g-assismoraes/DeltaP2S-Llama2-13B-DeltaP2S-CodeLlama7B-SameFormula-S13

TEXT GENERATIONPricing:Input $1.5 / Output $2.1Concurrent Unit Cost:1Model Size:13BQuant:FP8Context Size:4kPublished:Aug 28, 2026Architecture:Transformer Featherless Exclusive Cold

The g-assismoraes/DeltaP2S-Llama2-13B-DeltaP2S-CodeLlama7B-SameFormula-S13 model is a 13 billion parameter language model, a merged checkpoint resulting from a Delta-P2S experiment. This model combines elements from Llama2-13B and CodeLlama7B, specifically trained using a 'SameFormula-S13' configuration. It is designed for tasks benefiting from the combined strengths of these foundational models, particularly in areas where code-related understanding and generation are relevant, leveraging its 4096 token context length.

Loading preview...

Model Overview

This model, developed by g-assismoraes, is a 13 billion parameter language model derived from a Delta-P2S experiment. It represents a merged checkpoint, integrating components from both Llama2-13B and CodeLlama7B architectures. The training process utilized a specific configuration identified as 'SameFormula-S13', indicating a tailored approach to its development.

Key Characteristics

  • Architecture Hybrid: Combines features from Llama2-13B and CodeLlama7B.
  • Parameter Count: 13 billion parameters, offering substantial capacity for complex tasks.
  • Context Length: Supports a context window of 4096 tokens.
  • Experimental Origin: Product of a standalone Delta-P2S experiment, suggesting a focus on specific research or optimization goals.

Potential Use Cases

Given its foundational models, this merged checkpoint is likely suitable for applications requiring:

  • Code-related tasks: Benefiting from the CodeLlama7B component, it may excel in code generation, completion, or understanding.
  • General language understanding and generation: Leveraging the Llama2-13B base for broader NLP tasks.
  • Research and experimentation: Particularly for those interested in the Delta-P2S methodology and its impact on model performance.