Fireworks Research Launches Ember-1 Model with 40% Lower Token Usage
September 24, 2026
Ember-1, a specialized model built on the Kimi K3 architecture, is now available in Research Preview on Fireworks Serverless. The model achieves comparable quality to larger counterparts while using 40% fewer tokens.
HOW THIS AFFECTS YOU
●
builderYou can reduce inference costs and latency by switching to this more efficient model architecture.