Azure99/Blossom-V7-27B

Overview

Blossom-V7-27B is a 27 billion parameter, general-purpose multimodal model from Azure99, built upon the Qwen3.5 architecture. It is designed for local deployment and integrates advanced capabilities for reasoning, tool use, and image understanding. The model supports a substantial context length, handling up to 262,144 tokens, with 131,072 tokens recommended for optimal performance.

Key Capabilities

• Efficient Adaptive Thinking: Dynamically scales reasoning depth based on task difficulty, producing concise reasoning traces significantly shorter than comparable models.
• Tool Use with Interleaved Thinking: Reasons and makes decisions before each tool call, enhancing performance in agentic tasks.
• Image Understanding: Processes and understands image inputs alongside text.
• Long Context: Supports up to 262,144 tokens, with 131,072 tokens recommended for best results.
• Faster Inference: Incorporates Multi-Token Prediction (MTP) for speculative decoding in vLLM and llama.cpp.
• Broad Application: Post-trained for general assistant use, covering everyday conversation, world knowledge, mathematics, reasoning, coding, web development, and data visualization.

Good For

• Agentic Workflows: Its interleaved thinking and tool use capabilities make it suitable for tasks requiring decision-making and interaction with external tools.
• Multimodal Applications: Ideal for scenarios requiring both text and image understanding.
• Local Deployment: Optimized for efficient local execution across various hardware configurations, with variants tailored for GPU, CPU, and memory-constrained environments.
• Complex Reasoning Tasks: Benefits from adaptive thinking to tackle challenging problems in mathematics, coding, and general reasoning.