OpenThaiGPT 1.0.0 Beta 13B Chat
OpenThaiGPT 1.0.0 Beta is a 13B LLaMA v2 Chat fine-tune trained on translated Thai instructions. Vocabulary was expanded by 10,000+ Thai tokens to speed up tokenization. Both Thai and English are supported; it uses the standard Llama 2 chat prompt format.
This is an early-stage community project, not a production-ready Thai model. The translated-instruction approach and beta label both signal you should expect rough edges in Thai output quality. If you have the VRAM and just need something to prototype Thai chat flows, it works — but set expectations low. Skip it for anything customer-facing until a stable release with stronger benchmarks exists.
›Why this rating
Auto-generated rating (Opus 4.7 judge, claude-opus-4-7). Overall 9.02/10. License is explicit (Apache 2.0) on both HF tags and card, and commercial use is confirmed by the card itself. Metadata (13B, LLaMA v2 base, Thai+English) matches the card; context length of 4096 is the Llama 2 default and a reasonable inference. The editorial voice is honest about beta status, translated training data, and weak community traction. Best use case is reasonably specific (Thai instruction following/Q&A), though it could be sharper. Brand fit is moderate — niche Thai-language use case but legitimately serves the local-AI builder audience in that region.
Flags: - Context length 4096 is inferred from Llama 2 base, not explicitly stated in the card excerpt — acceptable but worth noting - Low popularity (992 downloads, 2 likes) — correctly disclosed in weaknesses
Overview
OpenThaiGPT 1.0.0 Beta is a 13B LLaMA v2 Chat fine-tune trained on translated Thai instructions. Vocabulary was expanded by 10,000+ Thai tokens to speed up tokenization. Both Thai and English are supported; it uses the standard Llama 2 chat prompt format.
Strengths
- 10,000+ Thai tokens added to vocabulary — faster tokenization vs. base LLaMA v2
- Fine-tuned on Thai instruction data for basic Thai task following
- Apache 2.0 license — commercial use allowed
- Llama 2 Chat base means prompt format is well-documented
Weaknesses
- Still a beta release — accuracy and reliability are explicitly unguaranteed
- Training data is translated instructions, not native Thai — quality ceiling is real
- Very low community traction (992 downloads, 2 likes on HF) — limited real-world validation
- 13B at fp16 needs ~26 GB VRAM; not lightweight
Quantization variants
Each quantization trades model quality for file size and VRAM. Q4_K_M is the most popular starting point.
| Quantization | File size | VRAM required |
|---|---|---|
| Q4_K_M | 7.2 GB | 10 GB |
Get the model
HuggingFace
Original weights
Source repository — direct quantization required.
Hardware that runs this
Cards with enough VRAM for at least one quantization of OpenThaiGPT 1.0.0 Beta 13B Chat.
Models worth comparing
Same parameter band, plus what's one tier above and below — so you can decide what actually fits your hardware.
Frequently asked
What's the minimum VRAM to run OpenThaiGPT 1.0.0 Beta 13B Chat?
Can I use OpenThaiGPT 1.0.0 Beta 13B Chat commercially?
What's the context length of OpenThaiGPT 1.0.0 Beta 13B Chat?
Source: huggingface.co/openthaigpt/openthaigpt-1.0.0-beta-13b-chat-hf
Reviewed by RunLocalAI Editorial. See our editorial policy for how we research and verify model claims.
Related — keep moving
Verify OpenThaiGPT 1.0.0 Beta 13B Chat runs on your specific hardware before committing money.