Ember-1

Fireworks Research unveiled Ember-1, a specialized AI model that delivers the same quality as Kimi K3 while using 40% fewer tokens through more efficient reasoning. Developed using over 50 training experiments, Ember-1 was designed to address the high costs of reasoning models in coding applications and multi-turn agent workloads, where lengthy internal reasoning traces accumulate exponentially. The model has been validated across external benchmarks, live A/B tests with customers, and internal developer workloads with no quality loss.