AI Agent StoreFireworks · Provider documented router
FireRouter with Opus
A coding router that chooses Claude Opus or a Fireworks open model for each user turn. Consider it when routine coding work could run on a cheaper model, then compare completed tasks and total session bills.
Announced 2026-09-28 · First reviewed 2026-10-10 · Checked 2026-10-15.
We still need comparable independent tests. You can compare the confirmed features and limits below, but we cannot yet give this version an Agent Fit or Practical Value rating. Test it on your own tasks before switching; results for another version or thinking setting may not apply.
What you can use today
Token prices measure the small pieces of text the model reads or produces. Capacity is the amount the service accepts in a request; check the limits below for any fallback to another model. A dash means the rate is unverified.
Exact provider API ID: firerouter/opus. This route chooses Claude Opus or a Fireworks open model for each turn. Opus currently means Claude Opus 5.5; the open mix includes GLM 5.3 and GLM 5.3 Flash. Read the response model field to see which answered. The simpler firerouter ID can choose Opus or GPT Sol depending on credentials and client; it is not the same pinned selection.
Use a Fireworks key and an Anthropic credential for Opus. Without the Anthropic credential, open models serve the turns. Routing preference 1 selects the closed model; other settings can use cheaper open models. The docs exclude accounts with data residency and Microsoft Foundry. A router is not a separately benchmarked base model.
The model page rounds context to 1000K. A guaranteed exact common limit, output cap and reasoning-effort mapping are unverified.
There is no single verified token rate for this route. Open turns bill at Fireworks rates; closed turns bill to the supplied provider account, with cache reads at that serving model’s rate. Vendor session savings are not an independent benchmark or a guaranteed saving. Your bill can also include tools, retries and charged thinking tokens. Check the offer used by your application.
Provider release and specifications ↗
Check availability and plans ↗
More setup and licensing details ↗
Test it before switching
Give each model the same examples, tools and definition of success. Record correct outcomes, response time, retries and total cost. A cheaper token price does not guarantee a cheaper completed task.
Compare caching and long-context costs ↗