shb777/Llama-3.3-8B-Instruct-128K

TEXT GENERATIONConcurrency Cost:1Model Size:8BQuant:FP8Ctx Length:32kPublished:Dec 30, 2025License:llama3.3Architecture:Transformer0.0K Cold

The shb777/Llama-3.3-8B-Instruct-128K is an 8 billion parameter instruction-tuned language model, based on the Llama 3.3 architecture. This model features an extended context length of 128K tokens, significantly enhanced by added `rope_scaling` and an updated generation configuration. It is primarily designed for conversational AI and instruction-following tasks requiring extensive context, making it suitable for applications like long-form content generation, summarization, and complex question-answering.

Loading preview...

Llama 3.3 8B 128K Instruct Overview

This model, shb777/Llama-3.3-8B-Instruct-128K, is an enhanced version of the original allura-forge/Llama-3.3-8B-Instruct. It is an 8 billion parameter instruction-tuned language model, specifically designed to handle very long contexts.

Key Capabilities

  • Extended Context Window: Features a substantial 128K token context length, enabling processing and generation of very long texts.
  • Improved Context Handling: Incorporates rope_scaling to enhance performance and stability with its large context window.
  • Optimized Instruction Following: Benefits from an updated generation configuration and an Unsloth chat template in its tokenizer, making it highly effective for instruction-based tasks and conversational AI.

Good for

  • Long-form Content Generation: Ideal for creating extensive articles, reports, or creative writing pieces that require maintaining coherence over many pages.
  • Advanced Summarization: Excels at summarizing large documents, books, or lengthy conversations.
  • Complex Question Answering: Suitable for answering questions that require synthesizing information from a vast amount of provided context.
  • Conversational AI: Its instruction-tuned nature and extended context make it well-suited for chatbots that need to remember and reference long dialogue histories.