LLM EXPLORER 61,918 MODELS INDEXED

Starchat Beta GPTQ by TheBloke

By TheBloke · 57 downloads
HF Repo ↗ Copy URL

Starchat Beta GPTQ is an open-source language model by TheBloke. Features: 15.8b LLM, VRAM: 8.9GB, License: bigcode-openrail-m, Quantized, Code Generating, LLM Explorer Score: 0.07.

  4-bit   Codegen   Endpoints compatible   Generated from trainer   Gpt bigcode   Gptq   Quantized   Region:us   Safetensors
Model Card on HF 🤗: https://huggingface.co/TheBloke/starchat-beta-GPTQ 

Starchat Beta GPTQ Parameters and Internals

Model Type 
GPT-like, coding assistant
Use Cases 
Areas:
research, educational purposes
Applications:
coding tasks, chat
Limitations:
uncensored model may generate problematic text, bias towards demographics in GitHub community
Additional Notes 
StarChat Beta is not aligned with human preferences and can produce problematic outputs. It is primarily trained on code data.
Supported Languages 
English (primary), Programming Languages (80+)
Training Details 
Data Sources:
openassistant-guanaco dataset, OpenAssistant/oasst1 dataset
Methodology:
fine-tuned using uncensored variant of dataset
Hardware Used:
multi-GPU setup with 8 devices
LLM NameStarchat Beta GPTQ
Repository 🤗https://huggingface.co/TheBloke/starchat-beta-GPTQ 
Model Size15.8b
Required VRAM8.9 GB
Updated2026-07-29
MaintainerTheBloke
Model Typegpt_bigcode
Model Files  8.9 GB
GPTQ QuantizationYes
Quantization Typegptq
Generates CodeYes
Model ArchitectureGPTBigCodeForCausalLM
Licensebigcode-openrail-m
Transformers Version4.28.1
Tokenizer ClassGPT2Tokenizer
Vocabulary Size49156
Torch Data Typebfloat16
Activation Functiongelu

Best Alternatives to Starchat Beta GPTQ

Best Alternatives
Context / RAM
Downloads
Likes
Sqlcoder2 GPTQ0K / 9.2 GB1211
Sqlcoder GPTQ0K / 9.2 GB119
Octocoder GPTQ0K / 9.2 GB68
Redmond Hermes Coder GPTQ0K / 9.2 GB5110
Starcoderplus GPTQ0K / 8.9 GB5225
Starcoder GPTQ0K / 8.9 GB6713
Note: green Score (e.g. "73.2") means that the model is better than TheBloke/starchat-beta-GPTQ.