Fireworks AI has introduced FireRouter with Opus, the market's first cache-aware router specifically optimized for the Claude Opus series, available as a serverless endpoint in the form of an independent routing model. This product can evaluate the suitability of each model for the current task, estimate processing costs, and select the model that offers the best balance of quality and cost after weighing routing costs and benefits. Currently, the routing pool includes Claude Opus5.5, GLM5.3, and GLM5.3Flash. Internal A/B testing results for coding traffic show that its accuracy in coding tasks reaches 98.1% of that achieved by using Opus alone, with an overall cost reduction of 57%. The cache hit rate is only 3.6 percentage points lower than when using Opus alone, a trade-off made to divert simple tasks to low-cost models. Currently, the product has been integrated into multiple toolchains, and users can enable it through specified commands.