Fireworks AI Releases Ember-1: A Post-Training Model Based on Kimi K3, Reducing Reasoning Tokens by Approximately 40%
Fireworks AI Releases Ember-1: A Post-Trained Kimi K3 That Uses About 40% Fewer Tokens
AI BriefingFireworks AI introduced Ember-1, a specialized model post-trained on Kimi K3, whose weights were open-sourced by Moonshot AI. By optimizing its internal reasoning process, the model reduces generated reasoning tokens by approximately 40% while maintaining task accuracy. Unlike simply lowering the level of reasoning effort, Ember-1 preserves useful self-reflection while eliminating redundant loops. Benchmarks show that it outperforms K3 Max on Terminal Bench 2.1 and DeepSWE 1.1, while achieving significant cost savings in production A/B tests. It is currently available only as a research preview through the Fireworks Serverless API, with its weights not publicly released.
Also reported by 2 other sources
- Fireworks Research Releases Ember-1: Achieving the Same Quality as Kimi K3 with Fewer TokensHacker News: Top AI Posts · September 28
- Fireworks releases Ember-1, matching Kimi K3 quality with about 40% fewer tokensFireworks AI · September 24
Source: MarkTechPost · marktechpost.com
Event follow-up1
View all event coverage- Fireworks releases Ember-1, matching Kimi K3 quality with about 40% fewer tokensFireworks AI · September 24