Join waitlist
Back to products
Qwen3.5

Qwen3.5

The 397B native multimodal agent with 17B active params

Website github.com
Overview

What it is

Qwen3 is the large language model series developed by Qwen team, Alibaba Cloud. - QwenLM/Qwen3

Intent

I need it when

Reduce inference costs while maintaining high performance for production workloads

Qwen3.5 provides multiple model sizes (0.6B to 235B) and MoE variants, enabling users to select the optimal model for their performance-cost tradeoff. Smaller models like 4B and 8B offer efficient inference while maintaining strong capabilities.

Process long documents and extended conversations without context limitations

Qwen3.5 supports 256K-token context windows extendable to 1 million tokens, enabling users to handle ultra-long inputs for document analysis, multi-turn dialogues, and complex reasoning tasks that require extensive context retention.

Build and deploy advanced reasoning AI applications with state-of-the-art performance

Qwen3.5 offers both thinking and non-thinking modes with improved reasoning capabilities for mathematics, coding, and logical tasks. Users can download open-weight models (0.6B to 235B parameters) and deploy them on their own infrastructure for full control and customization.

Implement agent-based systems with precise tool integration and autonomous task execution

Qwen3.5 demonstrates leading performance in agent capabilities with precise external tool integration in both thinking and non-thinking modes, enabling developers to build complex autonomous systems and multi-step workflows.

Integrate multilingual AI capabilities into applications across diverse markets

Qwen3.5 supports 100+ languages and dialects with strong multilingual instruction following and translation capabilities, allowing developers to build globally accessible applications without language barriers.

Drop

Not a fit when

  • User requires a managed API service with commercial support and SLA guarantees
  • User lacks GPU infrastructure or technical expertise to deploy and run LLMs locally
  • User needs real-time inference at scale without managing their own deployment infrastructure
  • User requires proprietary model weights with restricted commercial licensing
  • User cannot accommodate the computational requirements of 235B or 30B parameter models on available hardware
Commercials

Pricing

Open-source model available for free download and use