OrionLLM/GRM-2.6-Plus-0628

Hugging Face
VISIONConcurrent Unit Cost:2Model Size:27BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Jul 7, 2026License:apache-2.0Architecture:Transformer0.0K Open Weights Featherless Exclusive Warm

OrionLLM/GRM-2.6-Plus-0628 is a 27B-parameter reasoning model developed by OrionLLM, built on the Qwen3.6 architecture with a 32768 token context length. This model is optimized for difficult, high-complexity tasks, particularly excelling in long-horizon agentic workflows and structured reasoning. It delivers elite-level reasoning and strong performance for its size, making it suitable for advanced problem-solving, coding, and agentic applications.

Loading preview...

Overview

OrionLLM/GRM-2.6-Plus-0628 is a 27B-parameter reasoning model developed by OrionLLM, serving as an update to GRM-2.6-Plus. Built on the Qwen3.6 architecture, this model is designed for general-purpose AI with a strong focus on difficult, high-complexity tasks and long-horizon agentic workflows. It aims to provide elite-level reasoning and problem-solving capabilities while remaining practical and efficient for advanced local and research-oriented use.

Key Capabilities

  • Elite-Level Reasoning: Optimized for demanding reasoning workloads, providing clarity, consistency, and strong step-by-step problem-solving.
  • Improved Agentic Performance: Specifically targets long-horizon agentic workflows, maintaining coherence and effectiveness across multi-step tasks.
  • High Performance for Size: Delivers excellent capability relative to its 27B parameter scale, balancing intelligence with practical deployment.
  • Advanced Coding and Agentic Use: Well-suited for code generation, structured problem-solving, tool-style workflows, and local agentic applications.

Performance Highlights

GRM-2.6-Plus-0628 demonstrates strong performance across various benchmarks, often outperforming its predecessor GRM-2.6-Plus and other models in its class, including Qwen3.6-27B and google/gemma-4-31B-it. Notable scores include:

  • MMLU-Pro: 88.1
  • MMLU-Redux: 96.4
  • C-Eval: 92.4
  • GPQA Diamond: 90.1
  • LiveCodeBench v6: 86.5
  • SWE-bench Verified: 79.7

When to Use This Model

This model is ideal for users requiring a capable AI for advanced problem-solving, complex coding tasks, and agentic applications that demand structured reasoning and long-horizon planning. Its balance of strong intelligence and practical deployment makes it suitable for capable consumer and workstation hardware.

Finetunes
1 models