Open Source Kanana-2-vision-encoder
AI & ML interests
None defined yet.
Papers
Keep-or-Drop? Adaptive Tokenizer for Compact Video Representation
Let's Scale Step by Step: Compute-Efficient Hyperparameter Transfer for Large-Scale Mixture-of-Experts
Organization Card
Kakao’s AI, Kanana
We are excited to introduce Kanana, Kakao’s proprietary AI model family. Kanana is Kakao's in-house AI model lineup, featuring both Large Language Models(LLM) and Large Multimodal Models(LMM). We aim to deliver practical, high-performance, and efficient models that are optimized for real-world services.
models 28
kakaocorp/kanana-2-vision-encoder-tiny
Image Feature Extraction • 98.7M • Updated • 314 • 7
kakaocorp/kanana-2-vision-encoder-medium
Image Feature Extraction • 0.4B • Updated • 403 • 12
kakaocorp/lmspt
Audio-to-Audio • Updated • 3
kakaocorp/kanana-2-1.3b-instruct
Text Generation • 1B • Updated • 2.13k • 13
kakaocorp/kanana-2-30b-a3b-base
Text Generation • 31B • Updated • 256 • 29
kakaocorp/kanana-2-30b-a3b-instruct
Text Generation • 31B • Updated • 2.44k • 47
kakaocorp/kanana-2-30b-a3b-thinking
Text Generation • 31B • Updated • 216 • 39
kakaocorp/kanana-2-30b-a3b-base-2601
Text Generation • 31B • Updated • 264 • 31
kakaocorp/kanana-2-30b-a3b-mid-2601
Text Generation • 31B • Updated • 207 • 31
kakaocorp/kanana-2-30b-a3b-instruct-2601
Text Generation • 31B • Updated • 2.25k • 57
datasets 0
None public yet