Hy-MT2 — local AI model by Tencent
Written by Jakub Rusinowski · Last updated
Tencent's open translation models: 1.8B, 7B and a 30B mixture-of-experts with about 3B active, each translating among 33 languages and following translation instructions such as terminology and style. They translate; they are not general chat assistants, so they are kept out of the chat rankings. Apache-2.0, with official GGUF builds.
Variants
The smallest Hy-MT2 variant needs about 2 GB of VRAM at Q4_K_M — quantized weights plus framework overhead, before any KV cache.
| Model | VRAM at Q4 | VRAM | Context | Run it |
|---|---|---|---|---|
| Hy-MT2 1.8B → 1.8B | ~1.9 GB | 262,144 | ollama run hy-mt | |
| Hy-MT2 7B → 7B | ~5.4 GB | 262,144 | ollama run hy-mt | |
| Hy-MT2 30B-A3B → 30B (3B active) | ~19 GB | 262,144 | ollama run hy-mt |
Memory is quantized weights plus overhead at Q4_K_M, from the same engine as the GPU & VRAM checker.
How to run Hy-MT2 locally
Install Ollama, then pull the tag.
ollama run hy-mtPick a size above for its own VRAM figure, speed estimate and install command.
Licence
Commercial use permitted. No usage restrictions beyond attribution.
Applies to: Hy-MT2 1.8B, Hy-MT2 7B, Hy-MT2 30B-A3BRecommended GPU
The cheapest catalogued GPU that runs Hy-MT2 locally (min 2 GB VRAM) is the Intel Arc B570 (10 GB).
Hy-MT2 — frequently asked questions
How much VRAM does Hy-MT2 need?
Hy-MT2 needs about 2 GB VRAM at Q4_K_M quantization for its smallest variant. Variants: Hy-MT2 1.8B (2 GB, Q4_K_M); Hy-MT2 7B (5 GB, Q4_K_M); Hy-MT2 30B-A3B (19 GB, Q4_K_M). On Apple Silicon, unified memory counts toward this requirement.
Can I run Hy-MT2 on an RTX 4090 (24 GB)?
Yes — Hy-MT2 runs on an RTX 4090 (24 GB) and other 24 GB cards such as the RTX 3090. Smaller variants also fit comfortably on 8–16 GB GPUs at Q4_K_M.
What quantization should I use for Hy-MT2?
Q4_K_M is the best balance of quality and VRAM for Hy-MT2 in most cases. Choose Q8_0 for near-lossless quality if you have spare VRAM, or smaller quants (Q3/Q2) only when memory is tight.
How do I run Hy-MT2 with Ollama?
Hy-MT2 has no local Ollama tag — the published tag is cloud-hosted, so running it sends your prompts to a hosted GPU rather than your own machine.