Arc Pro B70 has 32GB of GDDR6 ECC and 608 GB/s of memory bandwidth, which decides how big a model fits and how fast it generates text. It comfortably runs models up to about 35B parameters: the largest in our catalog is Qwen3.6 35B-A3B at Q5_K_M (28GB with 8K context, ~62 tokens/sec).

## Key specs

- Memory: 32GB GDDR6 ECC
- Memory bandwidth: 608 GB/s
- Architecture: Xe2 (Battlemage) (2026)
- Power: 230W
- Typical price: ~$1,300 (checked 2026-10)
- Software: SYCL, Vulkan

## Best models for this GPU

- **Best overall (also coding, reasoning, vision, fastest, largest that fits):** Qwen3.6 35B-A3B at Q5_K_M, ~27.7GB with 8K context, ~62 tokens/sec
- **Runner-up:** Qwen3.8 27B at Q6_K, ~24.8GB with 8K context, ~14 tokens/sec
- **Stretch pick (CPU offload):** Qwen3.5 122B-A10B at Q4_K_M, ~77.0GB with 8K context, about 10 tokens/sec with CPU offload

## Compatibility (best quantization that fits, 8K context)

| Model | Params | Quant | Memory | Speed | Verdict |
| --- | --- | --- | --- | --- | --- |
| Gemma 4 E4B | 8B | Q8_0 | 10.5GB | ~33 tok/s | Runs well |
| Qwen3.5 9B | 9B | Q8_0 | 12.0GB | ~29 tok/s | Runs well |
| Phi-4 | 14.7B | Q8_0 | 18.3GB | ~21 tok/s | Runs well |
| gpt-oss-20b | 20.9B (3.6B active) | MXFP4 | 15.0GB | ~59 tok/s | Runs well |
| Gemma 4 26B A4B | 26B (3.8B active) | Q6_K | 23.2GB | ~48 tok/s | Runs well |
| Qwen3.8 27B | 27B | Q6_K | 24.8GB | ~14 tok/s | Runs well |
| Gemma 4 31B | 31B | Q5_K_M | 26.0GB | ~14 tok/s | Runs well |
| Qwen3.6 35B-A3B | 35B (3B active) | Q5_K_M | 27.7GB | ~62 tok/s | Runs well |
| GLM-4.5-Air | 106B (12B active) | Q4_K_M | 68.3GB | ~8.8 tok/s | Partial offload |
| Llama 4 Scout | 109B (17B active) | Q4_K_M | 69.5GB | ~6.3 tok/s | Partial offload |
| gpt-oss-120b | 117B (5.1B active) | MXFP4 | 68.3GB | ~21 tok/s | Partial offload |
| Qwen3.5 122B-A10B | 122B (10B active) | Q4_K_M | 77.0GB | ~10 tok/s | Partial offload |
| DeepSeek V4 Flash | 284B (13B active) | - | - | - | Won't fit |
| GLM-5.3-Flash | 320B (18B active) | - | - | - | Won't fit |
| GLM-5.3 | 744B (40B active) | - | - | - | Won't fit |

## Pros
- 32GB with ECC for less than rival 32GB cards
- 608 GB/s bandwidth
- 230W board power and a PCIe 5.0 x16 link

## Cons
- Intel's AI software stack is still maturing
- Price rose well above the $949 MSRP

Full page: https://llamabuilds.ai/gpus/arc-pro-b70 · Best models: https://llamabuilds.ai/gpus/arc-pro-b70/best-models

Sources: [StorageReview Arc Pro B70 review](https://www.storagereview.com/review/intel-arc-pro-b70-review), [Puget Systems Arc Pro B70 review](https://www.pugetsystems.com/labs/articles/intel-arc-pro-b70-review/), [GPU Poet Arc Pro B70 specs and pricing](https://gpupoet.com/gpu/learn/card/intel-arc-pro-b70)

---
Source: https://llamabuilds.ai/gpus/arc-pro-b70 · Updated October 14, 2026 · Llama Builds. Speeds marked ~ are estimates from our bandwidth model (https://llamabuilds.ai/methodology); measured results are labeled as such.