Hy Memory 30B A3B
Hy Memory 30B A3B is an open-weights language model from Tencent, released October 15, 2026. It has 30B parameters (mixture-of-experts, 3B active). The context window is 256K tokens.
Check it against your own PC →
Can you run Hy Memory 30B A3B locally?
The smallest common GPU it runs on is 6 GB of VRAM (with 32 GB of RAM): Q6_K, GPU + experts in RAM, roughly 14 tokens per second. Estimates assume 32 GB of DDR5 RAM, a 16K context and llama.cpp-style GGUF quantization. Enter your exact hardware in the live app.
| GPU memory | Verdict | Best quant | VRAM | RAM | tok/s |
|---|---|---|---|---|---|
| 6 GB VRAM | Runs · GPU + CPU experts | Q6_K (≈ lossless) | 5.2 GB | 21.7 GB | ~14 |
| 8 GB VRAM | Runs · GPU + CPU experts | Q6_K (≈ lossless) | 7.2 GB | 19.7 GB | ~16 |
| 12 GB VRAM | Runs · GPU + CPU experts | Q6_K (≈ lossless) | 11.2 GB | 15.7 GB | ~19 |
| 16 GB VRAM | Runs · GPU + CPU experts | Q6_K (≈ lossless) | 15.2 GB | 11.7 GB | ~24 |
| 24 GB VRAM | Runs on GPU | Q4_K_M (very good) | 20.6 GB | – | ~230 |
| 32 GB VRAM | Runs on GPU | Q6_K (≈ lossless) | 26.9 GB | – | ~300 |
Ready-made quantizations on Hugging Face: rapatsky/Hy-Memory-30B-A3B-GGUF · tencent/Hy-Memory-30B-A3B-GGUF · vinnipukh/Hy-Memory-30B-A3B-GGUF · abenzerps/Hy-Memory-30B-A3B-Quantized · rapatsky/Hy-Memory-30B-A3B-IQ4_XS-GGUF · hermitdave/Hy-Memory-30B-A3B-MLX-Mixed4-6
Benchmark results
Epoch AI has not published results for Hy Memory 30B A3B yet. New models are usually evaluated within days to weeks.
Links
Hugging Face: tencent/Hy-Memory-30B-A3B
More from Tencent
WeVisOne VL 4B · WeVisOne VL 2B · Youtu Parsing Omni · Tencent Hy4 preview · Hy4 preview · Tencent Hy3