Mincofficial/Minico-M1-Preview
Mincofficial/Minico-M1-Preview is a 0.35 billion parameter causal language model, fine-tuned from LiquidAI/LFM2.5-350M. This model is specifically designed to support visible reasoning blocks and configurable effort presets (low, medium, high, max) through its unique tokenizer chat template. It is optimized for tasks requiring explicit reasoning control and is suitable for research into model thought processes.
Loading preview...
Minico-M1-Preview Overview
Minico-M1-Preview is a 0.35 billion parameter language model developed by Mincofficial, fine-tuned from the LiquidAI/LFM2.5-350M base model. Its primary distinguishing feature is the integration of a specialized tokenizer chat template that enables visible <think>...</think> reasoning blocks and supports configurable reasoning effort presets, including low, medium, high, and max.
Key Capabilities
- Explicit Reasoning Control: Allows users to specify the level of reasoning effort the model should apply to a task.
- Visible Thought Process: Generates output that includes explicit reasoning steps within
<think>tags, offering transparency into the model's decision-making. - Fine-tuned for Reasoning: Leverages the QyrouNnet-AI/exp-reasoning-effort-control dataset for its fine-tuning, enhancing its ability to perform controlled reasoning.
Important Note on Checkpoint
The provided model.safetensors is a float16 dequantized export of a Q8_0 GGUF artifact (Minico-M1.gguf), not the original full-precision training checkpoint. Its weights are approximate and intended for interoperability and preview purposes.
Good For
- Research into model reasoning and thought processes.
- Applications requiring explicit control over the model's computational effort for generating responses.
- Use cases where understanding the model's internal steps is beneficial.