Arc Pro B65 Price & Used Prices: Which Local LLMs Can It Run?
Price alert
We'll email you when the eBay (US) price reaches your target.
The Arc Pro B65 is Intel's 2026 32GB, 200W workstation card for running models up to 35B locally. Listings mix it up with the B70.
eBay
Arc Pro B65 price history
eBay
As an eBay Partner, VRAMGlass may be compensated if you buy through the eBay links on this page. The price you pay stays the same.
Full Arc Pro B65 specifications
- Memory32 GB · GDDR6
- Bandwidth608 GB/s
- FP16 / BF1698.5 TFLOPS
- INT8197 TOPS
- Bus width
- 256-bit
- Architecture
- Xe2 Battlemage (BMG-G31)
- Cores
- 2,560
- Tensor cores
- 160
- Ecosystem
- oneAPI
- Interface
- PCIe 5.0 x16
- Power
- 200 W
- Slots
- 2
- Type
- Workstation
- Launch date
- Mar 25, 2026
What LLMs can a single Arc Pro B65 run?
One B65 is happiest with 30B-class MoE: about 113 tokens/s at 4-bit with full 256K context. 27B dense crawls at 15 to 16 at 8-bit; use 4-bit.
| Model | Size | Runs? | Speed t/s | Max context |
|---|---|---|---|---|
 Gemma 4 E4B8B | 3.8 GB | Runs | 102.9 | 128K |
 Qwen3.5 9B9.7B | 4.1 GB | Runs | 95.4 | 256K |
| Gemma 4 12B12B | 4.7 GB | Runs | 83.5 | 256K |
| Gemma 4 26B A4B25.2B (3.8B active) | 10.5 GB | Runs | 117.7 | 256K |
| Qwen3.6 27B27.8B | 11.8 GB | Runs | 36.7 | 256K |
| Qwen3.8 27B27.8B | 9.8 GB | Runs | 43.5 | 256K |
 GLM 4.7 Flash30B (3B active) | 11.9 GB | Runs | 129.6 | 198K |
| Gemma 4 31B30.7B | 11.8 GB | Runs | 35.3 | 246K |
| Qwen3.6 35B A3B36B (3B active) | 12.3 GB | Runs | 141.3 | 256K |
| Model | Size | Runs? | Speed t/s | Max context |
|---|---|---|---|---|
| Gemma 4 E4B8B | 5 GB | Runs | 81.4 | 128K |
| Qwen3.5 9B9.7B | 5.7 GB | Runs | 71.9 | 256K |
| Gemma 4 12B12B | 7.1 GB | Runs | 58.4 | 256K |
| Gemma 4 26B A4B25.2B (3.8B active) | 16.9 GB | Runs | 94.7 | 256K |
| Qwen3.6 27B27.8B | 16.8 GB | Runs | 26.3 | 245K |
| Qwen3.8 27B27.8B | 16.5 GB | Runs | 26.8 | 250K |
| GLM 4.7 Flash30B (3B active) | 18.3 GB | Runs | 110 | 198K |
| Gemma 4 31B30.7B | 18.3 GB | Runs | 23.7 | 168K |
| Qwen3.6 35B A3B36B (3B active) | 22.1 GB | Runs | 113.2 | 256K |
| Model | Size | Runs? | Speed t/s | Max context |
|---|---|---|---|---|
| Gemma 4 E4B8B | 8.2 GB | Runs | 52.2 | 128K |
| Qwen3.5 9B9.7B | 9.5 GB | Runs | 45.3 | 256K |
| Gemma 4 12B12B | 12.7 GB | Runs | 34.3 | 256K |
| Gemma 4 26B A4B25.2B (3.8B active) | 26.9 GB | Runs | 72.5 | 256K |
| Qwen3.6 27B27.8B | 28.6 GB | Runs | 15.8 | 69K |
| Qwen3.8 27B27.8B | 29 GB | Runs | 15.6 | 63K |
| GLM 4.7 Flash30B (3B active) | 31.8 GB | Runs | 83.4 | 26K |
| Gemma 4 31B30.7B | 32.6 GB | Too big | — | — |
| Qwen3.6 35B A3B36B (3B active) | 36.9 GB | Needs RAM offload | 33 | 256K |
RunsFits fully in VRAM; speed is estimated from VRAM bandwidth.
Needs RAM offloadMoE only: expert weights sit in system RAM, so speed is estimated from 70 GB/s RAM bandwidth.
Too bigWon’t fit on one card at this quantization, even with RAM offload.
Full calculation rules are on the Methodology page.
What LLMs can multiple Arc Pro B65 cards run?
