Intel Arc Pro B70 32GB
Key facts
The Intel Arc Pro B70 32GB has 32 GB graphics memory. 26 of 42 models we track fit fully at standard quality (estimate).
No checked euro price yet. Checked . Intel.
Intel's workstation card with 32 GB, when 24 GB is not enough.
Price
No price yet
We have no checked euro price yet. Price details below say what is known.
Memory
32 GB graphics memory
Largest model that fits
Up to Qwen3.6-35B-A3B
The largest well-known model it fits: a mid-size model, at standard quality.
Self-reported speed on Qwen3.8-27B
Self-reported by Josh Gold (github.com/killercatfish, 24 Oct 2026, quoted by name with written permission; SYCL prefill bug report in llama.cpp #30892 dated 20 Oct 2026). Headless Ubuntu 26.04 (kernel 7.0), Ryzen 5 7600X, 32 GB DDR5, B850 motherboard, two Intel Arc Pro B70 32 GB cards, one model per card. Card A runs vLLM (cyspiegel/vllm-xpu-b70) with Qwen3.8-27B Int4 (AutoRound) at about 30 tok/s decode; a 10K-token prompt takes about 7 s cold and 0.6 s warm with the prompt cache. Card B, in the B850 chipset PCIe 4.0 x4 slot, runs llama.cpp on Vulkan (Mesa 26.2.5) behind llama-swap with Qwen3.8-27B Q4_K_M at about 15 tok/s decode and 375–500 tok/s prompt reading (Vulkan: 0 faults). One card also faulted under Intel's compute stack; a slot swap later showed that card was defective, not the slot. On 26 Oct 2026 the faulty card failed 3 of 3 in the CPU x16 slot, and the healthy card ran 13 of 13 clean in the B850 chipset PCIe 4.0 x4 slot. The defective card is going back to ASRock for replacement. A single-card llama.cpp SYCL test build (b11755, GGML_SYCL_F16=OFF) measured about 270 tok/s prompt reading verified correct to 14K tokens; GGML_SYCL_F16=ON raised prompt speed to about 640 tok/s but produced !!!! NaN garbage (llama.cpp #30892; others point at PR #30604, which he had not retested). Splitting one model across both B70 cards over the B850 PCIe 4.0 x4 chipset slot dropped decode to 5.7 tok/s. Not a localLLMrig measurement.
- 1× Intel Arc Pro B70 32GB, vLLM (cyspiegel/vllm-xpu-b70), Int4 AutoRound decode [self-reported]: 30 tokens per second [self-reported]
- 1× Intel Arc Pro B70 32GB, llama.cpp Vulkan (Mesa 26.2.5), Q4_K_M decode [self-reported]: 15 tokens per second [self-reported]
- 2× Intel Arc Pro B70 32GB split across B850 PCIe 4.0 x4 chipset slot, decode [self-reported]: 5.7 tokens per second [self-reported]
Josh Gold (killercatfish), llama.cpp #30892. The written interview is Josh Gold's dual Arc Pro B70. Not a localLLMrig measurement.
Which models run on it
Each AI model needs a certain amount of memory. This is what fits at standard quality with a normal-length conversation (about 12 pages of text), and how fast the answers would come.
| AI model | Runs? | Speed |
|---|---|---|
| LFM3-1.7Bsmall | Runs | No measured speed yet |
| Llama 3.2 3Bsmall | Runs | No measured speed yet |
| Spark-X2.6-4Bsmall | Runs | No measured speed yet |
| Mistral 7B v0.3small | Runs | No measured speed yet |
| NVIDIA Nemotron 3.5 Bolt E4Bsmall | Runs | No measured speed yet |
| Llama 3.1 8Bsmall | Runs | No measured speed yet |
| Gemma 2 9Bsmall | Runs | No measured speed yet |
| Qwen2.5 14Bmid-size | Runs | No measured speed yet |
| Phi-4 14Bmid-size | Runs | No measured speed yet |
| DeepSeek R1 Distill Qwen 14Bmid-size | Runs | No measured speed yet |
| gpt-oss-20bmid-size | Runs | No measured speed yet |
| Mistral Small 3.2 24Bmid-size | Runs | No measured speed yet |
| Gemma 4 26B A4Bmid-size | Runs | No measured speed yet |
