Smilyai-labs/Mira-1-large-0919

TEXT GENERATIONPricing:Input $0.48 / Output $0.96Concurrent Unit Cost:1Model Size:14BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Sep 19, 2026License:otherArchitecture:Transformer0.0K Featherless Exclusive Cold

Mira-1-large-0919 by SmilyAI is a 14.8 billion parameter conversational and reasoning language model, built on Bc-AI/Nova-2-Large and Qwen3-14B, with a 40,960-token context window. It is fine-tuned for programming, mathematics, technical discussions, and complex reasoning tasks. This model excels in providing programming assistance, mathematical problem-solving, and engaging in technical conversations with a distinctive, expressive style.

Loading preview...

Overview

Smilyai-labs/Mira-1-large-0919 is a 14.8 billion parameter language model developed by SmilyAI, designed for technical and conversational interaction. Built upon the Bc-AI/Nova-2-Large base model, which itself is derived from Qwen3-14B and incorporates a DARE-TIES merge for code and mathematics, Mira features a substantial 40,960-token context window.

Key Capabilities

Mira's fine-tuning focuses on several core areas:

  • Reasoning: Excels in multi-step problem-solving, breaking down complex issues, identifying assumptions, and debugging.
  • Programming: Optimized for Python development, machine learning, AI engineering, code review, and debugging.
  • Mathematics: Assists with algebra, numerical problems, applied mathematics, and explaining mathematical concepts.
  • Conversational Style: Features a friendly, curious, energetic, and technically enthusiastic personality, including an optional "Roast Mode" for constructive code review.

Intended Use Cases

This model is particularly well-suited for:

  • Programming and AI/ML assistance
  • Debugging and code review
  • Mathematical problem solving and explanations
  • Technical research and engineering discussions
  • Conversational AI experimentation

Limitations

Users should be aware that Mira, like other generative models, can produce hallucinations, incorrect code, mathematical errors, and misinterpret ambiguous instructions. Outputs, especially for critical tasks, should always be independently verified. The model is licensed for proprietary, private use only, with commercial deployment requiring specific terms from SmilyAI.

Finetunes
1 models