SteelStorage/L3.1-MS-Astoria-70b-v2
Hugging Face
TEXT GENERATIONConcurrency Cost:4Model Size:70BQuant:FP8Ctx Length:32kPublished:Oct 21, 2024Architecture:Transformer0.0K Warm

SteelSkull's L3.1-MS-Astoria-70b-v2 is a 70 billion parameter language model built on the Llama 3.1 architecture, featuring a 32K context window. This model stock merge aims to combine the robust storytelling capabilities of multiple base models while maintaining high intelligence. It is designed for complex narrative generation and general-purpose conversational AI, supporting both Llama 3 and 'meth' instruction formats.

Loading preview...

L3.1-MS-Astoria-70b-v2: Merged Storytelling and Intelligence

L3.1-MS-Astoria-70b-v2 is a 70 billion parameter model developed by SteelSkull, based on the Llama 3.1 architecture. This model represents a "model stock" merge, combining several Llama 3.1-based models to achieve a balance between robust storytelling and general intelligence. It leverages a 32,768-token context window, making it suitable for extended conversations and complex narrative generation.

Key Capabilities

  • Enhanced Storytelling: Designed to integrate the narrative strengths of multiple models, aiming for coherent and engaging long-form content.
  • Maintained Intelligence: Focuses on preserving the core intelligence of its constituent Llama 3.1 models.
  • Flexible Instruction Formats: Supports both the standard Llama 3 instruction format and a specialized "meth" format for stepped thinking, particularly useful when Llama 3's default behavior is insufficient.
  • Model Stock Merge: Built from a diverse set of Llama 3.1-70B models, including mlabonne/Llama-3.1-70B-Instruct-lorablated, migtissera/Tess-3-Llama-3.1-70B, NeverSleep/Lumimaid-v0.2-70B, Sao10K/L3.1-70B-Euryale-v2.2, and ArliAI/Llama-3.1-70B-ArliAI-RPMax-v1.2, and nbeerbower/Llama3.1-Gutenberg-Doppel-70B.

Good For

  • Applications requiring advanced narrative generation and creative writing.
  • Use cases benefiting from a large context window for detailed interactions.
  • Developers seeking a model that balances creative output with strong reasoning capabilities.
Popular Sampler Settings

Top 3 parameter combinations used by Featherless users for this model. Click a tab to see each config.

temperature
top_p
top_k
frequency_penalty
presence_penalty
repetition_penalty
min_p