# Fireworks: Ember-1

Provider: fireworks  
Category: llm  
Model ID: `fireworks/ember-1`

Ember-1 is a specialized reasoning model from Fireworks Research, built on [Kimi K3](https://openrouter.ai/moonshotai/kimi-k3). It is designed to make every token go further: it produces shorter reasoning traces, using roughly 40%...

## Specs

- Context length: 1048576 tokens
- Max output: 943718 tokens
- Modalities: text, image
- Released: 2026-09-24

## Pricing

- Input: $3.00 per million tokens
- Output: $15.00 per million tokens

## Providers

- **fireworks** — ctx 1048576, input $3.00/M, output $15.00/M

## Verdict

Ember-1 is Fireworks' first proprietary model, designed for high-throughput production workloads where cost and speed matter more than frontier reasoning. With a 1M token context window and aggressive pricing ($3 input, $15 output per Mtok), it targets teams running high-volume inference on structured tasks like classification, extraction, and routing. Vision support adds utility for document workflows. Trade-off: no public benchmarks yet, so you're betting on Fireworks' infrastructure reputation without independent performance data. Best for teams already on Fireworks who need a fast, cheap workhorse for repetitive tasks and can validate quality in-house.

---
Last verified: 2026-10-19T02:30:34.239Z  
Canonical URL: https://switchy.build/directory/models/ember-1