{ "id": "epfdw4e7es", "slug": "sea-lion-epfdw4e7es", "title": "SEA-LION", "updated": "2026-10-21", "topic_path": [ { "id": "technology", "label": "Technology and the built world", "api_url": "https://www.edgepedia.org/api/v1/topics/technology" }, { "id": "technology.computing", "label": "Computing and digital systems", "api_url": "https://www.edgepedia.org/api/v1/topics/technology.computing" }, { "id": "technology.computing.modern_ai", "label": "Modern AI: foundation models, generative AI, and the AI industry", "api_url": "https://www.edgepedia.org/api/v1/topics/technology.computing.modern_ai" }, { "id": "technology.computing.modern_ai.models", "label": "Model families and named models", "api_url": "https://www.edgepedia.org/api/v1/topics/technology.computing.modern_ai.models" }, { "id": "technology.computing.modern_ai.models.llm_families", "label": "Large language model families", "api_url": "https://www.edgepedia.org/api/v1/topics/technology.computing.modern_ai.models.llm_families" }, { "id": "technology.computing.modern_ai.models.llm_families.open-weight-model-families", "label": "Open-weight model families", "api_url": "https://www.edgepedia.org/api/v1/topics/technology.computing.modern_ai.models.llm_families.open-weight-model-families" } ], "geo": [ { "id": "geo.asia.t2021.technology.computing.modern_ai", "label": "Asia (other) · 2021 and later: Modern AI: foundation models, generative AI, and the AI industry", "api_url": "https://www.edgepedia.org/api/v1/geo/geo.asia.t2021.technology.computing.modern_ai", "path": [ { "id": "geo.asia", "label": "Asia (other)", "api_url": "https://www.edgepedia.org/api/v1/geo/geo.asia" }, { "id": "geo.asia.t2021", "label": "Asia (other) · 2021 and later", "api_url": "https://www.edgepedia.org/api/v1/geo/geo.asia.t2021" }, { "id": "geo.asia.t2021.technology", "label": "Technology and the built world", "api_url": "https://www.edgepedia.org/api/v1/geo/geo.asia.t2021.technology" }, { "id": "geo.asia.t2021.technology.computing", "label": "Computing and digital systems", "api_url": "https://www.edgepedia.org/api/v1/geo/geo.asia.t2021.technology.computing" }, { "id": "geo.asia.t2021.technology.computing.modern_ai", "label": "Modern AI: foundation models, generative AI, and the AI industry", "api_url": "https://www.edgepedia.org/api/v1/geo/geo.asia.t2021.technology.computing.modern_ai" } ] } ], "excerpt": "SEA-LION (Southeast Asian Languages in One Network) is a family of open large language models built by AI Singapore, first released in December 2023 to serve Southeast Asian languages.", "snippet": "SEA-LION (Southeast Asian Languages in One Network) is a family of open large language models built by AI Singapore, first released in December 2023 to serve Southeast Asian languages.", "node": "technology.computing.modern_ai.models.llm_families.open-weight-model-families", "markdown": "# SEA-LION\n\nSEA-LION (Southeast Asian Languages in One Network) is a family of open large language models built by [AI Singapore](https://www.edgepedia.org/ai-singapore) to serve the languages of Southeast Asia, first released in December 2023 and, as of September 2026, spanning text, multimodal, reasoning, embedding, and safety model lines from 300 million to 70 billion parameters.[1](https://github.com/aisingapore/sealion/blob/main/README.md) • [2](https://docs.sea-lion.ai/) The project is anchored by the Products Pillar of AI Singapore and forms part of Singapore's National Multi-Modal LLM Project (NMLP).[3](https://sea-lion.ai/about/) AI Singapore's stated motivation is to create models for under-represented population groups and low-resource languages in the region, lowering adoption barriers for governments, enterprises, academia, and end users, and expanding Southeast Asian language and cultural representation in mainstream LLMs.[3](https://sea-lion.ai/about/)\n\nThe family covers eleven Southeast Asian languages: English, Chinese, Indonesian, Vietnamese, Malay, Thai, Burmese, Lao, Filipino, Tamil, and Khmer.