SEA-LION v4.5 (Latest)
SEA-LION version 4.5, released in May 2026, is our latest collection of foundational, agentic, and multimodal models optimized for Southeast Asia. Utilizing advanced post-training methodologies—such as knowledge distillation and model merging—this suite delivers state-of-the-art regional fluency, precise tool use, and high computational efficiency.
Gemma-SEA-LION-v4.5 (E2B Series)
Our highly efficient 4-billion parameter series built on Gemma 4. It is optimized for precise function-calling, structured JSON outputs, and autonomous agentic tool-use with minimal memory overhead.
Refer to the detailed page to access model cards, usage scripts, and deployment configurations for the entire Gemma family:
Core Agentic Foundations
Qwen-SEA-LION-v4.5 (27B Series)
Our high-capacity flagship causal and vision-language series built on Qwen3.6. It features a native 262K context window, robust multi-turn reasoning, and repository-level coding adapted for regional linguistic and cultural contexts.
Refer to the detailed page to explore full benchmarks, technical specifications, and configuration details for the entire Qwen family:
Base & Multimodal Foundations
High-Throughput Booster (SpecDecoder version)
Last updated