astom-M/matsuo-llm-advanced-phase-f4a

TEXT GENERATIONConcurrency Cost:1Model Size:7.6BQuant:FP8Ctx Length:32kPublished:Feb 22, 2026Architecture:Transformer Cold

The astom-M/matsuo-llm-advanced-phase-f4a is a 7.6 billion parameter language model, merged using the DARE TIES method with Qwen/Qwen2.5-7B-Instruct as its base. This model combines the strengths of two previous merges, 'phase_d' and 'phase_e2b', aiming to balance strong performance in both 'ALF' and 'DB' metrics. It is designed for general language understanding and generation tasks, leveraging its merged architecture for enhanced capabilities.

Loading preview...

Model Overview

The astom-M/matsuo-llm-advanced-phase-f4a is a 7.6 billion parameter language model created through a sophisticated merging process. It utilizes the DARE TIES merge method, building upon the robust Qwen/Qwen2.5-7B-Instruct as its foundational base model.

Key Characteristics

  • Merged Architecture: This model is a composite of two prior merges, phase_d and phase_e2b, strategically combined to leverage their individual strengths.
  • DARE TIES Method: The use of the DARE TIES method, as detailed in the arxiv.org/abs/2311.03099 paper, indicates a focus on efficient and effective parameter merging.
  • Balanced Performance Goal: The merge configuration explicitly aims to integrate the 'ALF' strength from phase_d (56%) with the 'DB' strength from phase_e2b (53.47%), suggesting an intent to achieve well-rounded performance across different evaluation criteria.
  • Base Model: Inherits capabilities from the Qwen/Qwen2.5-7B-Instruct model, known for its strong general language understanding and generation.

Intended Use Cases

This model is suitable for applications requiring a balanced performance profile, particularly where the combined strengths in 'ALF' and 'DB' metrics are beneficial. Its merged nature suggests potential for improved generalization and robustness compared to its individual constituent merges.