Qwen (Alibaba Group) released Qwen3‑Coder 30B A3B Instruct, a Mixture‑of‑Experts coding model. It has 30.5 billion total parameters but only 3.3 billion are active per token. Fireworks now hosts it for on‑demand GPU deployments, offering the full 262,144‑token context window and up to 65,536‑token outputs. The model can be fine‑tuned with LoRA through Fireworks’ Reserved Fine‑Tuning service. It is under an Apache 2.0 license, so commercial use is allowed.
Why it matters
Developers can run a large‑scale code‑assistant with very long context on Fireworks without rate limits, enabling more complex coding tasks.
