# Cline says Ember-1 matches Kimi K3 performance with fewer tokens

Cline says Fireworks Research’s Ember-1 is built on Kimi K3 and uses about 40% fewer tokens at the same benchmark performance. In a live coding A/B test, it used 71% fewer reasoning tokens and 39% fewer total tokens than K3 at the same success rate.

Language: en
Time zone: America/Los_Angeles
Current time (UTC, cached as of): 2026-11-17T15:13:00.000Z
Query date (America/Los_Angeles): 2026-11-17

HTML: https://didcodexreset.com/news/546f6c56e89143bd8f525c25.html

Content language: en
Localization state: sameLanguage

Source: Cline · Published 9/27/2026, 15:02:32

Cline says Ember-1, from the Fireworks Research team, is built on Kimi K3 and uses about 40% fewer tokens while matching K3’s benchmark performance. The reduction comes from post-training K3 to repeat less of its reasoning.

Tags: Ember-1, Fireworks, Kimi K3, reasoning models

[View original post](https://x.com/cline/status/2104329978146115998)
