Azure99/Blossom-V6.4-9B

VISIONConcurrency Cost:1Model Size:9BQuant:FP8Ctx Length:32kTool Calling:SupportedPublished:May 30, 2026License:apache-2.0Architecture:Transformer0.0K Open Weights Cold

Azure99/Blossom-V6.4-9B is a 9 billion parameter open-source conversational large language model developed by Azure99. It is designed to be a cost-effective, locally accessible general-purpose model, building upon the V6.3 training recipe with added multimodal samples. Blossom-V6.4-9B excels in generating reproducible post-training data through a sophisticated data synthesis workflow involving multiple advanced LLMs.

Loading preview...

Blossom-V6.4-9B: An Open-Source Conversational LLM

Blossom-V6.4-9B is a 9 billion parameter open-source conversational large language model from Azure99, focused on providing a powerful and cost-effective general-purpose model accessible locally. This iteration largely maintains the V6.3 training methodology, incorporating a small number of multimodal samples to retain base model capabilities.

Key Differentiators & Capabilities

  • Reproducible Post-Training Data: A core focus on generating high-quality, reproducible post-training data.
  • Advanced Data Synthesis Workflow: Employs a sophisticated, cost-effective data synthesis process utilizing Deepseek-V3.1, Gemini 2.5 Flash, and Qwen3-235B-A22B-Instruct-2507. This workflow includes:
    • Objective Scenario Validation: For tasks like mathematics, models cross-verify responses, with inconsistencies filtered out.
    • Subjective Scenario Cross-Evaluation: Models evaluate each other's responses, with roles (respondent/evaluator) randomly assigned to mitigate bias.
  • Rule-Based Filtering: Incorporates N-Gram filtering to remove repetitive data and discards toxic content.
  • Multimodal Preservation: Includes multimodal samples to maintain capabilities from its base models.

Ideal Use Cases

Blossom-V6.4-9B is well-suited for developers and researchers seeking:

  • A general-purpose conversational model with a strong emphasis on data quality and reproducibility.
  • Applications requiring robust data synthesis and filtering mechanisms.
  • Cost-effective local deployment of a capable LLM.