Intel Arc Pro B70 24GB
The Intel Arc Pro B70 24GB has 24 GB VRAM and 456 GB/s memory bandwidth. It can run 42 of our 71 tracked models natively in VRAM at 8k context.
With 24 GB GDDR6, the Intel Arc Pro B70 24GB is a workstation-tier GPU that can run 42 models natively. It handles 70B-class models at Q4 quantization.
The Intel Arc Pro B70 is Intel's flagship Battlemage workstation GPU with 24GB of GDDR6 and ECC support, announced at CES 2025. It targets CAD, media, and AI inference workloads with ISV certifications. The 24GB framebuffer fits 13B models at Q8 or 30B models at aggressive quantization, with the Vulkan backend providing usable LLM inference speeds on Linux and Windows.
Intel Arc Pro B70 24GB: 2025 Xe2-HPG Battlemage workstation GPU with 24GB ECC GDDR6 — Intel's pro flagship.
13B at Q8 or 30B at Q4 natively. ~8-12 t/s for 7B via Vulkan.
Vulkan via llama.cpp works cross-platform. SYCL backend available with oneAPI. ISV-certified workstation card.
| Vendor | Intel |
| Architecture | Xe2-HPG (Battlemage) |
| VRAM | 24 GB |
| Memory type | GDDR6 |
| Memory bandwidth | 456 GB/s |
| Compute backend | VULKAN |
| Tier | Workstation |
| Released | 2025 |
| Models (native) | 42 / 71 |
| Models (offload) | 11 / 71 |
Software: Vulkan backend works in llama.cpp; SYCL backend available with oneAPI toolkit. Primarily a workstation/professional card.
Popular models for this GPU
Models this GPU runs natively in VRAM (42)
- Mixtral 8x7B Instruct v0.146.7B · MMLU-Pro 29.7Q2_K · ~118.2 t/s
- Qwen 3.5 35B-A3B (MoE)35B · MMLU-Pro 84.2Q3_K_M · ~388.8 t/s
- Qwen 3.6 35B35B · MMLU-Pro 85.2Q3_K_M · ~30.3 t/s
- Yi 1.5 34B Chat34.4B · MMLU-Pro 37.0Q3_K_M · ~30.8 t/s
- Qwen3 32B32.8B · MMLU-Pro 65.5Q4_K_M · ~24.7 t/s
- Qwen 2.5 32B Instruct32.5B · MMLU-Pro 69.0Q3_K_M · ~32.6 t/s
- Qwen 2.5 Coder 32B Instruct32.5B · MMLU-Pro 50.4Q3_K_M · ~32.6 t/s
- DeepSeek R1 Distill Qwen 32B32.5B · MMLU-Pro 65.0Q3_K_M · ~32.6 t/s
- Nemotron 3 Nano 30B32B · MMLU-Pro 78.3Q4_K_M · ~297 t/s
- Gemma 4 31B31B · MMLU-Pro 85.2Q3_K_M · ~34.2 t/s
- Qwen3 30B-A3B (MoE)30B · MMLU-Pro 61.5Q5_K_M · ~259.6 t/s
- Gemma 2 27B Instruct27.2B · MMLU-Pro 38.0Q4_K_M · ~29.8 t/s
- Gemma 3 27B Instruct27B · MMLU-Pro 67.5Q5_K_M · ~26.2 t/s
- Qwen 3.6 27B27B · MMLU-Pro 86.2Q5_K_M · ~26.2 t/s
- Gemma 4 26B (MoE)26B · MMLU-Pro 82.6Q5_K_M · ~205 t/s
- Mistral Small 3.1 24B Instruct24B · MMLU-Pro 66.8Q5_K_M · ~29.5 t/s
- Mistral Small 22B22.2B · MMLU-Pro 49.2Q6_K · ~25 t/s
- GPT-OSS 20B21B · MMLU-Pro 67.9Q6_K · ~152.9 t/s
- Qwen3 14B14.8B · MMLU-Pro 61.0Q8_0 · ~30.8 t/s
