AISEAN

Latest SEA-LION generation built on NVIDIA Nemotron 3 (Nano & Super). Includes continued-pretrained base checkpoints and post-trained instruct variants. Adapted with SEA, reasoning, code, and multilingual parallel data. Supports 256K context. Available in GGUF, NVFP4, FP8 quantizations.

SEA-LION's first multimodal (vision-language) release, built on Gemma 3. Ranked #1 for Tamil and Filipino and #4 overall on SEA-HELM leaderboard out of 55 models. Efficient enough to run on a 32GB consumer laptop. Supports image + text inputs with SEA cultural grounding.

Built on Qwen3-32B foundation. Significant improvement in multilingual accuracy and cultural contextual understanding for SEA. Transitions to BPE tokenizer for superior text handling. Includes Qwen3-VL-based vision-language variants (4B, 8B). Efficient enough to run on a 32GB RAM laptop.

Indonesia's sovereign LLM co-built with AI Singapore on SEA-LION foundations. Upgraded to 70B parameters (Llama 3.1 70B base) in 2025 with 128K context. Supports 6 languages including Batak Toba and Balinese. All data and GPU infrastructure stored within Indonesian territory (GPU Merdeka sovereign cloud). Available via sahabat-ai.com and GoPay app.

Malaysia's first domestically built government-backed sovereign LLM, launched August 2025 by the National AI Office. Reflects Malaysian cultural values and constitutional principles. Hosted within YTL's 500MW Green Data Centre in Malaysia. Multilingual across Malaysia's four main language communities (Malay, English, Mandarin, Tamil).

Vietnam's pioneering open-source generative LLM series, pre-trained from scratch on a 102B-token / 482GB cleaned Vietnamese corpus. Uses a custom byte-level BPE tokenizer with 20K vocabulary types tailored for Vietnamese morphology. PhoGPT-4B-Chat fine-tuned on 300K+ conversations. Excels in conversational agents, document processing, and domain-specific QA.

First open-source Vietnamese reasoning LLM, released September 2025. Represents Vietnam's next-generation sovereign AI model after PhoGPT, with built-in chain-of-thought reasoning for Vietnamese. Packaged on NVIDIA NIM for single-H100 deployability.

Thailand's leading open-weight Thai LLM family from SCB 10X — covers text, reasoning, audio (speech-in/out), OCR (document parsing), and multimodal variants. Outperforms other Thai LLMs on O-NET / A-Level exams. Deployed in Thai government services via SCBX–OPDC partnership.

Alibaba DAMO's SEA-focused multilingual LLM series, instruction-tuned for Southeast Asian languages with emphasis on cultural alignment. Regularly benchmarked on SEA-HELM alongside SEA-LION and Sailor2.

Sovereign AI platform for Malaysian enterprises and government agencies. Hosts AI entirely within Malaysia using Malaysian data, ensuring data privacy, sovereignty, and local regulatory compliance.

Flagship open-source multilingual multimodal LLM family for SEA. v4 introduced multimodal (image+text) with 256K native context; v4.5 added reasoning & agentic tool-use via knowledge distillation; v4.8 (Nemotron, MoE) extends to 13 regional languages with enhanced multi-turn chat, cultural competence (KALAHI benchmark) and country-entity knowledge. Companion SEA-Guard safety layer and SEA-LION-Embedding suite also released.

Family of multilingual LLMs optimised for SEA languages. SeaLLMs-v3 achieves SOTA on diverse tasks with reduced hallucination and improved cultural safety. Babel extends coverage to the top 25 global languages (>90% of global speakers). SeaLLMs-Audio adds multimodal audio support for SEA languages.

State-of-the-art open-source Vietnamese generative LLMs pre-trained from scratch on a 102B-token Vietnamese corpus. Includes base pre-trained models and instruction-following/chat variants (PhoGPT-4B-Chat, PhoGPT-7B5-Instruct). Primary practical choice for Vietnamese-specific fine-tuning and deployment.

Open-source multilingual LLM family purpose-built for Southeast Asia. Continued pre-training on 200B SEA-language tokens with instruction-tuning, alignment, and model merging. Achieves state-of-the-art on SEA-HELM benchmark. Companion tools: SEA-HELM (eval) and SEA-Guard (safety).

Thai-first LLM family with highest performance among Thai open models. Includes multimodal variants for image and audio (Typhoon2-Audio). Built on Llama 3.1. Evaluated on ThaiExam, M3Exam, IFEval-TH, MT-Bench-TH.

