ForSureTesterSim/Qwen2.5-R1-Minny-1.5B-v2

TEXT GENERATIONConcurrent Unit Cost:1Model Size:1.5BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Aug 23, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

ForSureTesterSim/Qwen2.5-R1-Minny-1.5B-v2 is a 1.5 billion parameter Small Language Model (SLM) developed by ForSureTesterSim, built upon the DeepSeek-R1-Distill-Qwen architecture. This model utilizes a novel Sens-Stock Fusion merging paradigm to achieve a Pareto-optimal balance of mathematical reasoning, code generation, and conversational instruction-following. It is specifically optimized for complex mathematical problems and syntax generation, leveraging a 32K context length.

Loading preview...

What is Qwen2.5-R1-Minny-1.5B-v2?

Qwen2.5-R1-Minny-1.5B-v2 is an experimental 1.5 billion parameter Small Language Model (SLM) from ForSureTesterSim, designed for advanced mathematical reasoning, code generation, and instruction-following. It introduces a unique Sens-Stock Fusion methodology, which combines gradient-based layer routing (Sens-Merging) with geometric projection (Model Stock) to precisely merge specialized expert models without traditional backpropagation.

Key Capabilities & Unique Features

  • Novel Merging Paradigm: Employs "Sens-Stock Fusion" to mathematically address directional and magnitude issues in model merging, resulting in a highly specialized SLM.
  • Optimized for Reasoning: Constructed by merging three expert fine-tunes (Code, Math, Chat) onto a DeepSeek-R1-Distill-Qwen base, ensuring strong performance in pure mathematical reasoning and syntax generation.
  • "Golden Triangle" Topology: Merges DeepSeek-R1-Distill-Qwen-1.5B (base) with DeepCoder-1.5B-Preview (code), RLinf-math-1.5B (math), and DeepSeek-R1-ReDistill-Qwen-1.5B-v1.1 (chat/structural).
  • Context Length: Supports a substantial 32,768 token context window.
  • Chain-of-Thought (CoT) Integration: Inherently relies on <think> tags for complex reasoning, utilizing the Qwen ChatML format.

When to Use This Model

This model is particularly well-suited for use cases requiring:

  • Mathematical Problem Solving: Excels in tasks demanding pure mathematical reasoning and logical deduction.
  • Code Generation: Strong capabilities in generating syntactically correct and optimized code.
  • Complex Instruction Following: Designed to handle intricate instructions, especially when combined with Chain-of-Thought prompting.
  • Resource-Constrained Environments: As a 1.5B parameter model, it offers advanced capabilities in a smaller footprint.

Finetunes
1 models