my-ai-stack/Stack-3.0-Omni-Nexus

TEXT GENERATIONConcurrency Cost:1Model Size:7.6BQuant:FP8Ctx Length:32kPublished:Apr 16, 2026License:apache-2.0Architecture:Transformer0.0K Open Weights Cold

Stack 3.0 Omni Nexus is an 8x7B Mixture-of-Experts (MoE) model developed by my-ai-stack, featuring approximately 56 billion total parameters with 14 billion active during inference. Optimized for enterprise workloads, it excels in advanced code generation, complex reasoning, and multilingual tasks. The model supports a substantial context length of 131,072 tokens, making it suitable for processing extensive inputs.

Loading preview...

Stack 3.0 Omni Nexus: An 8x7B Mixture-of-Experts Model

Stack 3.0 Omni Nexus is an 8x7B Mixture-of-Experts (MoE) model from my-ai-stack, designed for enterprise applications requiring robust performance in code generation, complex reasoning, and multilingual processing. With 56 billion total parameters, it efficiently operates with only 14 billion active parameters per forward pass, balancing performance and resource usage.

Key Capabilities & Performance

This model demonstrates strong performance across various benchmarks, often surpassing larger models like Llama 3.1 70B and Mixtral 8x7B in specific areas:

  • Code Generation: Achieves 82.0% on HumanEval (pass@1) and 78.5% on MBPP (pass@1), indicating superior coding abilities.
  • Reasoning: Scores 91.2% on GSM8K (5-shot) for mathematical reasoning.
  • Competitive Benchmarks: Rated 1842 on CodeForces, outperforming comparative models.
  • Extensive Context: Supports a significant context window of 131,072 tokens (128K), enabling processing of large documents and complex prompts.

Architecture & Efficiency

The MoE architecture utilizes 8 experts, with 2 active per forward pass, contributing to its efficiency. It offers various GGUF quantization options, including a balanced Q4_K_M variant requiring approximately 3.5 GB of VRAM.

Ideal Use Cases

Stack 3.0 Omni Nexus is particularly well-suited for:

  • Software Development: Generating full-stack applications, code refactoring, and complex programming tasks.
  • Finance: Quant modeling and developing trading systems.
  • Healthcare & Legal: Building specialized software, ensuring compliance, and automating document processing.
  • Education: Creating course content and generating educational materials.

Finetunes
2 models