Open-weight Malay-first LLM pre-trained from scratch on a 90B-token Malaysian text corpus. Captures local slang, code-switching, and regional dialects with greater fidelity than fine-tuned alternatives. Includes instruction-tuned chat variants. Designed for data-sovereign, on-premise deployment.

Malaysia's sovereign LLM developed in partnership with Nvidia, hosted in YTL's 500MW green data centre. Natively processes colloquial Malaysian Bahasa, mixed dialects, and code-switching. Deployed to 3M+ users via YTL/Yes mobile. Targets top-10 open-source benchmark rankings for non-localised disciplines.

Flagship instruction-tuned multilingual LLM from AI Singapore's National Multi-Modal LLM Project, built via continued pre-training and multi-stage instruction fine-tuning on top of Llama 3. Part of the SEA-LION family which also includes a 70B variant, a Gemma-based 9B model, a vision-language model (Gemma-SEA-LION-v4-27B-VL), and the SEA-Guard safety models. Evaluated via the SEA-HELM benchmark co-developed with Stanford CRFM.

Cutting-edge instruction-tuned vision-language model from the SEA-LION v4 family, built on Gemma architecture. Designed for multimodal (vision + text) tasks across Southeast Asian languages and cultural contexts. Released March 2026.

Malaysia's first fully sovereign LLM, 100% developed, owned, and operated in Malaysia by YTL AI Labs. Showcased at the ASEAN AI Malaysia Summit 2025. Trained on local Malaysian language data and cultural context. Emphasis on AI sovereignty — all systems fully controlled within Malaysian jurisdiction. Available via early access chatbot.

Southeast Asian Languages in One Network — the flagship open-source SEA LLM family from Singapore's national AI programme. Continued pre-training on ~500B tokens across 11 SEA languages with full instruction-tuning, alignment, and model-merging pipeline. Companion tools: SEA-HELM (eval benchmark) and SEA-Guard (safety).

Indonesia's sovereign LLM initiative by GoTo Group (Gojek/Tokopedia). Built on Llama 3 with continued pre-training and instruction fine-tuning on Indonesian-language data. Aimed at enterprise and government deployments in Indonesia, with a focus on Bahasa Indonesia fluency and local cultural context.

Flagship open multimodal SEA LLM, continually pre-trained on Gemma 3 27B with ~500B tokens across 11 SEA languages. Supports vision–language tasks, long-context understanding, and native function calling. Developed with Google Cloud VTC infrastructure.

Tops the SEA-HELM open-source leaderboard (under 200B params). Built on Qwen3-32B, continually pre-trained on ~100B tokens across 7 SEA languages. Efficient enough to run on a 32 GB consumer laptop.

Latest open-source Thai LLM milestone with agentic design (tool use, multi-step reasoning), improved Thai fluency and tone, and high throughput efficiency. Previous Typhoon 2 offered 5 model sizes with multimodal (image + audio) variants.

Production-grade Thai LLM series. Typhoon 1.5X featured a 70B variant rivalling leading models. Typhoon 2 added multimodal (image and audio) processing. Offers a free API via opentyphoon.ai and AWS-native deployment.

Multilingual SEA LLM series from Alibaba DAMO Academy, built via continual pre-training on SEA language corpora, targeting linguistic and cultural alignment across the region. One of the earliest dedicated regional multilingual LLM initiatives.

Open Thai LLM series from Thailand's national electronics and computer technology centre (NECTEC) and VISTEC. Fine-tuned on Thai instruction datasets. WangchanX is the broader initiative covering Thai NLP tooling and benchmarks.

Flagship open SEA LLM built on Qwen 32B; tops SEA-HELM leaderboard for open instruct models <200B. Pre-trained on 100B regional tokens from SEA-PILE-v2. 32k context window. Multimodal & reasoning-capable. Companion models: SEA-Guard (safety, Feb 2026) & SEA-LION-Embedding (Mar 2026).

Latest generation of Thailand's premier Thai-English bilingual LLM family, built on Qwen3. Includes Typhoon2.5 4B and Typhoon2.5 30B A3B (MoE, 3B active). Family also includes reasoning model T1, multimodal, and translation-specific variants. Highest-performing Thai LLM on ThaiExam & M3Exam benchmarks.