- Qwen 2.5 14B Instruct14.7B · MMLU-Pro 63.7Q8_0 · ~31 t/s
- Phi-4 14B Instruct14B · MMLU-Pro 70.4Q8_0 · ~32.6 t/s
- Mistral Nemo 12B Instruct12.2B · MMLU-Pro 35.6Q8_0 · ~37.4 t/s
- Gemma 3 12B Instruct12.2B · MMLU-Pro 60.6Q8_0 · ~37.4 t/s
- Gemma 2 9B Instruct9.2B · MMLU-Pro 32.0Q8_0 · ~49.6 t/s
- Llama 3.1 8B Instruct8B · MMLU-Pro 48.3BF16 · ~28.5 t/s
- DeepSeek R1 Distill Llama 8B8B · MMLU-Pro 41.0BF16 · ~28.5 t/s
- Qwen3 8B8B · MMLU-Pro 56.7BF16 · ~28.5 t/s
- Qwen 2.5 7B Instruct7.6B · MMLU-Pro 56.3BF16 · ~30 t/s
- Mistral 7B Instruct v0.37.25B · MMLU-Pro 30.0BF16 · ~31.4 t/s
- Gemma 3 4B Instruct4B · MMLU-Pro 43.6FP32 · ~28.5 t/s
- Gemma 4 E4B4B · MMLU-Pro 69.4FP32 · ~28.5 t/s
- Phi-3.5 Mini Instruct3.8B · MMLU-Pro 47.4FP32 · ~30 t/s
- Llama 3.2 3B Instruct3.2B · MMLU-Pro 24.0FP32 · ~35.6 t/s
- Qwen 2.5 3B Instruct3.1B · MMLU-Pro 32.4FP32 · ~36.8 t/s
- Gemma 2 2B Instruct2.6B · MMLU-Pro 17.8FP32 · ~43.8 t/s
- Gemma 4 E2B2B · MMLU-Pro 60.0FP32 · ~57 t/s
- SmolLM2 1.7B Instruct1.7B · MMLU-Pro 19.0FP32 · ~67.1 t/s
- Qwen 2.5 1.5B Instruct1.5B · MMLU-Pro 16.8FP32 · ~76 t/s
- Llama 3.2 1B Instruct1.24B · MMLU-Pro 12.5FP32 · ~91.9 t/s
- Gemma 3 1B Instruct1B · MMLU-Pro 14.7FP32 · ~114 t/s
- Qwen 2.5 0.5B Instruct0.5B · MMLU-Pro 10.0FP32 · ~228 t/s
- SmolLM2 360M Instruct0.36B · MMLU-Pro 8.0FP32 · ~316.7 t/s
Models that fit with CPU offload (11)
These use system RAM for layers that don't fit in VRAM — expect much slower inference.
- Qwen 3.5 122B-A10B (MoE)122B · MMLU-Pro 86.7Q2_K · ~34.7 t/s
- Nemotron 3 Super 120B120B · MMLU-Pro 83.7Q2_K · ~28.9 t/s
- GPT-OSS 120B117B · MMLU-Pro 80.7Q2_K · ~69.3 t/s
- Llama 4 Scout 109B109B · MMLU-Pro 74.3Q2_K · ~20.4 t/s
- GLM-4.5 Air 106B106B · MMLU-Pro 81.4Q2_K · ~28.9 t/s
- GLM-4.6V 106B106B · MMLU-Pro 79.9Q2_K · ~28.9 t/s
- Qwen 2.5 72B Instruct72B · MMLU-Pro 71.1Q4_K_M · ~2.8 t/s
- Llama 3.3 70B Instruct70B · MMLU-Pro 68.9Q4_K_M · ~2.9 t/s
- DeepSeek R1 Distill Llama 70B70B · MMLU-Pro 70.0Q4_K_M · ~2.9 t/s
- Llama 3.1 70B Instruct70B · MMLU-Pro 66.4Q4_K_M · ~2.9 t/s
- Command-R 35B35B · MMLU-Pro 33.0Q6_K · ~4 t/s
Too large for this GPU (18)
Models mentioned
Mixtral 8x7B Instruct v0.146.7B (12.9B active)
Mistral AIQ4_K_M rec.
Qwen 3.5 35B-A3B (MoE)35B (3B active)
AlibabaQ4_K_M rec.
Qwen 3.6 35B35B
AlibabaQ4_K_M rec.
Yi 1.5 34B Chat34.4B
01.AIQ4_K_M rec.
Qwen3 32B32.8B
AlibabaQ4_K_M rec.
Qwen 2.5 32B Instruct32.5B
AlibabaQ4_K_M rec.
Qwen 2.5 Coder 32B Instruct32.5B
AlibabaQ4_K_M rec.
DeepSeek R1 Distill Qwen 32B32.5B
DeepSeekQ4_K_M rec.