[4](https://aclanthology.org/2025.ijcnlp-long.30.pdf) AI Singapore describes SEA-LION as Southeast Asia's first family of open-source LLMs designed to understand the region's contexts, languages, and cultures, with variants built on Gemma, Qwen, and Llama bases.[5](https://sea-lion.ai/models/) It is a family rather than a single model: base, instruction-tuned, multimodal, embedding, and safety-oriented derivatives are distributed through GitHub, Hugging Face, and the sea-lion.ai API.[6](https://sgai.md/opensource/sea-lion/)\n\n| Key fact | Detail |\n|---|---|\n| Maker | AI Singapore, under the National Multi-Modal LLM Project[3](https://sea-lion.ai/about/) |\n| First release | December 2023 (v1, pre-trained from scratch)[1](https://github.com/aisingapore/sealion/blob/main/README.md) |\n| Latest release | v4.5 (E2B and 27B instruction models)[1](https://github.com/aisingapore/sealion/blob/main/README.md) |\n| Languages | 11 SEA languages, English through Khmer[4](https://aclanthology.org/2025.ijcnlp-long.30.pdf) |\n| Parameter range | 300M (embedding) to 70B (v3/v3.5)[1](https://github.com/aisingapore/sealion/blob/main/README.md) |\n| Context length | 2,048 tokens (v1) to 262K tokens (v4.5 27B)[1](https://github.com/aisingapore/sealion/blob/main/README.md) |\n| Licensing | MIT where possible; Llama-based variants carry the Llama3 Community License[2](https://docs.sea-lion.ai/) • [7](https://docs.sea-lion.ai/models/sea-lion-v2) |\n\n## Release timeline and versions\n\n**Version 1 (December 2023).** The first SEA-LION was pre-trained from scratch at 3B and 7B parameter sizes on the MPT architecture, with a 256K vocabulary and a 2,048-token context, trained on 980 billion tokens across the 11 SEA languages and released under the MIT license.[1](https://github.com/aisingapore/sealion/blob/main/README.md) The vendor reports that v1 outperformed most models on SEA-HELM (Southeast Asian Holistic Evaluation of Language Models) when it was released.[1](https://github.com/aisingapore/sealion/blob/main/README.md)\n\n**Version 2 (July 2024).** SEA-LION v2 shifted strategy: an 8-billion-parameter model continued-pretrained on top of Llama 3 8B Instruct, with an 8,192-token context, using approximately 48 billion tokens across five SEA languages (English, Indonesian, Tamil, Thai, and Vietnamese).[7](https://docs.sea-lion.ai/models/sea-lion-v2) The instruct variant was fine-tuned on around 100,000 English instruction-completion pairs plus roughly 50,000 pairs from other ASEAN languages such as Indonesian, Thai, and Vietnamese.[7](https://docs.sea-lion.ai/models/sea-lion-v2) Unlike v1, v2 was released under the Llama3 Community License rather than MIT.[7](https://docs.sea-lion.ai/models/sea-lion-v2)\n\n**Version 3.** The peer-reviewed technical report (IJCNLP-AACL 2025, with an April 2025 arXiv version) introduces Llama SEA-LION-8B-IT, continued-pretrained and fine-tuned on Llama-3.1-8B-Instruct, and Gemma-SEA-LION-9B-IT, built on Gemma-2-9B.[4](https://aclanthology.org/2025.ijcnlp-long.30.pdf) • [8](https://arxiv.org/html/2504.05747) Training used 200 billion tokens of English, code, and SEA languages for continued pre-training and 16.8 million English and SEA-language instruction-answer pairs for post-training.[4](https://aclanthology.org/2025.ijcnlp-long.30.pdf)\n\n**Version 3.5.** A reasoning line of 8B and 70B models, each a supervised fine-tune of the corresponding Llama-SEA-LION-v3 IT model with 128K context.[1](https://github.com/aisingapore/sealion/blob/main/README.md)\n\n**Version 4 (August 2025).** AI Singapore's first collection of multimodal models trained on Southeast Asian text, handling image-plus-text inputs with up to 256K native context windows and specialized regional OCR.[9](https://github.com/aisingapore/sealion/blob/main/models/sea-lion-v4/sea-lion-v4.md) • [1](https://github.com/aisingapore/sealion/blob/main/README.md) The line includes 4B and 8B vision-language variants fine-tuned from Qwen3-VL-4B-Instruct and Qwen3-VL-8B-Instruct, a 32B instruct model fine-tuned from Qwen3-32B (with 4-bit and 8-bit quantized releases), and a 27B variant continued-pretrained from [Gemma 3](https://www.edgepedia.org/gemma-3) 27B IT with 128K context.