| OrcaCyber Nano 1.0mid-size | Runs | No measured speed yet |
| Gemma 2 27Bmid-size | Runs | No measured speed yet |
| Gemma 3 27Bmid-size | Runs | No measured speed yet |
| Qwen3.8-27Bmid-size | Runs | No measured speed yet |
| Muse Glimmer 30Bmid-size | Runs | No measured speed yet |
| NVIDIA Nemotron 3.5 Lightningmid-size | Runs | No measured speed yet |
| Qwen3 Coder 30B A3Bmid-size | Runs | No measured speed yet |
| Gemma 4 31Bmid-size | Runs | No measured speed yet |
| Qwen3 32Bmid-size | Runs | No measured speed yet |
| Qwen2.5 32Bmid-size | Runs | No measured speed yet |
| Nanbeige4.3-33Bmid-size | Runs | No measured speed yet |
| Qwen3.6-35B-A3Bmid-size | Runs | No measured speed yet |
| Recursive Flintmid-size | Runs | No measured speed yet |
| Llama 3.3 70Blarge | Doesn't fit | – |
| DeepSeek-R1-Distill 70Blarge | Doesn't fit | – |
| Qwen2.5 72Blarge | Doesn't fit | – |
| NVIDIA Nemotron 3.5 Bolt 74B-A9Blarge | Doesn't fit | – |
| Kolibri-1large | Doesn't fit | – |
| Llama 4 Scoutvery large | Doesn't fit | – |
| gpt-oss-120bvery large | Doesn't fit | – |
| Laguna S 2.1very large | Doesn't fit | – |
| NVIDIA Nemotron 3.5 Super VLvery large | Doesn't fit | – |
| Ornith-1.5-122Bvery large | Doesn't fit | – |
| Qwen3.8-Flash-Nextvery large | Doesn't fit | – |
| MiniMax M3.1 Flashvery large | Doesn't fit | – |
| Reflection Beam 1.0very large | Doesn't fit | – |
| DeepSeek V4.1 Flash 1017very large | Doesn't fit | – |
| Meta Muse Spark Openvery large | Doesn't fit | – |
| Kimi K3.1 Provery large | Doesn't fit | – |
No measured speed yet. We do not publish a speed from memory bandwidth for this machine.
Other quality levels
Higher quality needs more memory. Standard quality is what most people run at home.
- Q4 · standard (smallest that keeps quality)
- Q5 · a bit sharper, a bit bigger
- Q8 · near full quality, almost twice Q4
- FP16 · full quality, over three times Q4
Swipe the table sideways to see every quality level.
| AI model | Q4 | Q5 | Q8 | FP16 |
|---|---|---|---|---|
| LFM3-1.7B | Runs | Runs | Runs | Runs |
| Llama 3.2 3B | Runs | Runs | Runs | Runs |
| Spark-X2.6-4B | Runs | Runs | Runs | Runs |
| Mistral 7B v0.3 | Runs | Runs | Runs | Runs |
| NVIDIA Nemotron 3.5 Bolt E4B | Runs | Runs | Runs | Runs |
| Llama 3.1 8B | Runs | Runs | Runs | Runs |
| Gemma 2 9B | Runs | Runs | Runs | Runs |
| Qwen2.5 14B | Runs | Runs | Runs | Doesn't fit |
| Phi-4 14B | Runs | Runs | Runs | Doesn't fit |
| DeepSeek R1 Distill Qwen 14B | Runs | Runs | Runs | Doesn't fit |
| gpt-oss-20b | Runs | Runs | Runs | Runs |
| Mistral Small 3.2 24B | Runs | Runs | Runs | Doesn't fit |
| Gemma 4 26B A4B | Runs | Runs | Runs | Doesn't fit |
| OrcaCyber Nano 1.0 | Runs | Runs | Runs | Doesn't fit |
| Gemma 2 27B | Runs | Runs | Doesn't fit | Doesn't fit |
| Gemma 3 27B | Runs | Runs | Runs | Doesn't fit |
| Qwen3.8-27B | Runs | Runs | Runs | Doesn't fit |
| Muse Glimmer 30B | Runs | Runs | Doesn't fit | Doesn't fit |
| NVIDIA Nemotron 3.5 Lightning | Runs | Runs | Doesn't fit | Doesn't fit |
| Qwen3 Coder 30B A3B | Runs | Runs | Doesn't fit | Doesn't fit |
| Gemma 4 31B | Runs | Runs | Doesn't fit | Doesn't fit |
| Qwen3 32B | Runs | Runs | Doesn't fit | Doesn't fit |
| Qwen2.5 32B | Runs | Runs | Doesn't fit | Doesn't fit |
| Nanbeige4.3-33B | Runs | Runs | Doesn't fit | Doesn't fit |
| Qwen3.6-35B-A3B | Runs | Runs | Doesn't fit | Doesn't fit |
| Recursive Flint | Runs | Runs | Doesn't fit | Doesn't fit |
| Llama 3.3 70B | Doesn't fit | Doesn't fit | Doesn't fit | Doesn't fit |
