Fireworks: Ember-1
ember-1- Input / Output
- $3.00 / $15.00
- Context
- 1M
- Modalities
- Chat
- Capabilities
- Tool usePrompt CacheReasoning
About Ember-1
Ember-1 is a reasoning model from Fireworks Research, built on Moonshot's Kimi K3 for coding and agentic work. It was trained to drop the reasoning that Kimi K3 spends without improving answers, so Fireworks reports roughly 35 to 40 percent fewer tokens at comparable quality. It calls tools and handles a million-token context, and Fireworks released it as a research preview.
Where it works well
- Fireworks reports about 40 percent fewer generated tokens than Kimi K3 at comparable quality, which shortens long agent runs.
- It was trained on math, coding, instruction following, conversation, search and tool use, rather than on one narrow task.
- Fireworks evaluated it on SWE-bench Verified, Terminal Bench 2.1, DeepSWE 1.1 and tau-2 Bench Airline, all agent and coding measures.
- Tool calling and implicit prompt caching suit agents that repeat a large system prompt on every step.
When to choose another model
- It is a research preview, and Fireworks ties continued access to community demand.
- The evidence for its token savings and quality comes from Fireworks' own testing rather than an independent benchmark.
- It takes text only, so screenshot-driven work needs a vision model.
Getting started
Create API key
Create a key in Console, then use it with every model on the platform.
Send your first request
Copy the example for your language and run it against the endpoint.
Text to textChat Completions APIPOST/v1/chat/completionscurl https://api.tokenlab.sh/v1/chat/completions \ -H "Content-Type: application/json" \ -H "Authorization: Bearer sk-xxx" \ -d '{ "model": "ember-1", "messages": [ {"role": "user", "content": "Hello!"} ] }'
Pricing
The Verified price applies to Verified, which costs less on most models. The Official price is the model maker's published price and applies to the more reliable Official route. Auto bills the route that completes the request.
Default spec
per 1M tokens- Official price
- Input $3.00 / Output $15.00 / Cache read $0.30
- Verified price
- —
- Discount
- —
Cache read
- Official price
- $0.30
- Verified price
- —
- Discount
- —
| Official priceper 1M tokens | Verified priceper 1M tokens | Discount | |
|---|---|---|---|
| Default spec | Input $3.00 / Output $15.00 / Cache read $0.30 | — | — |
| Prompt cache pricing | |||
| Cache read | $0.30 | — | — |
Usage & activity
Success rate is the share of requests that completed. Latency is how long a full response takes; P95 means 95% of requests finished within that time.
Usage & availability
Last 24 hours- Requests
- Success rate
- P95 latency
- Total tokens
Data is based on aggregate user requests, excluding status checks.
Open in Console
Open Ember-1 in Console with a prompt ready to edit or send.
Help me try ember-1 with a short message at /v1/chat/completions. Show the reply, latency, and cost.
Use cases
Best for- Reasoning
- Code
Coding agents that think a lot
Run terminal or repository agents where Kimi K3's reasoning turns are long and you want the same answers with fewer generated tokens.
Tool-heavy search workflows
Chain web or database lookups and let the model decide each next call while keeping the full history in the long context.
Reasoning cost experiments
Compare it with Kimi K3 on your own task set to see whether the shorter reasoning holds quality for your workload.
Large-codebase review
Load many files at once and ask for a design critique or a root-cause analysis of a cross-module failure.
Prompt examples
The repo tree and the failing CI log are attached. Use the shell tool to find the cause, patch it, and rerun only the affected tests.
Search the changelog and issue tracker for every change to the retry policy since v2, then summarize which ones could explain the new timeouts.
Review these twelve files for race conditions in the job scheduler and rank the findings by how likely they are to occur in production.
This model has conditional pricing. Monthly token totals alone cannot produce a reliable estimate; use the detailed pricing for the request specification, cache, and applicable time window.
FAQ
What exactly is Ember-1 from Fireworks?
A research-preview reasoning model from Fireworks Research, built on Kimi K3. Fireworks trained it to keep the reasoning that helps recover from mistakes and remove loops that add tokens without improving results, aiming at coding and agentic workloads.
How is Ember-1 different from Kimi K3?
It targets the same quality with less thinking. Fireworks says Kimi K3 spends most of its generated tokens on reasoning, sometimes over 90 percent, while Ember-1 uses about 35 to 40 percent fewer tokens per task. Quality claims come from Fireworks' evaluations.
Is Ember-1 a stable production model?
Not yet. Fireworks launched it as a research preview with time-limited free serverless access, to be made permanent depending on demand. Applications should keep a second model available in case the preview ends.
Does Ember-1 support tool calling?
Yes. Fireworks describes function calling and implicit prompt caching. Define your functions in the request as you would for any OpenAI-style chat model, and keep the system prompt stable so caching can help.
Does Ember-1 accept images?
No. It takes text only and answers in text. For screenshots or diagrams, use a vision model, and keep Ember-1 for the coding and agent steps that follow.
How much does Ember-1 cost?
On TokenLab, Ember-1 costs Input $3.00 / Output $15.00 per 1M tokens. The pricing table above shows the full breakdown. Rates depend on the billing unit, specification, and usage. Compare matching conditions in the model's detailed pricing; a single rate does not determine the total cost.
What are the context window and output limit of Ember-1?
Ember-1 accepts up to 1,048,576 tokens of context.
Which endpoint should Ember-1 use?
Use https://api.tokenlab.sh/v1/chat/completions for Ember-1. The request example below shows the matching code shape.
Which operations does Ember-1 support?
Ember-1 supports Text to text. Select an operation above to see its endpoint and request example.
Compare Ember-1
Sources
Reviewed Oct 2, 2026
