Radeon AI PRO R9700 Price & Used Prices: Which Local LLMs Can It Run?
Price alert
We'll email you when the eBay (US) price reaches your target.
AMD's Radeon AI PRO R9700 (2025) is a 32GB blower card that stacks well, running 27B-class with long context at about 28 tokens/s. The S model is fanless.
eBay
Radeon AI PRO R9700 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 Radeon AI PRO R9700 specifications
- Memory32 GB · GDDR6
- Bandwidth640 GB/s
- FP16 / BF16191 TFLOPS
- FP8 / INT8383 / 383 TOPS
- Bus width
- 256-bit
- Architecture
- RDNA 4 (Navi 48)
- Cores
- 4,096
- Tensor cores
- 128
- Ecosystem
- ROCm
- Interface
- PCIe 5.0 x16
- Power
- 300 W
- Power connector
- 1× 12V-2x6
- Slots
- 2
- Type
- Workstation
- Launch date
- Jul 23, 2025
- MSRP
- $1,299
Best Radeon AI PRO R9700 deals on eBay
Hand-picked deals from trusted eBay sellers. Shop with confidence.
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.
UsedASRock AMD Radeon AI Pro R9700 32GB$1,950Buy on eBay
NewASRock Creator Radeon AI PRO R9700 32GB GDDR6 PCI Express 5.0 x16 Graphics Card$2,090Buy on eBay
NewXFX AMD Radeon AI PRO R9700 32GB GDDR6 Professional Graphics CardTop Rated$2,100Buy on eBay
UsedGigabyte Radeon AI PRO R9700 AI TOP 32G GDDR6 Graphics Card$2,150Buy on eBay
NewAsus AMD Radeon AI PRO R9700 Graphic Card - 32 GB GDDR6 (turbo-ai-pro-r9700-32g)$2,314Buy on eBay
What LLMs can a single Radeon AI PRO R9700 run?
One R9700 lives at 4-bit: 30B-class MoE tops 100 tokens/s, 27B-class dense reaches the 20s. 8-bit is a stretch; a 31B no longer fits.
| Model | Size | Runs? | Speed t/s | Max context |
|---|---|---|---|---|
 Gemma 4 E4B8B | 3.8 GB | Runs | 107.5 | 128K |
 Qwen3.5 9B9.7B | 4.1 GB | Runs | 99.7 | 256K |
| Gemma 4 12B12B | 4.7 GB | Runs | 87.3 | 256K |
| Gemma 4 26B A4B25.2B (3.8B active) | 10.5 GB | Runs | 120.4 | 256K |
| Qwen3.6 27B27.8B | 11.8 GB | Runs | 38.5 | 256K |
| Qwen3.8 27B27.8B | 9.8 GB | Runs | 45.6 | 256K |
 GLM 4.7 Flash30B (3B active) | 11.9 GB | Runs | 132.1 | 198K |
| Gemma 4 31B30.7B | 11.8 GB | Runs | 37.1 | 246K |
| Qwen3.6 35B A3B36B (3B active) | 12.3 GB | Runs | 143.7 | 256K |
| Model | Size | Runs? | Speed t/s | Max context |
|---|---|---|---|---|
| Gemma 4 E4B8B | 5 GB | Runs | 85.1 | 128K |
| Qwen3.5 9B9.7B | 5.7 GB | Runs | 75.3 | 256K |
| Gemma 4 12B12B | 7.1 GB | Runs | 61.2 | 256K |
| Gemma 4 26B A4B25.2B (3.8B active) | 16.9 GB | Runs | 97.4 | 256K |
| Qwen3.6 27B27.8B | 16.8 GB | Runs | 27.6 | 245K |
| Qwen3.8 27B27.8B | 16.5 GB | Runs | 28.1 | 250K |
| GLM 4.7 Flash30B (3B active) | 18.3 GB | Runs | 112.7 | 198K |
| Gemma 4 31B30.7B | 18.3 GB | Runs | 24.9 | 168K |
| Qwen3.6 35B A3B36B (3B active) | 22.1 GB | Runs | 115.9 | 256K |
| Model | Size | Runs? | Speed t/s | Max context |
|---|---|---|---|---|
| Gemma 4 E4B8B | 8.2 GB | Runs | 54.7 | 128K |
| Qwen3.5 9B9.7B | 9.5 GB | Runs | 47.5 | 256K |
| Gemma 4 12B12B | 12.7 GB | Runs | 36 | 256K |
| Gemma 4 26B A4B25.2B (3.8B active) | 26.9 GB | Runs | 74.9 | 256K |
| Qwen3.6 27B27.8B | 28.6 GB | Runs | 16.6 | 69K |
| Qwen3.8 27B27.8B | 29 GB | Runs | 16.4 | 63K |
| GLM 4.7 Flash30B (3B active) | 31.8 GB | Runs | 86 | 26K |
| Gemma 4 31B30.7B | 32.6 GB | Too big | — | — |
| Qwen3.6 35B A3B36B (3B active) | 36.9 GB | Needs RAM offload | 33.2 | 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 Radeon AI PRO R9700 cards run?
