- Services
Every open model in our directory, ranked by how well it runs on ASRock Radeon AI PRO R9700 Creator (32GB) at 4-bit. Grades weigh whether the model fits in memory and how fast it should run.
ASRock Radeon AI PRO R9700 Creator has 32 GB of VRAM, enough to run 261 of the 358 open models we track at 4-bit. The largest that still fits well is Falcon 40B Instruct (40B), and the top-rated is LFM2.5-24B-A2B at about 115 tok/s.
261 of 358 models run comfortably on ASRock Radeon AI PRO R9700 Creator.
See full ASRock Radeon AI PRO R9700 Creator specs and pricingThe top models this device can run at 4-bit, ranked by fit and speed.
| Model | Grade | Speed | VRAM |
|---|---|---|---|
| LFM2.5-24B-A2BLiquid AI | SS | 114.7 tok/s | 17.0 GB |
| DiffusionGemma 26B-A4BGoogle | SS | 49.1 tok/s | 22.9 GB |
| Gemma 4 26B-A4B ITGoogle | SS | 46.8 tok/s | 23.8 GB |
| Nemotron 3.5 LightningNvidia | SS | 220.3 tok/s | 18.0 GB |
| North Mini CodeCohere | SS | 61.5 tok/s | 24.0 GB |
| Nemotron 3 Nano OmniNVIDIA | SS | 60.4 tok/s | 24.2 GB |
| Qwen3-30B-A3BAlibaba | SS | 95.7 tok/s | 21.0 GB |
| Holo4-35B-A3BHcompany | SS | 220.3 tok/s | 20.9 GB |
| Llama 3.1 8B InstructMeta | SS | 38.6 tok/s | 13.3 GB |
| Llama 2 13B ChatMeta | SS | 60.9 tok/s | 8.5 GB |
| Mellum2.1JetBrains | SS | 253.6 tok/s | 7.5 GB |
| Tess-4.5-27BContinuity Research | SS | 30.2 tok/s | 17.0 GB |
LFM2.5-24B-A2BLiquid AI | 24B(2.45B active) | SS | 114.7 tok/s | 17.0 GB | |
DiffusionGemma 26B-A4BGoogle | 25.2B(3.8B active) | SS | 49.1 tok/s | 22.9 GB | |
Gemma 4 26B-A4B ITGoogle | 26B(4B active) | SS | 46.8 tok/s | 23.8 GB | |
Nemotron 3.5 LightningNvidia | 30B(3B active) | SS | 220.3 tok/s | 18.0 GB | |
North Mini CodeCohere | 30B(3B active) | SS | 61.5 tok/s | 24.0 GB | |
Nemotron 3 Nano OmniNVIDIA | 30B(3B active) | SS | 60.4 tok/s | 24.2 GB | |
Qwen3-30B-A3BAlibaba | 30B(3B active) | SS | 95.7 tok/s | 21.0 GB | |
Holo4-35B-A3BHcompany | 35B(3B active) | SS | 220.3 tok/s | 20.9 GB | |
Ad Tools for AI BuildersAll the hardware and tools you need to run private AI at your office or at home.Shop Now | |||||
| 8B | SS | 38.6 tok/s | 13.3 GB | ||
Llama 2 13B ChatMeta | 13B | SS | 60.9 tok/s | 8.5 GB | |
Mellum2.1JetBrains | 12B(2.5B active) | SS | 253.6 tok/s | 7.5 GB | |
Tess-4.5-27BContinuity Research | 27B | SS | 30.2 tok/s | 17.0 GB | |
Holo4-27BHcompany | 27B | SS | 30.2 tok/s | 17.0 GB | |
| 8.78B | SS | 35.3 tok/s | 14.6 GB | ||
Raon-1 mini previewKRAFTON | 28.78B(4.3B active) | SS | 42.9 tok/s | 26.2 GB | |
Xing4.0-29B-A4BChina Telecom AI | 29B(4B active) | SS | 46.0 tok/s | 25.7 GB | |
Ad Vast.aiAffordable on-demand GPU rentals for training and inference. Pick from thousands of hosts.Rent a GPU | |||||
Qwen3.6 35B-A3BAlibaba | 35B(3B active) | SS | 60.4 tok/s | 27.1 GB | |
Qwen3.5-35B-A3BAlibaba | 35B(3B active) | SS | 60.4 tok/s | 27.1 GB | |
MiniCPM-V 4.7 35B-A3Bopenbmb | 35.21B(3B active) | SS | 60.4 tok/s | 27.2 GB | |
LFM2.5-8B-A1BLiquid AI | 8.3B(1.5B active) | SS | 177.3 tok/s | 6.8 GB | |
| 4B | SS | 46.0 tok/s | 11.2 GB | ||
| 3.98B | SS | 46.2 tok/s | 11.2 GB | ||
LensVLM-9BApple | 9B | SS | 85.7 tok/s | 6.0 GB | |
| 9B | SS | 85.7 tok/s | 6.0 GB | ||
| AdTools for AI BuildersAll the hardware and tools you need to run private AI at your office or at home.Shop Now | |||||
| 8B | AA | 91.0 tok/s | 5.7 GB | ||
Model Compatibility
Use the calculator to weigh this device against any other, model by model, with speed and fit scores.
Straight answers to the questions we hear most often.
Can't find what you're looking for? Book a Discovery Call
LFM2.5-24B-A2B is the highest-rated model that runs on ASRock Radeon AI PRO R9700 Creator at 4-bit, at about 115 tok/s. In total, ASRock Radeon AI PRO R9700 Creator can run 261 of the 358 open models we track.
We size each model at 4-bit (Q4_K_M), add the KV cache for its context length and a fixed runtime overhead, then compare that to this device VRAM. Each model is graded on how comfortably it fits and how fast it should run on this card.
It is an estimate of decode speed in tokens per second at 4-bit, based on this device memory bandwidth and each model size. Real speed depends on your runtime, batch size, and context length, so use it to compare models, not as a hard promise.
Lower-bit quantization shrinks a model so more of them fit. These grades already assume 4-bit, which is the common local default. Going below 4-bit can squeeze a larger model in at some quality cost; going above needs more VRAM and may not fit.
Mixture-of-experts models only activate a fraction of their parameters at a time, so their runtime memory is closer to the active size than the headline size. We score fit on the active parameters, which is why some very large models still fit.
