If you use Claude Code, there’s a small but meaningful change you should know about. As of May 14, 2026, Fast Mode now defaults to Opus 4.7 Anthropic’s most capable publicly available model. Previously, it defaulted to Opus 4.6.
The update shipped with Claude Code v2.1.142, and it was announced directly on the @ClaudeDevs Twitter account with a simple message: “Fast mode now defaults to Opus 4.7 in Claude Code. Try it out today with /fast.”
WHAT IS FAST MODE?
Fast Mode is a high-speed configuration for Claude Opus inside Claude Code. It’s not a different model it’s the exact same Opus with a different infrastructure setup that prioritizes response speed over cost efficiency.
The result: up to 2.5x higher output tokens per second compared to standard Opus. The trade-off: it costs 6x more per token.
To be clear you’re not getting a stripped-down or simplified Claude. Same reasoning, same capabilities, same full 1 million-token context window. You’re just getting it faster because Anthropic routes your requests through a higher-priority serving path.
HOW TO USE IT
Activating Fast Mode is simple. Just type /fast in your Claude Code session. You’ll see a lightning bolt icon next to your prompt confirming it’s on. To toggle it off, type /fast again.
Fast Mode works in:
– Claude Code CLI (terminal)
– Claude Code VS Code Extension
Requirements: Claude Code v2.1.36 or later. Check your version with: claude –version
WHAT CHANGED ON MAY 14?
Before this update, typing /fast would run Fast Mode on Opus 4.6. Now it runs on Opus 4.7 by default.
If you want to stay on Opus 4.6, you can override this by setting the environment variable: CLAUDE_CODE_OPUS_4_6_FAST_MODE_OVERRIDE=1
This variable takes precedence, so your Fast Mode sessions will continue using Opus 4.6 regardless of the new default.
WHAT’S THE COST?
Fast Mode pricing runs at 6x standard Opus rates. Standard Opus 4.7 costs $5 per million input tokens and $25 per million output tokens. Fast Mode on Opus 4.7 runs at $30 per million input and $150 per million output.
One important detail: if you switch Fast Mode on mid-conversation, you pay the full Fast Mode input price for your entire existing conversation context not just from that point forward. This makes it significantly more expensive to toggle Fast Mode halfway through a session. The most cost-efficient approach is to enable it at the start.
Also worth knowing: Fast Mode tokens don’t count toward your plan’s included usage. Every Fast Mode token goes directly to extra usage billing.
IS IT WORTH IT?
That depends entirely on what you’re doing. Fast Mode is a precision tool, not a blanket upgrade.
It makes sense when:
– You’re iterating quickly and latency is breaking your flow
– You’re debugging live and each response pause costs you concentration
– You’re on a tight deadline and speed has real dollar value
– You’re doing interactive pair programming where every pause disrupts your rhythm
It doesn’t make sense when:
– You’re running background tasks or batch jobs where you’re not watching in real time
– You’re working on long, complex tasks where cost matters more than wait time
– The work can be done at standard speed without affecting your output
In a typical debugging session with 25 back-and-forth exchanges, Fast Mode saves roughly 15-20 minutes compared to standard Opus. At 6x the token cost, that’s a math problem every developer has to solve for themselves.
ABOUT OPUS 4.7
For context on why defaulting to Opus 4.7 is a meaningful upgrade: Opus 4.7 is Anthropic’s current flagship model. It scores 64.3% on SWE-bench Pro ahead of GPT-5.4 at 57.7% and Gemini 3.1 Pro at 54.2%. It also supports a 1 million-token context window, high-resolution image processing (up to 3.75 megapixels), and a new xhigh effort level designed specifically for agentic coding tasks.
Fast Mode on Opus 4.7 is currently in research preview. Pricing, availability, and behavior may change as Anthropic collects feedback.
BOTTOM LINE
Fast Mode now runs Anthropic’s best publicly available model by default. If you’re already using /fast in Claude Code, you just got an automatic upgrade to Opus 4.7 faster responses, same intelligence, higher cost. If you’ve never tried Fast Mode, now’s a good time to experiment with it on a session where speed genuinely matters
