appvoid/palmer-003
TEXT GENERATIONConcurrency Cost:1Model Size:1.1BQuant:BF16Ctx Length:2kPublished:Jan 26, 2024License:apache-2.0Architecture:Transformer Open Weights Cold

palmer-003 is a 1.1 billion parameter general language model developed by appvoid, designed to be a highly capable small model. It features a 2048-token context length and demonstrates competitive performance against other models in its size class across various benchmarks, including MMLU, ARC-C, and HellaSwag. This model is optimized for general language understanding and generation tasks, aiming for broad applicability in resource-constrained environments.

Loading preview...

palmer-003: A Capable Small Language Model

palmer-003 is appvoid's latest 1.1 billion parameter general language model, developed with the goal of creating a highly effective small-scale model. It is designed for broad applicability in various language understanding and generation tasks.

Key Capabilities & Performance

This model demonstrates competitive performance across several common benchmarks when compared to other models of similar or slightly larger parameter counts. Its evaluation scores include:

  • MMLU: 0.2523
  • ARC-C: 0.3439
  • OBQA: 0.3740
  • HellaSwag: 0.6208
  • PIQA: 0.7524
  • Winogrande: 0.6590

These results indicate that palmer-003 achieves an average score of 0.5004 across these benchmarks, positioning it favorably against models like tinyllama-3t and zyte-1b, and closely approaching the performance of the larger qwen-1.8b in some areas. The model operates with a context length of 2048 tokens.

Ideal Use Cases

palmer-003 is well-suited for applications requiring a compact yet capable language model, particularly where computational resources or deployment size are constraints. Its general-purpose nature makes it adaptable for tasks such as text generation, summarization, and question answering in scenarios where larger models might be impractical.