Kimi-VL is the efficient open MoE VLM (2.8B active) from Moonshot AI. Excels at multimodal reasoning, long context (128K) & agent tasks. 'Thinking' variant available.
Kimi-VL from Moonshot AI is an efficient open-weight MoE Vision-Language Model packing strong multimodal reasoning and agent skills into just 2.8B active parameters (16B total).
It comes in a general Instruct version, which is great for agents and a Thinking version optimized with CoT/RL for complex visual reasoning. Both handle 128K context and leverage their MoonViT vision encoder.
Congrats on your recent Product Hunt launch ā I came across your product and would be happy to provide feedback! I'm also currently collecting feedback on my own tool, WebChatter, which helps users interact with web content (articles, PDFs, and videos) using AI.
If you're up for it, Iād love to set up a quick 20-minute call ā 10 minutes to hear about your product and share thoughts, and 10 minutes for you to try out WebChatter and offer feedback in return.
Replies
Flowtica Scribe
Hi everyone!
Kimi-VL from Moonshot AI is an efficient open-weight MoE Vision-Language Model packing strong multimodal reasoning and agent skills into just 2.8B active parameters (16B total).
It comes in a general Instruct version, which is great for agents and a Thinking version optimized with CoT/RL for complex visual reasoning. Both handle 128K context and leverage their MoonViT vision encoder.
The models are openly available. You can try it here.
Congrats on your recent Product Hunt launch ā I came across your product and would be happy to provide feedback! I'm also currently collecting feedback on my own tool, WebChatter, which helps users interact with web content (articles, PDFs, and videos) using AI.
If you're up for it, Iād love to set up a quick 20-minute call ā 10 minutes to hear about your product and share thoughts, and 10 minutes for you to try out WebChatter and offer feedback in return.
Looking forward to connecting!
Best,
Hossein