// Tencent Hunyuan / Hy · Translation
Hy-MT2-7B: VRAM & local deployment
8.03B parameters. About 6.9 GB at 4-bit and 19.2 GB at BF16 with one 8K-token request.
released 2026-05-11licence: Commercial OKmemory figures are estimates
01Specs
| Publisher | Tencent · tencent |
|---|---|
| Family / series | Tencent Hunyuan / Hy · Hy-MT2 |
| Type | Translation (translation) |
| Parameters | 8.03B Source: counted from official safetensors weights |
| Context length | 256K tokens (262,144)max_position_embeddings in config.json |
| Architecture | 32 layers · GQA |
| Released weights | BF16 · 16.1 GBdownload |
| Licence | Apache-2.0 · Commercial OKPermissive licence (Apache-2.0 / MIT / CC-BY / OpenMDW): commercial use allowed. |
| Released | 2026-05-11 (Hugging Face repo created) |
02How much VRAM it needs (estimate)
Total = weights + 10% + 0.5GB overhead + KV cache for one request (FP16 KV). More concurrent users multiply the KV column.
| Precision | Weights | Total @ 8K | Total @ 32K | Total @ 128K | Smallest GPU | Mac |
|---|---|---|---|---|---|---|
| BF16 / FP16 (original) | 16.1 GB | 19.2 GB | 22.5 GB | 35.3 GB | RTX 4090 24GB | Mac 36GB |
| FP8 / INT8 / Q8_0 | 8.5 GB | 11.0 GB | 14.2 GB | 27.1 GB | 16GB GPU | Mac 24GB |
| 4-bit (Q4_K_M / AWQ / GPTQ-Int4) | 4.9 GB | 6.9 GB | 10.1 GB | 23.0 GB | 8GB GPU | Mac 16GB |
KV cache per request: 8K 1.1 GB · 32K 4.3 GB · 128K 17.2 GB
Bits per weight: BF16 16, FP8 8.5, 4-bit 4.85. “Fits” = total ≤ 90% of memory; Mac GPU can use ~2/3 of unified memory (≤36GB) or ~3/4 (larger). For MoE, all experts must be in memory, so size by total parameters. Method →
03How to run it
Only runtimes the publisher or the runtime’s official library names for this model.
- vLLM — mentioned in the official model card
- SGLang — mentioned in the official model card
- llama.cpp — mentioned in the official model card
- Transformers — mentioned in the official model card
04Official repositories & quantised builds
3 official repo(s) from tencent.
- Hy-MT2-7B-GGUF GGUF
- Hy-MT2-7B main
- Hy-MT2-7B-FP8 FP8
05Other Tencent Hunyuan / Hy models
All 19 →| Model | Params | 4-bit total | Released |
|---|---|---|---|
| Hy-MT2-1.8B · same series | 2.04B | 2.4 GB | 2026-05-11 |
| Hy-MT2-30B-A3B · same series | 30.1B / 3B | 21.4 GB | 2026-05-11 |
| Hy-Memory-30B-A3B | 30.1B / 3B | 21.4 GB | 2026-10-15 |
| Hy4-preview | 770B / 49B | 515 GB | 2026-08-27 |
| Hy3 | 295B / 21B | 200 GB | 2026-07-02 |
| Hy3-preview | 295B / 21B | 200 GB | 2026-04-13 |
| HY-MT1.5-1.8B | 2.04B | 2.4 GB | 2025-12-25 |
| HunyuanOCR | 1.12B | 2.1 GB | 2025-11-18 |
| HunyuanVideo-1.5 | — | — | 2025-11-18 |
| HunyuanImage-3.0-Instruct | 80B / 13B | 53.9 GB | 2025-09-25 |
06Sources
Data checked 2026-05-26. Parameters, context and licence are copied from the official repository; memory figures are our estimates.