Fireworks launches FireRouter with Opus for coding harnesses
Fireworks says FireRouter with Opus is available as a standalone router model and through its CLI, including in Claude Code, Codex and Cursor. It reports lower cost and slightly lower accuracy than Opus-only in an internal coding test.
SessionWatcher editorial · Published · Updated · Source announcement: 2026-09-28
What FireRouter with Opus does
Fireworks says FireRouter with Opus is available in its CLI and as a standalone router model that Fireworks accounts can use as a serverless endpoint. It routes each user turn among Claude Opus 5.5, GLM 5.3 and GLM 5.3 Flash, weighing task suitability, estimated model cost and the potential cost of losing prompt-cache hits when switching models.
The post says the model set may change as new models and versions launch. Fireworks lists Claude Code, Codex and Cursor IDE among the supported harnesses. Its example command is `fireconnect claude --model firerouter/opus`.
Sources: Introducing FireRouter with Opus
Fireworks' internal cost and accuracy results
In an internal test of coding traffic, Fireworks randomly assigned sessions to FireRouter with Opus or an Opus-only control group across the same workloads and users. It reports average cost per session of $6.63 with the router versus $15.36 with Opus-only, a 57% reduction (±19 percentage points).
Fireworks reports that the router scored 78.7% of graded turns, compared with 80.2% for Opus-only, or 98.1% of Opus-only's overall accuracy. Cache hit rate was 94.2% with FireRouter and Opus versus 97.8% with Opus-only. These are results from the company's internal test, not an independent benchmark.
Sources: Introducing FireRouter with Opus
AI assisted reporting, checked against the linked official sources. Source pages checked 2026-09-30. Editorial process and corrections.