Four B65s (128GB) are the practical build, running Qwen3.8 Flash Next at 4-bit at about 73 tokens/s; DeepSeek V4 Flash needs eight. No NVLink.
| Model | Size | Cards | Total price | Total VRAM | Speed t/s | Total power |
|---|---|---|---|---|---|---|
| Qwen3.8 Flash Next176B (6B active) | 78.9 GB | 4 cards | No price | 128 GB | 88.4 | 800 W |
 DeepSeek V4 Flash284B (13B active) | 96.8 GB | 4 cards | No price | 128 GB | 69.5 | 800 W |
 Hy3295B (21B active) | 107.2 GB | 4 cards | No price | 128 GB | 30.9 | 800 W |
| GLM 5.3 Flash320B (18B active) | 108.7 GB | 4 cards | No price | 128 GB | 54.9 | 800 W |
| Model | Size | Cards | Total price | Total VRAM | Speed t/s | Total power |
|---|---|---|---|---|---|---|
| Qwen3.8 Flash Next176B (6B active) | 111.3 GB | 4 cards | No price | 128 GB | 73.1 | 800 W |
| DeepSeek V4 Flash284B (13B active) | 155.1 GB | 8 cards | No price | 256 GB | 49.7 | 1,600 W |
| Hy3295B (21B active) | 182.2 GB | 8 cards | No price | 256 GB | 22.8 | 1,600 W |
| GLM 5.3 Flash320B (18B active) | 199.7 GB | 8 cards | No price | 256 GB | 34.3 | 1,600 W |
| Model | Size | Cards | Total price | Total VRAM | Speed t/s | Total power |
|---|---|---|---|---|---|---|
| DeepSeek V4 Flash284B (13B active) | 161.9 GB | 8 cards | No price | 256 GB | 48.1 | 1,600 W |
| Qwen3.8 Flash Next176B (6B active) | 188.2 GB | 8 cards | No price | 256 GB | 51.8 | 1,600 W |
Full calculation rules are on the Methodology page.
Who it suits
Good fit
- One person using it locally, with generation speed as the top priority
It's as fast as the B70, at about 27 tokens/s on Qwen3.8 27B at 4-bit
- 24GB isn't enough for you
Qwen3.8 27B at 8-bit (29 GB) fits, and at 4-bit you can open up to about 250K of context
- Your power supply has little headroom
The card draws 200W, 30W less than the B70
- You want to pool 128GB for big MoE models
Four cards run Qwen3.8 Flash Next at 4-bit
Not a good fit
- You often feed it long documents, do RAG, or share it among several people
It has only a little over half the B70's compute, so prompt processing is slow
- You want an Intel-made card or ready-made documentation
The B65 comes only from board partners, and docs such as llm-scaler's don't list it separately
Before you buy
Don't confuse the B65 with the B70
Both cards have 32GB, and listing titles easily mix them up. But the B65 has only 20 Xe cores enabled while the B70 has 32, so the B70 has nearly twice the compute. Before buying, check the full model name rather than just “32GB”.
Board-partner versions only
The B70 comes in an Intel-made version, but the B65 is made only by board partners; at launch, Intel said supply would not start until mid-April 2026. Cooling, power connectors and dimensions vary by brand. If you plan to pack several cards closely, choose a dual-slot blower version, and read the spec page for the exact model before buying.
Software docs often don't list the B65
Intel's llm-scaler (vLLM) documentation names the B60 and B70, and the OpenVINO 2026.1 release notes mention only the B70. The B65 uses the same chip as the B70, so in principle it runs the same software; but when you hit a problem, it may not be among the officially validated configurations.
Compared with similar GPUs
Arc Pro B65 vs Arc Pro B70, Radeon AI PRO R9700, and Arc Pro B60: price, VRAM, bandwidth, compute and local LLM speed compared.
| Spec | Arc Pro B65 | Arc Pro B70 | Radeon AI PRO R9700 | Arc Pro B60 |
|---|---|---|---|---|
| Memory | 32 GB | 32 GBSame | 32 GBSame | 24 GB25% less |
| Bandwidth | 608 GB/s | 608 GB/sSame | 640 GB/s5% higher | 456 GB/s25% lower |
| BF16 / FP16 compute | 99 TFLOPS | 184 TFLOPS86% higher | 191 TFLOPS94% higher | 99 TFLOPSSame |
| INT8 compute | 197 TOPS | 367 TOPS86% higher | 383 TOPS94% higher | 197 TOPSSame |
| Qwen3.8 27B speed (2-bit) | 43.5 t/s | 43.5 t/sSame | 45.6 t/s5% faster | 33.1 t/s24% slower |
| Largest dense model (4-bit) | 51B | 51BSame | 51BSame | 37B27% less |
| Power | 200 W | 230 W15% higher | 300 W50% higher | 200 WSame |
| Interface | PCIe 5.0 x16 | PCIe 5.0 x16 | PCIe 5.0 x16 | PCIe 5.0 x8 |
| Released | 2026-03 | 2026-03 | 2025-07 | 2025-08 |
| eBay median |
As an eBay Partner, VRAMGlass may be compensated if you buy through the eBay links on this page. The price you pay stays the same.
Amazon.jp
Xianyu