Thai-English bilingual LLM built on Gemma 3 base. Available in 4B and 12B sizes. Part of the broader Typhoon 2 family which also includes 1B, 3B, 8B instruct variants and a dedicated Typhoon-Translate 4B model for Thai↔English translation.

Malaysia's government-backed sovereign LLM launched August 2025, tailored for Bahasa Malaysia and hosted within YTL's 500MW Green Data Centre in Malaysia. Developed in partnership with NVIDIA. Targets e-commerce, media, and telecommunications anchor use cases. Aligned with Malaysia's national AI roadmap.

Cambodia's first dedicated open-source Khmer LLM, trained on 50M tokens of Khmer text. Developed by Angkor Intelligence. A separate Khmer variant of SEA-LION 7B is being co-developed via an MoU between AI Forum Cambodia and AI Singapore (signed Jan 2025), marking the first official Khmer integration into the SEA-LION ecosystem.

Southeast Asian Languages in One Network — the flagship open-source SEA LLM family by AI Singapore, continued-pretrained on Llama 3 and Gemma 2 backbones over ~1T tokens spanning 11 SEA languages. Includes SEA-HELM evaluation benchmark and SEA-Guard safety layer.

Thai-first bilingual LLM series by SCB 10X, optimized for Thai language understanding and instruction-following. Typhoon 2.5 is built on Qwen3; includes multimodal (vision + audio) variants and a Thai/English reasoning model (T1). Available via opentyphoon.ai API.

Thailand's leading open-source LLM series optimised for Thai language, culture, and tokenisation. Typhoon 2.x adds multimodal (Typhoon Vision), audio (Typhoon Audio), Thai–English translation, and speech-to-text (incl. Isan dialect). Used in Thai government services and enterprise deployments.

The first open-source Vietnamese reasoning LLM, released by GreenNode (VNG's AI-cloud unit) in September 2025. Packaged on NVIDIA NIM and deployable on a single H100 GPU. Positioned as Vietnam's post-VinAI sovereign reasoning model.

Southeast Asian Large Language Models by Alibaba DAMO Academy, released as part of the SEA open-model ecosystem. SeaLLMs v3 (7B) supports a broad suite of SEA languages via continued pre-training and instruction tuning. One of the earliest (Dec 2023) SEA-focused open LLMs alongside SEA-LION, covering Burmese, Khmer, Lao, Malay, Thai, Vietnamese, Indonesian, and more.

First open-source Vietnamese reasoning LLM, released September 2025. Packaged on NVIDIA NIM for single-H100 deployability. Represents Vietnam's next-generation sovereign AI model after PhoGPT's freeze, with built-in chain-of-thought reasoning for Vietnamese.

Southeast Asia's first family of open-source, multilingual, multimodal LLMs. v4 is built on Gemma 3 27B with 128K context, image+text understanding, function calling, and SEA-focused post-training. v4.5 expands agentic capabilities with a speculative decoder for up to 6× efficiency.

Thailand's flagship Thai-optimised LLM series. Typhoon 2.5 focuses on agentic AI with multi-step reasoning, improved function calling, high-throughput inference (3,000+ tokens/sec on H100), and superior Thai fluency. Multimodal variants (audio + vision) also available.

Southeast Asian Languages In One Network — a family of open-source LLMs fine-tuned for SEA languages and cultures. v4 (late 2025) introduced multimodality on Gemma 3 27B; v4.5 (March 2026) adds agent capabilities, a speculative decoder for 6× throughput, SEA-Guard safety models, and SEA-LION-Embedding suite. Supports 11 SEA languages. Part of Singapore's National Multi-Modal LLM Project.

Thailand's leading open-source Thai LLM series. Typhoon 2 features 5 model sizes, multimodal (image + audio) capabilities, and state-of-the-art Thai instruction-following. Typhoon 2.1 Gemma is featured in Google DeepMind's Gemmaverse. Typhoon 2.5 is available via opentyphoon.ai API. Deployed in Thai government services (OPDC partnership).

Indonesia's sovereign open-source LLM, built on SEA-LION with AISG support. Upgraded to 70B parameters in June 2025 with multilingual chat service. Operates across Bahasa Indonesia and 4 local languages. All data and GPU infrastructure stored within Indonesian territory (GPU Merdeka sovereign cloud). Available via sahabat-ai.com and GoPay app.

