Miiyamoto255/Aetheris-Core-2B

TEXT GENERATIONPricing:Input $0.32 / Cached $0.064 / Output $1.6Concurrent Unit Cost:1Model Size:2.6BQuant:BF16Context Size:8kPublished:Aug 30, 2026Architecture:Transformer0.0K Featherless Exclusive Cold

Aetheris-Core-2B by Miiyamoto255 is a compact 2.6 billion parameter language model built on the Gemma2ForCausalLM architecture, featuring an 8192-token context length. It is specifically designed for fast, local inference with a strong emphasis on coding, reasoning, creative generation, and open-ended problem solving. This model aims to minimize unnecessary refusals, providing direct and developer-friendly responses for technical, creative, and research-oriented requests. Its core strength lies in its balanced capability for local deployment while excelling in programming and complex reasoning tasks.

Loading preview...

Aetheris-Core-2B: A Developer-Friendly 2.6B Parameter Model

Aetheris-Core-2B, developed by Miiyamoto255, is a 2.6 billion parameter language model built from scratch on the Gemma2ForCausalLM architecture. Designed for efficient local inference, it balances capability with memory usage and speed, making it suitable for a wide range of applications on consumer hardware. The model features an 8192-token context length, enabling it to handle substantial inputs for complex tasks.

Key Capabilities

  • Programming & Software Engineering: Excels in various languages (Python, JavaScript, C/C++), debugging, algorithms, and code architecture.
  • Reasoning: Capable of multi-step analysis, problem decomposition, troubleshooting, and technical explanations.
  • Developer Assistance: Provides guidance on refactoring, optimization, error diagnosis, and project planning.
  • Creative Generation: Supports character, world, dialogue, and story creation, along with general brainstorming.
  • General Knowledge: Offers explanations, summaries, learning assistance, and question answering.

Design Philosophy & Differentiation

Aetheris-Core-2B is engineered to be highly useful and direct, minimizing unnecessary refusals often seen in heavily constrained assistant-style models. It intelligently interprets requests, avoiding automatic rejections based on keywords, and handles legitimate experimentation, research, and fictional scenarios without undue restriction. The model prioritizes clear and efficient communication, avoiding excessive disclaimers or artificial verbosity. This approach makes it particularly valuable for developers and researchers seeking an unhindered, technically focused AI assistant.

Finetunes
1 models