| DeepSeek-R1-Distill 70B | Doesn't fit | Doesn't fit | Doesn't fit | Doesn't fit |
| Qwen2.5 72B | Doesn't fit | Doesn't fit | Doesn't fit | Doesn't fit |
| NVIDIA Nemotron 3.5 Bolt 74B-A9B | Doesn't fit | Doesn't fit | Doesn't fit | Doesn't fit |
| Kolibri-1 | Doesn't fit | Doesn't fit | Doesn't fit | Doesn't fit |
| Llama 4 Scout | Doesn't fit | Doesn't fit | Doesn't fit | Doesn't fit |
| gpt-oss-120b | Doesn't fit | Doesn't fit | Doesn't fit | Doesn't fit |
| Laguna S 2.1 | Doesn't fit | Doesn't fit | Doesn't fit | Doesn't fit |
| NVIDIA Nemotron 3.5 Super VL | Doesn't fit | Doesn't fit | Doesn't fit | Doesn't fit |
| Ornith-1.5-122B | Doesn't fit | Doesn't fit | Doesn't fit | Doesn't fit |
| Qwen3.8-Flash-Next | Doesn't fit | Doesn't fit | Doesn't fit | Doesn't fit |
| MiniMax M3.1 Flash | Doesn't fit | Doesn't fit | Doesn't fit | Doesn't fit |
| Reflection Beam 1.0 | Doesn't fit | Doesn't fit | Doesn't fit | Doesn't fit |
| DeepSeek V4.1 Flash 1017 | Doesn't fit | Doesn't fit | Doesn't fit | Doesn't fit |
| Meta Muse Spark Open | Doesn't fit | Doesn't fit | Doesn't fit | Doesn't fit |
| Kimi K3.1 Pro | Doesn't fit | Doesn't fit | Doesn't fit | Doesn't fit |
Ran a model on this card? Send us your llama-bench result.
Speeds and memory are our estimate from the hardware specs, not a measured test. Prices are dated and may change. How we calculate
Sources and method
Fit is our estimate from memory size. No measured speed yet, and we do not publish a speed from memory bandwidth for this machine. Checked . Intel.
Price details and parts
Intel's specification lists 32 GB of GDDR6, a 256-bit bus, 608 GB/s and ECC. It does not state a euro price, so we do not invent one. No speed from that bandwidth. A self-reported Qwen3.8-27B run by Josh Gold is on the model page (llama.cpp #30892).
Source: Intel, checked 24 Oct 2026.
What this is
- Intel Arc Pro B70, 32 GB GDDR6, 256-bit, 608 GB/s, ECC, from Intel's specification
Prices exclude shipping. Euro prices include VAT. Dollar prices are converted from our euro prices without EU VAT and before US sales tax, so they are an estimate.
Technical specs
- Graphics memory
- 32 GB graphics memory
- Usable for AI models (our estimate)
- About 31.4 GB. A little is kept free for the graphics driver.
- Memory speed (bandwidth)
- 608 GB/s. How fast the chip can read the model; higher means faster answers. Source
- In short
- 32 GB workstation card. Intel lists 608 GB/s. We do not publish a speed from that figure. A self-reported Qwen3.8-27B run is on the model page.
Common questions
- Can the Intel Arc Pro B70 32GB run Llama 3.3 70B?
- No, by our estimate. Llama 3.3 70B needs about 46 GB at standard quality, and the Intel Arc Pro B70 32GB has about 31.4 GB it can use.
- Which AI models run fully on the Intel Arc Pro B70 32GB?
- By our estimate, 26 of the 42 models we track run fully in its memory at standard quality: LFM3-1.7B, Llama 3.2 3B, Spark-X2.6-4B, Mistral 7B v0.3, NVIDIA Nemotron 3.5 Bolt E4B, Llama 3.1 8B, Gemma 2 9B, Qwen2.5 14B, Phi-4 14B, DeepSeek R1 Distill Qwen 14B, gpt-oss-20b, Mistral Small 3.2 24B, Gemma 4 26B A4B, OrcaCyber Nano 1.0, Gemma 2 27B, Gemma 3 27B, Qwen3.8-27B, Muse Glimmer 30B, NVIDIA Nemotron 3.5 Lightning, Qwen3 Coder 30B A3B, Gemma 4 31B, Qwen3 32B, Qwen2.5 32B, Nanbeige4.3-33B, Qwen3.6-35B-A3B, Recursive Flint.
- How fast is the Intel Arc Pro B70 32GB with AI models?
- Qwen3.6-35B-A3B fits fully in its memory, but we have no speed estimate for it on this machine. No measured speed yet.