LLM EXPLORER 61,652 MODELS INDEXED

ThaiLLM 8B Instruct by KBTG-Labs

By KBTG-Labs · 176 downloads
HF Repo ↗ Copy URL

ThaiLLM 8B Instruct is an open-source language model by KBTG-Labs. Features: 8b LLM, VRAM: 16.4GB, Context: 40K, License: apache-2.0, Instruction-Based, LLM Explorer Score: 0.22.

  Arxiv:2601.04597   Base model:merge:qwen/qwen3-8b Base model:merge:qwen/qwen3-8b... Base model:merge:thaillm/thail...   Base model:qwen/qwen3-8b   Base model:qwen/qwen3-8b-base   Base model:thaillm/thaillm-8b   Conversational   En   Instruct   Merge   Mergekit   Qwen3   Region:us   Safetensors   Sharded   Tensorflow   Th
Model Card on HF 🤗: https://huggingface.co/KBTG-Labs/ThaiLLM-8B-Instruct 

ThaiLLM 8B Instruct Parameters and Internals

LLM NameThaiLLM 8B Instruct
Repository 🤗https://huggingface.co/KBTG-Labs/ThaiLLM-8B-Instruct 
Base Model(s)  ThaiLLM/ThaiLLM-8B   Qwen3 8B   Qwen3 8B Base   ThaiLLM/ThaiLLM-8B   Qwen/Qwen3-8B   Qwen/Qwen3-8B-Base
Model Size8b
Required VRAM16.4 GB
Updated2026-09-19
MaintainerKBTG-Labs
Model Typeqwen3
Instruction-BasedYes
Model Files  4.9 GB: 1-of-4   4.9 GB: 2-of-4   5.0 GB: 3-of-4   1.6 GB: 4-of-4
Supported Languagesen th
Model ArchitectureQwen3ForCausalLM
Licenseapache-2.0
Context Length40960
Model Max Length40960
Transformers Version4.57.3
Tokenizer ClassQwen2Tokenizer
Padding Token<|endoftext|>
Vocabulary Size151936
Errorsreplace

Best Alternatives to ThaiLLM 8B Instruct

Best Alternatives
Context / RAM
Downloads
Likes
DeepAnalyze 8B128K / 16.4 GB51293
...epSeek R1 0528 Qwen3 8B Esper3128K / 32.7 GB153
WebWorld 8B40K / 16.4 GB95064
WebWorld 32B40K / 65.8 GB34275
WebWorld 14B40K / 29.5 GB50428
...nBias Plus SFT Instruct Legacy40K / 16.4 GB3831
...n3 8B UnBias Plus SFT Instruct40K / 16.4 GB541
QVikhr 3 8B Instruction40K / 16.4 GB384610
ATLAS 8B Instruct40K / 16.4 GB232
Qwen3 8B ShiningValiant340K / 32.7 GB113
Note: green Score (e.g. "73.2") means that the model is better than KBTG-Labs/ThaiLLM-8B-Instruct.