Kwai-Keye/Keye-VL-2.0-30B-A3B-GGUF · Hugging Face
Kwai-Keye releases Keye-VL-2.0-30B-A3B-GGUF on Hugging Face, a 30B-parameter model focused on long-video understanding and agent capabilities. The model leads open-source competitors and matches Gemini-3-Flash on temporal grounding benchmarks. Keye-VL-2.0-30B-A3B features a DSA-native long-context architecture with sparse attention. You can access the model via Hugging Face for local deployment.
Key takeaways
- 30B-parameter model for long-video understanding and agent capabilities.
- Leads open-source competitors and matches Gemini-3-Flash on temporal grounding.
- DSA-native architecture with sparse attention for long-context handling.
Kwai-Keye releases Keye-VL-2.0-30B-A3B-GGUF on Hugging Face, a 30B-parameter model focused on long-video understanding and agent capabilities. The model leads open-source competitors and matches Gemini-3-Flash on temporal grounding benchmarks. Keye-VL-2.0-30B-A3B features a DSA-native long-context architecture with sparse attention. You can access the model via Hugging Face for local deployment.
Key takeaways
- 30B-parameter model for long-video understanding and agent capabilities.
- Leads open-source competitors and matches Gemini-3-Flash on temporal grounding.
- DSA-native architecture with sparse attention for long-context handling.