FlagRelease/DeepSeek-R1-Distill-Qwen-1.5B-hygon-FlagOS

TEXT GENERATIONPricing:Input $0.04 / Cached $0.008 / Output $0.08Concurrent Unit Cost:1Model Size:1.5BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Jul 6, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

FlagRelease/DeepSeek-R1-Distill-Qwen-1.5B-hygon-FlagOS is a 1.5 billion parameter distilled language model based on the Qwen architecture, developed by FlagRelease. It is derived from the DeepSeek-R1 reasoning model and specifically optimized for deployment on Hygon hardware within the FlagOS ecosystem, featuring a 32768 token context length. This model is designed for efficient inference and integration into FlagOS-powered environments, offering strong reasoning capabilities for various tasks.

Loading preview...

Model Overview

FlagRelease/DeepSeek-R1-Distill-Qwen-1.5B-hygon-FlagOS is a 1.5 billion parameter language model distilled from the DeepSeek-R1 reasoning model, built upon the Qwen architecture. DeepSeek-R1 itself is an advanced reasoning model developed through large-scale reinforcement learning, achieving performance comparable to OpenAI o1 on mathematical, coding, and general reasoning tasks. This specific distilled version is tailored for efficient deployment and operation on Hygon hardware within the FlagOS software stack.

Key Capabilities & Features

  • Reasoning Focus: Inherits strong reasoning capabilities from the DeepSeek-R1 foundation model.
  • FlagOS Integration: Optimized for seamless deployment using FlagOS-Hygon container images, providing out-of-the-box inference scripts.
  • Hardware Acceleration: Designed to leverage the FlagOS stack for enhanced performance on Hygon AI accelerators.
  • Benchmark Performance: Demonstrates competitive performance on benchmarks like musr_generative, mmlu_pro, and gpqa_generative_cot within the FlagOS environment.
  • Unified Ecosystem: Part of the broader FlagRelease platform, which aims to unify the "model–system–chip" layers for diverse AI accelerators.

Ideal Use Cases

  • Edge/On-premise Deployment: Excellent for applications requiring efficient LLM inference on Hygon-based hardware.
  • Reasoning Tasks: Suitable for tasks demanding strong logical reasoning, mathematical problem-solving, and code understanding.
  • FlagOS Ecosystem Development: Perfect for developers and researchers working within the FlagOS framework who need a performant, integrated language model.
  • Benchmarking & Evaluation: Useful for evaluating model performance and the FlagOS stack on specific hardware configurations.

Finetunes
1 models