Ember-1 is a specialised reasoning model from Fireworks Research, built on top of Kimi K3. Its design goal is narrow and concrete: make every token go further by producing shorter reasoning traces. A reasoning model bills its thinking tokens at the output rate, so a model that reaches the same answer with less deliberation costs less per task even at the same headline price.
It inherits Kimi K3's long context — 1,048,576 tokens in, up to 943,718 out — and accepts images alongside text.
Reasoning tokens bill at the output rate.
Shorter reasoning traces
The model is tuned to think less per answer rather than to think better. On a reasoning workload the thinking tokens are usually the bulk of the bill, so trace length is the lever that moves cost.
Kimi K3's context, unchanged
A 1,048,576-token window holds whole repositories or document sets in one request, with room for very long replies.
Agentic plumbing
Tool calling, structured output and streaming are all supported, alongside image input.
High-volume reasoning
Pipelines that run many reasoning calls, where the thinking-token bill decides whether the workload is viable.
Long-context analysis
Repository-scale code review and long-document research in a single request.
Agent loops
Multi-step work that plans, calls tools and checks itself, where every step pays for its own reasoning.
Kimi K3 is the base model, at $3.9 / 1M input and $19.5 / 1M output, with a 1M-token context and web search on top. Ember-1 is Fireworks Research's derivative at $4.1262 / $20.631: a slightly higher sticker price in exchange for reaching answers with fewer reasoning tokens. Which one is cheaper in practice depends on how much your workload makes the model think — worth measuring on your own traffic rather than assuming.
import requests r = requests.post( "https://api.aimlapi.com/v1/chat/completions", headers={"Authorization": "Bearer " + AIMLAPI_KEY}, json={ "model": "fireworks/ember-1", "messages": [ { "role": "user", "content": "Hello!" } ] }, ) print(r.json())
const r = await fetch("https://api.aimlapi.com/v1/chat/completions", { method: "POST", headers: { Authorization: `Bearer ${process.env.AIMLAPI_KEY}`, "Content-Type": "application/json", }, body: JSON.stringify({ "model": "fireworks/ember-1", "messages": [ { "role": "user", "content": "Hello!" } ] }), }); console.log(await r.json());
curl -X POST https://api.aimlapi.com/v1/chat/completions \ -H "Authorization: Bearer $AIMLAPI_KEY" \ -H "Content-Type: application/json" \ -d '{"model":"fireworks/ember-1","messages":[{"role":"user","content":"Hello!"}]}'
OpenAI-compatible — swap the base URL and it works with your existing SDK.
| Type | Price |
|---|---|
| Input | |
| Output | |
| Cached input |
Reasoning tokens are billed at the output rate — which is the point of the model: shorter traces mean fewer of them per answer.
Create an Account
Sign up at aimlapi.com
Generate an API Key
Create it in your dashboard and make sure it's active.
Set the Base URL
Use https://api.aimlapi.com/v1
Make an API Call
Send chat messages to /chat/completions with fireworks/ember-1 as the model id.
| Model | Input | Output | Context | Best for |
|---|---|---|---|---|
Ember-1 This page | Reasoning with fewer thinking tokens | |||
| Long-context, multimodal & agentic workflows | ||||
| Frontier reasoning + agents | ||||
| Balanced coding + agents | ||||
| Reasoning + agents |
Ember-1 is a reasoning model from Fireworks Research, built on Moonshot AI's Kimi K3. It is tuned to produce shorter reasoning traces — to reach an answer while spending fewer thinking tokens on the way there.
On AI/ML API it costs $4.1262 per 1M input tokens and $20.631 per 1M output tokens, with cached input at $0.41262 per 1M. Reasoning tokens are billed at the output rate.
Kimi K3 is the base model Ember-1 is built on, at $3.9 / $19.5 per 1M tokens. Ember-1 lists slightly higher at $4.1262 / $20.631 but is tuned to use fewer reasoning tokens per answer, so which one is cheaper depends on how much your workload makes the model think. Kimi K3 also offers web search, which Ember-1 does not.
1,048,576 tokens in and up to 943,718 tokens out, inherited from Kimi K3 — enough for repository-scale code and very long replies in a single request.
Send a POST request to https://api.aimlapi.com/v1/chat/completions with fireworks/ember-1 as the model id and your AI/ML API key in the Authorization header. The endpoint is OpenAI-compatible, so an existing chat client only needs the base URL swapped.
