Anthropic's Claude Code introduces fast mode for Opus 4.6, delivering quicker responses through API optimization at higher token costs, with a 50% discount running through mid-February.
Anthropic has launched fast mode for Claude Code, a research preview feature that accelerates Opus 4.6 responses by optimizing API configuration rather than using a different model. The feature trades cost efficiency for speed, charging $30 per 150 million input tokens and $150 per 150 million output tokens for conversations under 200K tokens.
How Fast Mode Works
Fast mode isn't a separate model but an API configuration that prioritizes response speed over cost efficiency. Users get identical Opus 4.6 quality and capabilities but with reduced latency. The feature can be toggled on with /fast in the Claude Code CLI or VS Code Extension, or by setting "fastMode": true in user settings.
Once enabled, fast mode persists across sessions. If you're on a different model when activating it, Claude Code automatically switches to Opus 4.6. The model doesn't revert when you disable fast mode—you'll stay on Opus 4.6 and need to use /model to switch elsewhere.
Cost Considerations
The pricing structure creates a significant premium over standard Opus 4.6. Fast mode charges apply from the first token, even if you switch mid-conversation, meaning you'll pay the full fast mode uncached input token price for the entire conversation context. This makes it more cost-effective to enable fast mode at the start of a session rather than toggling it mid-conversation.
A 50% discount runs through 11:59pm PT on February 16 for all subscription plans (Pro/Max/Team/Enterprise) and Claude Console users. However, fast mode usage is billed separately from subscription rate limits and counts against extra usage only.
When to Use Fast Mode
Fast mode shines for interactive work where response latency matters more than cost:
- Rapid iteration on code changes
- Live debugging sessions
- Time-sensitive work with tight deadlines
Standard mode remains better for long autonomous tasks, batch processing, CI/CD pipelines, and cost-sensitive workloads.
Users can combine fast mode with lower effort levels for maximum speed on straightforward tasks, though this may reduce quality on complex problems.
Requirements and Limitations
Several restrictions apply:
- Not available on third-party cloud providers (Amazon Bedrock, Google Vertex AI, Microsoft Azure Foundry)
- Requires extra usage enabled on your account
- For Teams and Enterprise, admins must explicitly enable the feature
- Separate rate limits from standard Opus 4.6
When hitting fast mode rate limits or running out of extra usage credits, the system automatically falls back to standard Opus 4.6 with a grayed-out indicator showing cooldown status.
Research Preview Status
As a research preview, fast mode may change based on user feedback, with availability and pricing subject to modification. The underlying API configuration could evolve as Anthropic gathers usage data and refines the feature.
For teams, admins can enable fast mode through Console preferences for API customers or Admin Settings > Claude Code for Teams and Enterprise organizations.
The introduction of fast mode reflects growing demand for interactive AI experiences where response speed becomes critical for developer workflows, even at premium pricing. Whether the cost premium justifies the latency reduction will likely depend on specific use cases and organizational priorities.
Comments
Please log in or register to join the discussion