[2](https://docs.sea-lion.ai/) • [1](https://github.com/aisingapore/sealion/blob/main/README.md)\n\n**Version 4.5.** The latest line specializes existing open foundation models through knowledge distillation and model merging, delivering reasoning and agentic tool-use capabilities: an E2B model (128K context, a fine-tune of gemma-4-E2B-it) and a 27B model (262K context, a fine-tune of Qwen3.6-27B).[1](https://github.com/aisingapore/sealion/blob/main/README.md)\n\n**Derivatives.** The family also includes SEA-GUARD safety-oriented models, listed at 4B and 8B sizes (128K context) in vendor documentation, with the GitHub README additionally describing 12B variants built on Qwen-SEA-LION-v4 VLM, Llama-SEA-LION-v3-8B-IT, and Gemma 3 12B IT bases; the two vendor sources do not fully agree on the size list.[1](https://github.com/aisingapore/sealion/blob/main/README.md) • [2](https://docs.sea-lion.ai/) SEA-LION Embedding models exist at 300M and 600M sizes with 8K context, built from ModernBERT (trained from scratch) and E5-Large respectively.[1](https://github.com/aisingapore/sealion/blob/main/README.md)\n\n## Architecture and training as published\n\nThe family's training approach changed direction after v1. Version 1 was the only release pre-trained from scratch, on the MPT architecture with a custom 256K-vocabulary tokenizer, at a cost of 980 billion training tokens.[1](https://github.com/aisingapore/sealion/blob/main/README.md) From v2 onward, every release has been a continued pre-training or supervised fine-tune of an existing open foundation model: Llama 3 and [Llama](https://www.edgepedia.org/llama-model-family) 3.1 for the 8B and 70B lines, Gemma 2 and Gemma 3 for the 9B and 27B lines, and Qwen3 and Qwen3-VL for the 32B and vision-language variants.[7](https://docs.sea-lion.ai/models/sea-lion-v2) • [8](https://arxiv.org/html/2504.05747) • [2](https://docs.sea-lion.ai/)\n\nFor v3, the published recipe combines large-scale multilingual continued pre-training (200 billion tokens of English, code, and SEA languages) with a multi-stage post-training regime of instruction fine-tuning, alignment, and model merging.[4](https://aclanthology.org/2025.ijcnlp-long.30.pdf) Stage 2 instruction fine-tuning drew on 22 existing datasets in English, Thai, and Vietnamese plus 22 newly formulated synthetic SEA instruction datasets, totalling 7,298,828 instructions.[4](https://aclanthology.org/2025.ijcnlp-long.30.pdf) Later versions compress this pipeline: v4.5 relies on knowledge distillation and model merging to specialize new foundation models quickly.[1](https://github.com/aisingapore/sealion/blob/main/README.md)\n\n## Benchmarks: vendor versus independent\n\nAll benchmark claims in the record are vendor-authored. The technical report's authors state that SEA-LION models achieve state-of-the-art performance on SEA-HELM and the [Open LLM Leaderboard](https://www.edgepedia.org/open-llm-leaderboard) against similar-size regional models such as Sailor 2 and SeaLLMs 3.[4](https://aclanthology.org/2025.ijcnlp-long.30.pdf) The vendor's own site claims SEA-LION v4 (Gemma 27B, Qwen 32B) ranked #1 among open models under 200 billion parameters for SEA tasks.[3](https://sea-lion.ai/about/) No independent third-party evaluation of SEA-LION on SEA-HELM or arena-style leaderboards appears in the available record, so these claims rest on the maker's own measurements.[4](https://aclanthology.org/2025.ijcnlp-long.30.pdf) • [3](https://sea-lion.ai/about/)\n\n## By the numbers\n\n- Parameters: 300M and 600M (embedding models) up to 70B (v3 and v3.5), with text models at 3B, 4B, 7B, 8B, 9B, 12B, 27B, and 32B across versions.[1](https://github.com/aisingapore/sealion/blob/main/README.md) • [2](https://docs.sea-lion.ai/)\n- Training tokens: 980 billion for v1 (from scratch), about 48 billion for v2's continued pre-training, 200 billion for v3's continued pre-training.[1](https://github.com/aisingapore/sealion/blob/main/README.md) • [7](https://docs.sea-lion.ai/models/sea-lion-v2) • [4](https://aclanthology.org/2025.ijcnlp-long.30.pdf)\n- Instruction data: 16.8 million instruction-answer pairs for v3 post-training; 7,298,828 instructions in v3's stage 2 alone; about 150,000 pairs total for v2.