Can I run Hy4 preview locally?
Hy4 preview by tencent has 780B parameters (40B active per token, 256 experts, 8 used per token). At 4-bit with an 8K context it needs about 445 GB of GPU memory. It fits on 1 of 65 GPUs and Macs we track.
Memory needed by quant and context length
GB of GPU memory = weights + KV cache (FP16) + runtime overhead. Weights use parameter count × bits per weight. How we calculate.
| Quant | Quality | 4K ctx | 8K ctx | 32K ctx | 128K ctx |
|---|---|---|---|---|---|
| 2-bit (Q2_K) | rough | 273 GB | 274 GB | 276 GB | 286 GB |
| 3-bit (Q3_K_M) | usable | 356 GB | 356 GB | 359 GB | 369 GB |
| 4-bit (Q4_K_M) | good, most popular | 445 GB | 445 GB | 448 GB | 458 GB |
| 5-bit (Q5_K_M) | very good | 519 GB | 519 GB | 522 GB | 532 GB |
| 6-bit (Q6_K) | near lossless | 597 GB | 597 GB | 600 GB | 610 GB |
| 8-bit (Q8_0) | effectively lossless | 773 GB | 773 GB | 776 GB | 786 GB |
| 16-bit (F16/BF16) | original weights | 1454 GB | 1454 GB | 1457 GB | 1467 GB |
Which GPUs and Macs run Hy4 preview?
Best quality quant that fits at 8K context. "Tight" = within 10% of usable memory. Speed is a ceiling from memory bandwidth (real speed is typically 50–80% of it).
Prefer an API?
Tencent: Hy4 preview is on OpenRouter at $0.834 per 1M input tokens and $2.501 per 1M output tokens. Price details and history →
Questions
How much VRAM does Hy4 preview need?
About 445 GB at 4-bit (Q4_K_M) with an 8K context: 444 GB of weights (estimated), 0.8 GB of KV cache and 0.6 GB of runtime overhead. At 8-bit it needs about 773 GB.
What is the smallest GPU that runs Hy4 preview?
No single consumer GPU in our table fits it at Q4_K_M with 8K context.
Can I run Hy4 preview on a Mac?
Not at Q4_K_M with 8K context on any Mac in our table, even a 512 GB Mac Studio, using the default GPU memory limit.
Model facts
| Hugging Face repo | tencent/Hy4-preview |
| Parameters | 779,960,992,733 (780B) · ~40B active (computed from config) |
| Layers / hidden size | 78 / 6144 |
| Attention | MLA; 64 heads, 8 KV heads, head dim 64 |
| KV cache per token (FP16) | 107.3 KiB |
| Max context | 1,048,576 tokens |
| License | apache-2.0 |
| Released on HF | Aug 27, 2026 |
| Downloads (30d) / likes | 11,357 / 564 |
Sources: Hugging Face model API · config.json (tencent/Hy4-preview) · GPU specs from vendor pages (list). Data fetched Nov 17, 2026. Spot an error? Report it.