Nemotron 3 Nano 30B32B (3B active)
NVIDIAQ5_K_M rec.
Gemma 4 31B31B
GoogleQ4_K_M rec.
Qwen3 30B-A3B (MoE)30B (3B active)
AlibabaQ4_K_M rec.
Gemma 2 27B Instruct27.2B
GoogleQ4_K_M rec.
Gemma 3 27B Instruct27B
GoogleQ4_K_M rec.
Qwen 3.6 27B27B
AlibabaQ4_K_M rec.
Gemma 4 26B (MoE)26B (3.8B active)
GoogleQ4_K_M rec.
Mistral Small 3.1 24B Instruct24B
Mistral AIQ4_K_M rec.
Mistral Small 22B22.2B
Mistral AIQ4_K_M rec.
GPT-OSS 20B21B (4B active)
OpenAIQ5_K_M rec.
Qwen3 14B14.8B
AlibabaQ5_K_M rec.
Qwen 2.5 14B Instruct14.7B
AlibabaQ5_K_M rec.
Phi-4 14B Instruct14B
MicrosoftQ5_K_M rec.
Mistral Nemo 12B Instruct12.2B
Mistral AIQ5_K_M rec.
Gemma 3 12B Instruct12.2B
GoogleQ5_K_M rec.
Gemma 2 9B Instruct9.2B
GoogleQ5_K_M rec.
Llama 3.1 8B Instruct8B
MetaQ5_K_M rec.
DeepSeek R1 Distill Llama 8B8B
DeepSeekQ5_K_M rec.
Qwen3 8B8B
AlibabaQ5_K_M rec.
Qwen 2.5 7B Instruct7.6B
AlibabaQ6_K rec.
Mistral 7B Instruct v0.37.25B
Mistral AIQ6_K rec.
Gemma 3 4B Instruct4B
GoogleQ6_K rec.
Gemma 4 E4B4B
GoogleQ5_K_M rec.
Phi-3.5 Mini Instruct3.8B
MicrosoftQ6_K rec.
Llama 3.2 3B Instruct3.2B
MetaQ6_K rec.
Qwen 2.5 3B Instruct3.1B
AlibabaQ6_K rec.
Gemma 2 2B Instruct2.6B
GoogleQ8_0 rec.
Gemma 4 E2B2B
GoogleQ8_0 rec.
SmolLM2 1.7B Instruct1.7B
Hugging FaceQ8_0 rec.
Qwen 2.5 1.5B Instruct1.5B
AlibabaQ8_0 rec.
Llama 3.2 1B Instruct1.24B
MetaQ8_0 rec.
Gemma 3 1B Instruct1B
GoogleQ8_0 rec.
Qwen 2.5 0.5B Instruct0.5B
AlibabaQ8_0 rec.
SmolLM2 360M Instruct0.36B
Hugging FaceQ8_0 rec.
Frequently asked questions
- How much VRAM does the Intel Arc Pro B70 24GB have?
- The Intel Arc Pro B70 24GB has 24 GB of GDDR6 with 456 GB/s memory bandwidth.
- What is the Intel Arc Pro B70 24GB best for?
- With 24 GB of VRAM, the Intel Arc Pro B70 24GB is well-suited for running 7B–32B models at Q4 with room for context, making it a great all-rounder for local LLM inference.
- What LLMs can the Intel Arc Pro B70 24GB run locally?
- The Intel Arc Pro B70 24GB can run 42 of the 71 open-weight models tracked by CanItRun natively in VRAM at 8k context. Top options include: Llama 3.1 8B Instruct at BF16, Llama 3.2 3B Instruct at FP32, Llama 3.2 1B Instruct at FP32.
- Can the Intel Arc Pro B70 24GB run Llama 3.3 70B Instruct?
- The Intel Arc Pro B70 24GB can run Llama 3.3 70B Instruct with CPU offload at Q4_K_M quantization, but inference will be slower than native VRAM execution.
- Can the Intel Arc Pro B70 24GB run Qwen 3.6 27B?
- Yes. The Intel Arc Pro B70 24GB runs Qwen 3.6 27B natively in VRAM at Q5_K_M quantization, achieving approximately 26.2 tokens per second.
- Can the Intel Arc Pro B70 24GB run Llama 3.1 8B Instruct?
- Yes. The Intel Arc Pro B70 24GB runs Llama 3.1 8B Instruct natively in VRAM at BF16 quantization, achieving approximately 28.5 tokens per second.