Elsephire/Qwen3.6-35B-A3B-Nex-N2-mini-80-20-Beta

TEXT GENERATIONConcurrent Unit Cost:3Model Size:35.1BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Jul 1, 2026License:mitArchitecture:Transformer Open Weights Featherless Exclusive Cold

Elsephire/Qwen3.6-35B-A3B-Nex-N2-mini-80-20-Beta is a 35.1 billion parameter language model created by Elsephire through a Task Arithmetic merge of Qwen3.6-35B-A3B and Nex-N2-mini. This model leverages an 80/20 weighting ratio using Minerve-Fusion, offering a unique blend of capabilities from its constituent models. With a 32768 token context length, it is designed for general language tasks, representing a beta release for community testing and feedback.

Loading preview...

Model Overview

Elsephire/Qwen3.6-35B-A3B-Nex-N2-mini-80-20-Beta is a 35.1 billion parameter language model developed by Elsephire. It was created using a Task Arithmetic merging technique, combining the strengths of two distinct base models: Qwen3.6-35B-A3B and Nex-N2-mini. The merge was executed with an 80/20 weighting ratio, indicating a primary emphasis on the Qwen3.6-35B-A3B component, and utilized the Minerve-Fusion tool, which is slated for future open-source release.

This model supports a substantial context length of 32768 tokens, allowing it to process and generate longer, more coherent texts. As a beta release, it is specifically put forth for community testing and feedback, encouraging developers to explore its performance and identify areas for improvement.

Key Capabilities

  • Merged Architecture: Combines features from Qwen3.6-35B-A3B and Nex-N2-mini.
  • Task Arithmetic: Utilizes a specific merging strategy for balanced performance.
  • Extended Context: Processes inputs up to 32768 tokens.

Good For

  • General Language Tasks: Suitable for a wide range of text generation and understanding applications.
  • Experimentation: Ideal for developers interested in testing merged models and providing feedback on beta releases.
  • Exploring Model Blends: Offers insights into the performance characteristics of models combined via Task Arithmetic.