AI & ML interests
None defined yet.
Recent Activity
Papers
View all Papers
A suit of multilingual MoE models with highly-sparse architectures
-
ATH-MaaS/Marco-Nano-Base
Text Generation • 8B • Updated • 76 • 17 -
ATH-MaaS/Marco-Mini-Base
Text Generation • 17B • Updated • 67 • 10 -
ATH-MaaS/Marco-Mini-Global-Base
Text Generation • 17B • Updated • 57 • 8 -
ATH-MaaS/Marco-Nano-Instruct
Text Generation • 8B • Updated • 956 • 43
Ovis-Image is a 7B text-to-image model specifically optimized for high-quality text rendering under stringent computational constraints.
Our next-generation MLLMs for native-resolution vision and advanced reasoning
Our latest advancement in multi-modal large language models (MLLMs)
Ovis1.5 is fully open-source: we release training datasets, training & inference codes, and model weights.
-
ATH-MaaS/Ovis-Omni-Embedding-3B
Feature Extraction • Updated • 700 • 64 -
ATH-MaaS/Ovis-VL-Embedding-2B
Feature Extraction • Updated • 489 • 40 -
ATH-MaaS/Ovis-VL-Embedding-9B
Feature Extraction • Updated • 497 • 48 -
Ovis-Embedding: Pushing the Frontiers of Universal Omni-Modal Embeddings
Paper • 2609.25165 • Published • 78
-
ATH-MaaS/Marco-DeepResearch-8B
Text Generation • 8B • Updated • 909 • 35 -
ATH-MaaS/Marco-DeepResearch-8B-GGUF
Text Generation • 8B • Updated • 682 • 7 -
ATH-MaaS/Marco-DeepResearch-8B-i1-GGUF
Text Generation • 8B • Updated • 996 • 8 -
ATH-MaaS/Marco-DeepResearch-8B-FP8
Text Generation • 8B • Updated • 32 • 1
Our advanced models and datasets for exploring the frontiers of MT
An unified model for multimodal understanding, text-to-image generation, and image editing.
With 29B parameters, Ovis1.6-Gemma2-27B achieves exceptional performance in the OpenCompass benchmark, ranking among the top-tier open-source MLLMs.
-
ATH-MaaS/Ovis1.6-Gemma2-27B
Image-Text-to-Text • 29B • Updated • 138 • 62 -
ATH-MaaS/Ovis1.6-Gemma2-9B
Image-Text-to-Text • 10B • Updated • 4.26k • 273 -
ATH-MaaS/Ovis1.6-Gemma2-9B-GPTQ-Int4
Image-Text-to-Text • Updated • 56 • 9 -
ATH-MaaS/Ovis1.6-Llama3.2-3B
Image-Text-to-Text • 4B • Updated • 7.53k • 50
-
ATH-MaaS/Ovis-Omni-Embedding-3B
Feature Extraction • Updated • 700 • 64 -
ATH-MaaS/Ovis-VL-Embedding-2B
Feature Extraction • Updated • 489 • 40 -
ATH-MaaS/Ovis-VL-Embedding-9B
Feature Extraction • Updated • 497 • 48 -
Ovis-Embedding: Pushing the Frontiers of Universal Omni-Modal Embeddings
Paper • 2609.25165 • Published • 78
-
ATH-MaaS/Marco-DeepResearch-8B
Text Generation • 8B • Updated • 909 • 35 -
ATH-MaaS/Marco-DeepResearch-8B-GGUF
Text Generation • 8B • Updated • 682 • 7 -
ATH-MaaS/Marco-DeepResearch-8B-i1-GGUF
Text Generation • 8B • Updated • 996 • 8 -
ATH-MaaS/Marco-DeepResearch-8B-FP8
Text Generation • 8B • Updated • 32 • 1
A suit of multilingual MoE models with highly-sparse architectures
-
ATH-MaaS/Marco-Nano-Base
Text Generation • 8B • Updated • 76 • 17 -
ATH-MaaS/Marco-Mini-Base
Text Generation • 17B • Updated • 67 • 10 -
ATH-MaaS/Marco-Mini-Global-Base
Text Generation • 17B • Updated • 57 • 8 -
ATH-MaaS/Marco-Nano-Instruct
Text Generation • 8B • Updated • 956 • 43
Ovis-Image is a 7B text-to-image model specifically optimized for high-quality text rendering under stringent computational constraints.
Our advanced models and datasets for exploring the frontiers of MT
Our next-generation MLLMs for native-resolution vision and advanced reasoning
An unified model for multimodal understanding, text-to-image generation, and image editing.
Our latest advancement in multi-modal large language models (MLLMs)
With 29B parameters, Ovis1.6-Gemma2-27B achieves exceptional performance in the OpenCompass benchmark, ranking among the top-tier open-source MLLMs.
-
ATH-MaaS/Ovis1.6-Gemma2-27B
Image-Text-to-Text • 29B • Updated • 138 • 62 -
ATH-MaaS/Ovis1.6-Gemma2-9B
Image-Text-to-Text • 10B • Updated • 4.26k • 273 -
ATH-MaaS/Ovis1.6-Gemma2-9B-GPTQ-Int4
Image-Text-to-Text • Updated • 56 • 9 -
ATH-MaaS/Ovis1.6-Llama3.2-3B
Image-Text-to-Text • 4B • Updated • 7.53k • 50
Ovis1.5 is fully open-source: we release training datasets, training & inference codes, and model weights.