AdarshSingh7647/HETU-GLM-Z1-9B-MathReasoning-CotCond

TEXT GENERATIONPricing:Input $0.431 / Cached $0.0862 / Output $1.12Concurrent Unit Cost:1Model Size:9BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Aug 26, 2026Architecture:Transformer Featherless Exclusive Cold

AdarshSingh7647/HETU-GLM-Z1-9B-MathReasoning-CotCond is a 9 billion parameter language model based on the zai-org/GLM-Z1-9B-0414 architecture, fine-tuned for advanced mathematical reasoning tasks. It utilizes the HETU (Hints Enable True Understanding) method, specifically trained with a compact conditioning signal (CotCond) instead of full chain-of-thought generation. This model excels in math reasoning benchmarks such as AIME, GSM8K, and MMLU, offering specialized performance for complex mathematical problem-solving.

Loading preview...

HETU-GLM-Z1-9B-MathReasoning-CotCond Overview

This model, developed by AdarshSingh7647, is a 9 billion parameter variant within the HETU (Hints Enable True Understanding) model suite, built upon the zai-org/GLM-Z1-9B-0414 base architecture. It is specifically designed and optimized for complex mathematical reasoning tasks.

Key Capabilities

  • Specialized Math Reasoning: The model is fine-tuned to perform exceptionally well on various math reasoning benchmarks, including AIME, GSM8K, MATH-500, Omni-MATH, GPQA-Diamond, and MMLU.
  • CotCond Training Method: It employs a unique training methodology called CotCond (Conditioning on compact signals), which is part of the HETU approach. This method uses a compact conditioning signal rather than a full generated chain-of-thought, potentially leading to more efficient reasoning.
  • Merged Model: This release represents the full merged model, combining the base weights with the LoRA adapter, and is provided in bf16 precision.

Good For

  • Mathematical Problem Solving: Ideal for applications requiring robust performance in solving intricate mathematical problems.
  • Research in Reasoning: Useful for researchers exploring alternative methods for enhancing model reasoning capabilities, particularly those interested in the CotCond approach detailed in the HETU paper.