Ember-1 from Fireworks now available on AI Gateway
By Steven Van ·
Fireworks' Ember-1 reasoning model joins Vercel AI Gateway as a two-week research preview with a 1M-token context window.
Fireworks' Ember-1 reasoning model is now available through Vercel AI Gateway. Built on Kimi K3 for coding and agentic workflows, Fireworks reports it generates roughly 40% fewer tokens than Kimi K3 at comparable quality, which can lower output costs and reduce the context carried into later steps for coding agents that make repeated model calls.
Ember-1 supports a 1M-token context window, text and image input, tool calling, and implicit prompt caching. The Fireworks endpoint supports Zero Data Retention and No Prompt Training. It's available as a research preview for an initial two-week window, priced at $3 per million input tokens and $15 per million output tokens.
To use it, run vercel ai-gateway setup with the Vercel CLI, which detects installed coding agents and configures their connection to the gateway, then select fireworks/ember-1 in the agent's model configuration. It can also be called via the AI SDK, OpenAI Chat Completions and Responses APIs, or the Anthropic Messages API using the model name fireworks/ember-1. The announcement also notes it can be tried in the model playground.