Explore additional models
Frequently
asked questions
Qwen3.6 35B A3B FFT is the full fine-tuning version of Qwen3.6 35B A3B. The base model is the mixture-of-experts model the Alibaba team released in April 2026 under the Apache 2.0 license. On Crusoe, Qwen3.6 35B A3B FFT is available for Serverless Fine-Tuning in Crusoe Intelligence Foundry.
Yes. Qwen3.6 35B A3B is a mixture-of-experts model with 35B total parameters and about 3B active parameters per token. Each token routes through 8 of its 256 experts plus 1 shared expert. Like other sparse MoE models, it runs fewer parameters per token than dense models of the same total size, which keeps inference cost per token down. It succeeds Qwen3.5 35B A3B, and its dense sibling is Qwen3.6 27B.
The Alibaba team focused Qwen3.6 35B A3B on agentic coding, particularly frontend workflows and repository-level reasoning. Teams can fine-tune it on their own coding tasks. Teams can also fine-tune it on domain-specific text tasks where you want the model to follow your formats, terminology, or style.
FFT stands for full fine-tuning. A full fine-tuning job trains on your dataset and updates every weight in Qwen3.6 35B A3B, including the attention layers and every expert. Teams typically choose full fine-tuning for larger datasets or broad changes, such as adapting the model to a new domain or a private codebase.


%204.png)
