MediaTek-Research
2 models across 1 family · Latest: Breeze 7B (2023-11)
Pioneering edge AI advancements in research
MediaTek-Research's portfolio covers 2 active models across 1 current family, spanning general LLM work. Open a model detail page to compare provider routes and sourced benchmarks.
Covers 0 workload areas across 2 active tracked models; last verified 2026-05-19.
Use it for
- Teams evaluating general LLM work across this lab's releases
- Comparing model families before committing to a flagship
- Migration and pricing follow-ups across 2 tracked models
Do not use it for
- Choosing a hosting provider without opening a model page for price ladders
Active models
2
Current models from this lab, excluding deprecated ones
Active families
1
Current model families from this lab
Open catalog
2 open
0 open source / 2 open weights
Lowest output price
Not tracked
No provider output pricing linked yet
Latest dated release
2023-11-10
Breeze 7B
Freshness
2026-05-19
Researched 60d ago
Information
Release cadence
Showing 2 recent dated releases (full timeline below). Latest: Breeze 7B (2023-11-10).
Where this lab wins
Not enough capability or benchmark coverage yet to call strengths for this lab.
Flagship quality / price signal
Flagship: Breeze 7B (best sourced coding quality-per-dollar in this portfolio).
Quality-per-dollar unavailable for this flagship — benchmark coverage or output token pricing is still missing.
MediaTek-Research is an AI research organization founded in 1997. Pioneering edge AI advancements in research. MediaTek-Research ships 1 model family totaling 2 models, with the most recent release Breeze 7B in 2023-11. Notable families include Breeze. Use it as a stable reference for lab background, release coverage, and follow-up model pages as they are added. Researchers and evaluators can scan counts, links. View official API endpoints, benchmark performance, and coding/agent fit for every MediaTek-Research model.
About
MediaTek Research, a specialized unit within the Global MediaTek Group, is pioneering advancements in artificial intelligence (AI) and machine learning (ML). Established with the vision of unlocking AI potential to create intelligent, ethical, secure, and sustainable systems, MediaTek Research has gathered accomplished researchers from diverse backgrounds, including computer science, engineering, mathematics, and physics. Their research centers located strategically in Cambridge (UK) and National Taiwan University enhance collaboration with global institutions and academics, providing a rich environment for innovation and knowledge exchange. Beyond foundational research, MediaTek is deeply committed to building an edge computing ecosystem to accelerate AI application development across diverse devices such as smartphones, IoT devices, vehicles, and smart homes. This effort is reflected in their partnerships with tech giants like Google and Meta to leverage cutting-edge technologies, such as Llama 2, to enhance on-device generative AI capabilities. The company’s efforts to seamlessly integrate generative AI into everyday technology demonstrate its leading role in the advancement of AI. MediaTek Research's contributions to generative AI and Large Language Models (LLMs) are particularly noteworthy. With the publication of numerous seminal papers covering Bayesian optimization, reinforcement learning, and the development of LLMs optimized for Traditional Chinese, MediaTek has positioned itself at the forefront of AI research. Notable among their innovations is the open-source Large Language Model, MediaTek Research Breeze-7B, optimized for both Traditional Chinese and English. Based on the Mistral model, this seven-billion parameter model excels in processing speed and efficiency, significantly enhancing the application's capabilities in handling complex Traditional Chinese inferences. Moreover, MediaTek has patented various technologies related to AI processing, such as optimizing deep neural networks and improving wireless communication systems. They have also extended efforts to bolster the pre-training processes of models like BLOOM for enhanced support of Traditional Chinese. These advancements underscore the company's technical expertise and commitment to pushing the boundaries of AI research. In collaboration with academic institutions and government bodies, MediaTek Research advances real-world applications of generative AI. Collaborations like those with the Taipei City Government and National Taipei University of Technology in developing data-searching chatbots enhance productivity while demonstrating the practical impacts of their research. Their active participation in industry events, such as the India Mobile Congress (IMC), emphasizes their leadership in shaping the AI sector's future. Through a strategic blend of research excellence, open-source initiatives, and meaningful collaborations, MediaTek Research continues to make substantial contributions to the field of generative AI and LLMs.
Featured models
| Model | Released | Context | Input price ($/1M) | Output price ($/1M) | License | Openness |
|---|---|---|---|---|---|---|
| Breeze 7B | 2023-11-10 | 32k | - | - | Llama 2 Community | Open weights |
| Breeze 8x7B | 2023-11-10 | 32k | - | - | Llama 2 Community | Open weights |
Model families
Recent releases
- Breeze 7B- 2023-11-10
- Breeze 8x7B- 2023-11-10
FAQ
Who founded MediaTek-Research and when?
MediaTek-Research was founded in 1997 and is associated with Hsinchu, Taiwan.
What models has MediaTek-Research released?
MediaTek-Research ships 2 models across 1 family: Breeze.
Is MediaTek-Research's technology open source?
All tracked models are released under Open Weights.
Where is MediaTek-Research headquartered?
MediaTek-Research is headquartered in Hsinchu, Taiwan.
What is MediaTek-Research known for?
Pioneering edge AI advancements in research. Its most prominent tracked family is Breeze.
How can I access MediaTek-Research's models?
MediaTek-Research's models are available via NVIDIA NIM.
Explore related pages
Last reviewed: 2026-05-19. Data sourced from public lab announcements and provider documentation.