Vercel added Ember-1 from Fireworks to AI Gateway. Ember-1 is a research-preview reasoning model for coding and agent workflows. It uses about 40% fewer tokens than Kimi K3 at similar quality. It supports a 1M-token context window, text and image input, tool calling and implicit prompt caching. The Fireworks endpoint offers Zero Data Retention and No Prompt Training. The model is available for an initial two-week research preview. Users can call it via `fireworks/ember-1` in APIs and coding agents. Vercel CLI can configure coding agents to use it. AI Gateway provides unified model access with usage tracking and budgets.
What changed
Ember-1 was not available on Vercel AI Gateway.
What it unlocks
Using a research-preview reasoning model for coding and agent workflows with reduced token usage.
- 40% fewer generated tokens
- 1M-token context window
- initial two-week research preview window
What you need to act on it
- access to Vercel AI Gateway
- research preview availability
Sources