Choose theme
AI Resources
Hy-MT1.5-1.8B-1.25bit
AngelSlim’s Hy-MT1.5-1.8B-1.25bit compresses a Tencent translation model for small-device use, with a separate GGUF download and Android demo.
It targets local translation across 33 languages. The GGUF instructions use a custom STQ1_0 kernel, so check the specific runtime path before treating this as a model you can load in any local app. Use this as a first read, not a recommendation. Open the original project before trusting details like terms, limits, privacy, cost, setup, or safety.
What it is
A compressed translation model
This is a 1.25-bit variant of HY-MT1.5-1.8B for translating text, not a general-purpose assistant.
Why it stands out
Small weights, specific runtime
The authors report a 440MB weight footprint. That is a storage figure; it does not describe the complete working memory or confirm compatibility with your device.
Availability
Model, GGUF and demo
The model card links a GGUF variant, an Android demo and the required runtime branch. These are different ways to work with the project, not interchangeable installations.
Why it matters
What makes it useful
Compare the two published weight footprints before choosing a variant: 440MB for 1.25-bit and 574MB for the linked 2-bit model. Match those figures to the space you have reserved for model weights, then follow the instructions for the variant you select. This is a storage choice, not a comparison of translation quality.
What to know
Where it fits
The authored GGUF guide points to llama.cpp pull request 22836 and its STQ1_0 implementation. A GGUF filename alone does not establish that a stock runtime can read this quantization. Follow that guide rather than relying only on a model platform’s generic launch suggestions.
Notable points
What stands out
Choose the artifact for your task. The model repository provides the conversion path; the separate GGUF repository provides the file for its documented runtime. If you want to try translation rather than convert weights, start with the GGUF instructions, not the conversion sequence.
Before using
What to review
The Android download is a demo APK. Check its origin and requested access before installing; a model’s local inference design does not establish the whole app’s network or data behavior.
The advertised background word-extraction mode can work with text from other apps. Use ordinary sample text first and check what access the demo requests.
Confirm the result for your language pair and important terms. Published examples do not establish the translation quality of your own text.
When a word needs a particular translation, make that choice explicit in the prompt. The linked base-model guide includes a terminology template that maps a source term to a preferred target term before the passage. Use that template for your chosen wording, then check the returned translation. This supplies terminology guidance; it does not guarantee that the compressed model will follow it correctly.
Reader fit
Who may find it relevant
Developers exploring a translation component on a constrained device.
Readers prepared to use the documented runtime or inspect a demo, rather than expecting a universal local-app download.
Someone seeking a general chatbot needs a different model purpose.
Editorial note
Why LifeHubber lists it
The 1.25-bit label describes a particular packing method, not simply a smaller download of an ordinary GGUF model. The authored GGUF card explains Sherry’s sparse ternary pattern: three nonzero weights and one zero are packed into five bits for four weights. That gives readers exploring small-device translation a concrete compression design to study alongside the weight footprint. Its custom STQ1_0 kernel is part of that choice; storage reduction alone does not establish equivalent translation quality.
Source links
Source materials
Reader note
Before relying on this entry
LifeHubber lists entries to help readers inspect AI projects, not to endorse them or prove they are safe, suitable, accurate, maintained, or right for a specific use. We do not verify every entry in depth. Before relying on anything listed, review the original materials, terms, privacy practices, limits, and risks that matter for your situation.
More in AI Models
Keep browsing this category
Explore more AI model resources.
Gemma 4
google/gemma-4
A Google DeepMind Gemma 4 model family collection with public checkpoints including Gemma 4 12B, a dense multimodal model Google describes around local agentic workflows, native audio input, and encoder-free vision/audio handling.
MiniMax H3 Integrations
MiniMax-AI/awesome-minimax-h3-integration
A community-maintained MiniMax H3 integration index that maps checkpoints, hardware and VRAM starting points, runtimes, ComfyUI nodes, prompting tools, acceleration routes, and deployment options.
MiniMax-M2.7
MiniMaxAI/MiniMax-M2.7
A large MiniMax model focused on agentic work, software engineering, tool use, and complex productivity workflows.
For project maintainers
Listed here? You can use the badge.
If you maintain a project with a current LifeHubber listing, you may add the optional “Listed on LifeHubber AI Resources” badge to its README, docs, or website. No introduction or permission request is needed.
