All tasks / Models
Hy-MT2-30B-A3B
Tencent's open-weights machine-translation models; Hy-MT2 ships in 1.8B, 7B and 30B-A3B sizes under Apache 2.0.
- Tencent
- Open weights
- Released May 2026
Where it ranks
- –TranslationModels · Local · no result on this task’s benchmarks yet
Facts
- Vendor
- Tencent
- Family
- Hunyuan-MT (Hy-MT)
- Released
- May 2026
- Weights
- Open: download and run it yourself
- License
- Apache-2.0
- Parameters
- 30.1B (3B active)
- GPU memory to run locally
- ~19 GB (30.1B total (HF safetensors), ~3B active per the A3B name; 4-bit ≈ 30.1 B × 0.57 GB + 1.5 GB ≈ 19 GB, estimate. 7B and 1.8B siblings and official GGUF builds exist.)
- API price
- $0.07 in / $0.29 out per 1M tokens (Tencent first-party endpoint on OpenRouter; the 7B model has the same price, the 1.8B $0.044/$0.177.)
- Access
- Free (open weights)
- Last verified
- 18 Oct 2026
Price from openrouter.ai, 18 Oct 2026.
Benchmark results
No public benchmark results recorded yet.
Where you can use it
- –llama.cppAPIs & offline engines · Local
- –OpenRouterAPIs & offline engines · Cloud
Earlier versions
- HY-MT1.5 · —
- Hunyuan-MT-7B · —