[4](https://aclanthology.org/2025.ijcnlp-long.30.pdf) • [7](https://docs.sea-lion.ai/models/sea-lion-v2)\n- Context length: 2,048 tokens (v1), 8,192 (v2), 128K (v3.5, v4, v4.5 E2B), and 262K (v4.5 27B).[1](https://github.com/aisingapore/sealion/blob/main/README.md) • [7](https://docs.sea-lion.ai/models/sea-lion-v2)\n\n## Licensing, availability and cost\n\nSEA-LION's stated licensing policy is to release under the fully open MIT license where possible; v1 was released under MIT, while the licensing of the v3 models varies by model and may include the underlying Llama or Gemma license terms.[4](https://aclanthology.org/2025.ijcnlp-long.30.pdf) • [1](https://github.com/aisingapore/sealion/blob/main/README.md) • [10](https://docs.sea-lion.ai/overview) However, exact terms vary with the underlying base model: Llama-based variants such as v2 may be bound by the Llama3 Community License, which places certain restrictions on commercial use, and Gemma-based variants carry Gemma's own terms.[7](https://docs.sea-lion.ai/models/sea-lion-v2) • [1](https://github.com/aisingapore/sealion/blob/main/README.md) • [2](https://docs.sea-lion.ai/) A company can therefore deploy and fine-tune the MIT-licensed releases without restriction, but must check the base-model license for Llama- and Gemma-derived variants. Models are distributed via GitHub, Hugging Face, and the sea-lion.ai API.[6](https://sgai.md/opensource/sea-lion/)\n\n## References\n\n1. aisingapore/sealion README (official repository) — https://github.com/aisingapore/sealion/blob/main/README.md\n2. SEA-LION Documentation — Model Collections — https://docs.sea-lion.ai/\n3. About SEA-LION (AI Singapore) — https://sea-lion.ai/about/\n4. SEA-LION: Southeast Asian Languages in One Network (IJCNLP-AACL 2025) — https://aclanthology.org/2025.ijcnlp-long.30.pdf\n5. Models | SEA-LION — https://sea-lion.ai/models/\n6. SEA-LION · Singapore AI Observatory — https://sgai.md/opensource/sea-lion/\n7. SEA-LION v2 model card — https://docs.sea-lion.ai/models/sea-lion-v2\n8. SEA-LION: Southeast Asian Languages in One Network (arXiv, April 2025) — https://arxiv.org/html/2504.05747\n9. SEA-LION v4 release notes (GitHub) — https://github.com/aisingapore/sealion/blob/main/models/sea-lion-v4/sea-lion-v4.md\n10. SEA-LION, SEA-LION Documentation, https://docs.sea-lion.ai/overview\n\n---\n*Topic: Encyclopedia › Technology and the built world › Computing and digital systems › Modern AI: foundation models, generative AI, and the AI industry › Model families and named models › Large language model families › Open-weight model families*\n\n*Initially written Sep 17, 2026 · Edited: Oct 11, 2026; Oct 15, 2026; Oct 21, 2026*\n\n*Copyright 2026 EdgeChat AI, a subsidiary of Biostate AI.*\n\nLicense: Edgepedia Community License 1.0, https://www.edgepedia.org/edgepedia/license\n", "same_as": [], "url": "https://www.edgepedia.org/sea-lion-epfdw4e7es", "markdown_url": "https://www.edgepedia.org/sea-lion-epfdw4e7es.md", "license": { "name": "Edgepedia Community License 1.0", "url": "https://www.edgepedia.org/edgepedia/license", "summary": "Free with credit, commercial use included. AI training is open to everyone. For other uses, organizations over USD 100M in revenue or 100M monthly users license separately.", "spdx": "LicenseRef-Edgepedia-Community-1.0" }, "credit": "\"SEA-LION\", Edgepedia (EdgeChat), https://www.edgepedia.org/sea-lion-epfdw4e7es. Edgepedia Community License 1.0.", "credit_md": "\"[SEA-LION](https://www.edgepedia.org/sea-lion-epfdw4e7es)\", Edgepedia (EdgeChat), [https://www.edgepedia.org/sea-lion-epfdw4e7es](https://www.edgepedia.org/sea-lion-epfdw4e7es). [Edgepedia Community License 1.0](https://www.edgepedia.org/edgepedia/license).", "credit_html": "\"SEA-LION\", Edgepedia (EdgeChat), https://www.edgepedia.org/sea-lion-epfdw4e7es. Edgepedia Community License 1.0.", "speakable": "SEA-LION is a family of open large language models built by AI Singapore, first released in December 2023 to serve Southeast Asian languages." }