signalHacker News Best2026-09-28
Ember-1
Fireworks Research releases Ember-1, a specialized model that matches Kimi K3's performance using 40% fewer tokens. Trained on Kimi K3, it shortens reasoning traces without losing accuracy, confirmed across seven benchmarks and two customers' production traffic. The model is available now, with training involving over 50 experiments and 200 evaluations.
- for who
- Developers running large-scale coding or agentic workloads with reasoning models
- why now
- Ember-1 launches today, cutting Kimi K3's token cost by 40% without quality loss.
- what changes
- They can now achieve the same quality at 40% lower token cost, making automated coding and agentic tasks more affordable.
- to do
- Adopt Ember-1 from Fireworks to reduce token expenses while preserving Kimi K3's reasoning quality.
key points
- Ember-1 cuts token usage by 40% while keeping Kimi K3's quality
- Built via 50 training experiments and 200 evaluations
- Validated on seven benchmarks and two customer A/B tests
#new model#cost optimization#token compression
score
score 8 out of 10. 0-10: how dense the facts are, multiplied by how much you can do with them after reading. 8+ means the topic's evidence bar is met: benchmarks and availability for a new model, amount and investors for a funding round, revenue figures for a solo-money story. Below 5 an item does not enter the digest. A press release scores 3 or less, a reprint loses 2, anything older than 14 days loses 1, a headline that misleads loses 3.
read the source