AI Gateway Adds Unified Fast Mode Support

AI Gateway has introduced a Unified Fast Mode abstraction in beta that lets users request all models in the same way. When users use the speed: 'fast' setting, the gateway preferentially serves the model's Fast tier if available, and automatically falls back to standard speed if it isn't.

The key features of Fast Mode are as follows:

• Performance optimization: While the cost per token is somewhat higher, latency can be lowered or throughput increased.
• Flexible implementation: A single setting lets you take advantage of the available fast path without needing to pin a specific provider or connect directly to one.
• Two request methods: You can increase speed while keeping the base model ID via the providerOptions.gateway.speed setting, or directly specify the Fast Slug of a fast variant model, such as anthropic/claude-opus-5-fast.

Currently, the number of models supporting Fast Mode continues to grow, and supported models can be identified by a lightning bolt icon in the model list. In addition, coding agents such as Claude Code can also take advantage of this feature immediately via the /fast command or model configuration.