Vietnam's pioneering open-source generative LLM, pre-trained from scratch on a 102B-token Vietnamese corpus (482 GB cleaned). PhoGPT-4B-Chat is fine-tuned on 70K instruction prompts and 290K conversations. Uses a custom byte-level BPE tokenizer with 20K vocabulary types tailored for Vietnamese morphology.

SEA-LION v4 adds multimodal capability (vision + language) to the family. Reasoning enhanced in v3.5. Part of Singapore's S$70M National Multimodal LLM Programme. Full SEA-LION ecosystem: base LLM, embeddings, safety guard.

SOTA on SEA-HELM multilingual benchmark. Continued pre-training on Llama-3.1-8B and Gemma-2-9B. 200B tokens, 16.8M instruction pairs. NVIDIA-optimized via TensorRT-LLM. MIT license. Widely deployed across ASEAN.

SOTA retrieval embeddings for 10 SEA languages. Tested on SEA-BED (Southeast Asia Embedding Benchmark) with human-curated native data. Sets new records on retrieval, reranking, and semantic textual similarity. Essential for RAG in SEA languages.

Dedicated safety layer for SEA-LION family. Culturally attuned — understands SEA-specific harmful content, cultural sensitivities, and local contexts. Launched Feb 4, 2026. Holistic NLP benchmarks + handcrafted SEA cultural diagnostic tests.

Multimodal Empathetic Reasoning and Learning in One Network. Singapore's first AudioLLM. Trained on 62M multimodal samples, 260K hours of audio. SOTA on Singlish ASR. Fuses MERaLiON-Whisper encoder + SEA-LION text decoder. Free API trial available.

Built on Qwen2.5, continual pre-training on 500B tokens (400B SEA-specific). Sailor2-20B achieves 50-50 win rate vs GPT-4o across SEA languages. 13 SEA languages. Reproducibility cookbook published. Community-driven.

Comprehensive SEA multilingual LLM. SeaLLM-13B outperforms ChatGPT-3.5 on SEA languages. Strong on low-resource languages like Lao and Khmer. SeaLLMs-Audio extension released Mar 2025.

First large audio-language model for SEA. Voice interactions across 5 languages. Built on Qwen2-Audio-7B. SOTA on SeaBench-Audio for Indonesian, Thai, and Vietnamese. Complements SeaLLMs text models.

Thai-English open LLM family from Siam Commercial Bank. Typhoon2-Instruct for chat, Typhoon2-Audio for Thai speech (SOTA). Collaboration with AI Singapore on cross-lingual audio. Most capable open Thai model.

Thai sovereign LLM trained to understand Thai context and culture. Released by NSTDA. NVIDIA subsidiary investment for development. Government-owned, deployed across Thai public sector.

Open-source Thai GPT fine-tuned on Thai instruction datasets. Community-driven. Built on Llama and Mistral. Active development with Thai government and academic support. Most-used open Thai instruction model.

Pre-trained RoBERTa-based Thai language model. Trained on 78.5GB Thai text. Best-performing Thai BERT-class model. Used as backbone for Thai NLP tasks across industry and academia.

Vietnamese-first LLM pre-trained from scratch on 102B Vietnamese tokens. Avoids copyright issues by training on curated corpus. PhoGPT-4B-Chat fine-tuned on 300K+ conversations.

Vietnamese instruction-tuned LLM based on Mistral-7B. Community-built. Strong Vietnamese chat performance. Most downloaded open Vietnamese chat model on HuggingFace.

Compact Vietnamese multimodal LLMs (vision + language). VinAI's efficient models optimized for mobile and edge deployment. Strong performance on Vietnamese visual understanding tasks.

Malaysia's first domestically built sovereign LLM. Reflects Malaysian cultural values and constitutional principles. Launched Aug 2025 by National AI Office. Multilingual for Malaysia's four main language communities.

Family of Malay language models trained on 90B tokens from Malaysian contexts. First significant dedicated Malay LLM. Used in Malaysian government and enterprise. Mesolitica also maintains Malaya NLP library.

Indonesian instruction-tuned LLM. Built on Llama-2-7B fine-tuned with Indonesian instruction datasets. Community-developed to address Indonesian language gap in major LLMs.

Multilingual BERT for Indonesian and regional languages. Covers Javanese and Sundanese — two of the world's most-spoken under-resourced languages. Used in IndoNLU benchmark evaluation.

First fully open-source SEA LLM — all training data and code disclosed. Uses parallel data for continual pre-training of OLMo 2. Addresses LLM data transparency risks. Aimed at reproducibility.