tencent
Tencent: Hy-MT2-30B-A3B
tencent/hy-mt2-30b-a3b
Our take
AI-assisted analysis based on the sources below. Suitability is an editorial assessment; test it on your own workload.
tencent/hy-mt2-30b-a3b is the largest, mixture-of-experts variant of Tencent's Hy-MT2 family, described by the vendor as a flagship translation model for complex real-world scenarios and as part of a "fast-thinking" design. Vendor documentation says it translates among 33 languages plus several Chinese dialect and minority-language pairs, and supports instruction templates for terminology, style, personalization, delimiter preservation, structured data and context-aware translation. The OpenRouter catalog lists it as text-to-text with an 8192-token context. It is released as open weights under Apache-2.0, with documented transformers, vLLM and SGLang deployment paths.[1][2]
Strengths
- Vendor documentation describes support for 33 languages plus several Chinese dialect and minority-language pairs, with a published table covering varieties such as Traditional Chinese, Cantonese, Tibetan, Kazakh, Mongolian and Uyghur.[1][2]
- Documented instruction workflows cover glossary and terminology guidance, style control, personalization, exact delimiter retention, context-aware prompts, and structured-data translation that preserves keys, placeholders and hierarchy.[1][2]
- Open weights under Apache-2.0, with a 30B-A3B FP8 variant, recommended inference parameters, and a training pipeline supporting both full-parameter and LoRA fine-tuning.[2]
Limitations
- Evidence here is catalog and vendor documentation only: the README's results are a vendor-published image and OpenRouter's benchmark fields were empty, so quality and outperform claims are vendor-reported, not independently measured in this pack.[1][2]
- It is text-only and, per the catalog, limited to an 8192-token context, so it is a specialist translation model that may not suit long-document or multimodal workloads.[1][2]
- The README's vLLM and SGLang instructions require building those frameworks from source, and hardware requirements for the 30B-A3B variant are not stated; validate throughput, cost and translation quality on your own pairs before production.[2]
Where to use it
- Multilingual translation pipelines, especially involving Chinese and less-common language pairs, where a dedicated translation model is preferred over a general-purpose LLM.[1][2]
- Localization of structured content such as JSON, code-adjacent text and delimited strings, where preserving keys, placeholders and delimiters matters.[1][2]
- Enterprise translation needing terminology control, style constraints or background context, using the vendor's documented prompt patterns.[1][2]
Sources
- [1] OpenRouter catalog · Tencent: Hy-MT2-30B-A3BSources checked: Nov 13, 2026
- [2] Model documentation · Tencent: Hy-MT2-30B-A3BSources checked: Nov 13, 2026
About this model
Description below is the original text supplied by OpenRouter, shown in the source language.
Hy-MT2-30B-A3B is Tencent's flagship translation model in the Hy-MT2 family. It supports 33 language pairs and five Chinese dialect and minority-language pairs, with workflows for structured, delimiter-based, contextual, glossary-based, and...
Specifications
- Model ID
- tencent/hy-mt2-30b-a3b
- Canonical ID
- tencent/hy-mt2-30b-a3b-20260521
- Creator
- tencent
- Added to OpenRouter
- 20.08.2026
- Context length
- 8,192
- Modality
- text->text
- Input modalities
- text
- Output modalities
- text
- Token pricing
- USD per 1M tokens: in $0.074 / out $0.295
- Knowledge cutoff
- —
- Expiration date
- —
Tasks and capabilities
A grade appears only where the source supplies a matching measurement for that task, the model is active and the cohort holds at least 10 comparable models. The capability column lists facts the source declares — supported formats, reasoning and tool parameters — which are not measured quality.
| Task | Grade | Source-declared capability |
|---|---|---|
| General quality | No measurement | — |
| Coding | No measurement | — |
| Agents | No measurement | parameters: structured_outputs |
| Writing | No measurement | output: text |
| Reasoning | No measurement | — |
| Document analysis | No measurement | — |
| Vision | No measurement | — |
| Image generation | No measurement | — |
| Audio | No measurement | — |
| Video | No measurement | — |
| Embeddings | No measurement | — |
Evaluations
No evaluations are supplied for this model.
How grades work
Each letter grade compares a model only against other models scored by the same source and metric in the current snapshot. S means the top 10%, A the next 20%, B the next 30%, C the rest. Models without a published measurement are shown as not rated — a supported format alone never implies measured quality.
- S — S — top 10% of the cohort
- A — A — next 20% of the cohort
- B — B — next 30% of the cohort
- C — C — below the B band
Providers
Optional live provider statistics from OpenRouter. Missing values are unknown, not zero.
| Provider | Context length | Price in /1M | Price out /1M | Latency p50 (30m, s) | Throughput p50 (30m, tok/s) | Uptime (30m) |
|---|---|---|---|---|---|---|
| Tencentfp8 | 8,192 | $0.074 | $0.295 | — | — | 100% |
Provider data retrieved: 13.11.2026
Related articles
No published articles identify this model yet.