AMD officially supports up to four R9700s (128GB), just enough for Qwen3.8 Flash Next at 4-bit at about 76 tokens/s; two suit gpt-oss 120B at 2-bit.
| Model | Size | Cards | Total price | Total VRAM | Speed t/s | Total power |
|---|---|---|---|---|---|---|
| Qwen3.8 Flash Next176B (6B active) | 78.9 GB | 4 cards | $8,360 | 128 GB | 91 | 1,200 W |
 DeepSeek V4 Flash284B (13B active) | 96.8 GB | 4 cards | $8,360 | 128 GB | 71.9 | 1,200 W |
 Hy3295B (21B active) | 107.2 GB | 4 cards | $8,360 | 128 GB | 32.3 | 1,200 W |
| GLM 5.3 Flash320B (18B active) | 108.7 GB | 4 cards | $8,360 | 128 GB | 57 | 1,200 W |
| Model | Size | Cards | Total price | Total VRAM | Speed t/s | Total power |
|---|---|---|---|---|---|---|
| Qwen3.8 Flash Next176B (6B active) | 111.3 GB | 4 cards | $8,360 | 128 GB | 75.5 | 1,200 W |
| DeepSeek V4 Flash284B (13B active) | 155.1 GB | 8 cards | $16,720 | 256 GB | 51.7 | 2,400 W |
| Hy3295B (21B active) | 182.2 GB | 8 cards | $16,720 | 256 GB | 23.9 | 2,400 W |
| GLM 5.3 Flash320B (18B active) | 199.7 GB | 8 cards | $16,720 | 256 GB | 35.8 | 2,400 W |
| Model | Size | Cards | Total price | Total VRAM | Speed t/s | Total power |
|---|---|---|---|---|---|---|
| DeepSeek V4 Flash284B (13B active) | 161.9 GB | 8 cards | $16,720 | 256 GB | 50 | 2,400 W |
| Qwen3.8 Flash Next176B (6B active) | 188.2 GB | 8 cards | $16,720 | 256 GB | 53.8 | 2,400 W |
Full calculation rules are on the Methodology page.
Who it suits
Good fit
- You want 32GB for long context
Qwen3.8 27B at 4-bit can open up to 250K, and Gemma 4 31B up to 168K
- Your workstation needs several cards side by side
Dual-slot blower cooling, and AMD's multi-GPU documentation supports up to four cards
- You want FP8
RDNA 4 supports FP8 compute
Not a good fit
- You're after single-card generation speed
On the same Qwen3.8 27B, the RX 7900 XTX does about 41 tokens/s versus about 28 on this card
- You want a 27B at 8-bit
Qwen3.8 27B at 8-bit (29 GB) fits, but runs at only about 16 tokens/s with about 63K of context
Before you buy
Power is 12V-2x6
The card draws 300W through a single 12V-2x6 connector. For one card, AMD recommends a 750W power supply, though the specific model you buy takes precedence. If you plan on four cards, you need four 12V-2x6 cables, so the power supply must have enough native connectors.
Don't end up with the passive R9700S
The R9700S, with an S at the end, is a passively cooled version for server racks. The two are listed side by side in ROCm's support table, and listings easily mix them up. For a desktop workstation, buy the R9700 with the blower fan.
Software status and known issues
It shares a chip target with the RX 9070 XT, and its software support is among the better ones for AMD cards: ROCm 10.0 officially supports it on both Linux and Windows, vLLM's documentation lists the RX 9000 series, and it appears in AMD's Windows PyTorch support table. AMD itself has flagged two known issues: running ONNX Runtime through MIGraphX may fail, and Unsloth QLoRA fine-tuning with PyTorch on RHEL may hit memory access errors. Neither affects llama.cpp inference.
Compared with similar GPUs
Radeon AI PRO R9700 vs Radeon RX 9070 XT, Radeon PRO W7800, and RTX PRO 4500: price, VRAM, bandwidth, compute and local LLM speed compared.
| Spec | Radeon AI PRO R9700 | Radeon RX 9070 XT | Radeon PRO W7800 |  RTX PRO 4500 |
|---|---|---|---|---|
| Memory | 32 GB | 16 GB50% less | 32 GBSame | 32 GBSame |
| Bandwidth | 640 GB/s | 640 GB/sSame | 576 GB/s10% lower | 896 GB/s40% higher |
| BF16 / FP16 compute | 191 TFLOPS | 195 TFLOPSSame | 90 TFLOPS53% lower | 211 TFLOPS10% higher |
| INT8 compute | 383 TOPS | 389 TOPSSame | 90 TOPS76% lower | 422 TOPS10% higher |
| Qwen3.8 27B speed (2-bit) | 45.6 t/s | 45.6 t/sSame | 41.3 t/s9% slower | 62.3 t/s37% faster |
| Largest dense model (4-bit) | 51B | 23B55% less | 51BSame | 51BSame |
| Power | 300 W | 304 WSame | 260 W13% lower | 200 W33% lower |
| Interface | PCIe 5.0 x16 | PCIe 5.0 x16 | PCIe 4.0 x16 | PCIe 5.0 x16 |
| Released | 2025-07 | 2025-03 | 2023-04 | 2025-03 |
| 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