Vercel Developers
@vercel_dev
Fast mode for Opus 4.7 is live on AI Gateway. Try it via API: ๐๐๐๐๐: '๐๐๐๐' https://t.co/i8v3oiLPph
38likes
View on XVercel Launches Fast Mode for Opus 4.7 to Accelerate Agentic Workflows
Vercel, the frontend cloud platform and creator of the AI SDK, launched Fast mode for Claude Opus 4.7 on its AI Gateway. This research preview enables a high-speed inference (running a model to generate outputs) tier that delivers approximately 2.5x faster output token generation than the standard model.
- Output speed increase
- ~2.5x faster
- Pricing (input)
- $30 per 1M tokens
- Pricing (output)
- $150 per 1M tokens
- Availability
- Research preview on AI Gateway
- Integration method
- AI SDK and Claude Code
This update mirrors OpenRouter's high-speed inference tier by addressing the latency bottleneck in high-reasoning models. While Opus 4.7 provides deep intelligence, its standard speed can disrupt autonomous tasks. By offering a speed-for-spend tradeoff, Vercel ensures its infrastructure remains competitive for Windsurf's Opus 4.7 integration.
To enable the feature, pass the speed: 'fast' parameter within the anthropic provider options of the AI SDK. It is also compatible with Claude Code via environment variables. The release follows Vercel's AI Gateway production index report. Fast mode is priced at a 6x premium, costing $30 per million input tokens and $150 per million output tokens.
Still wondering? A few quick answers below.
Fast mode is a high-speed inference tier for Anthropic's flagship Claude Opus 4.7 model, currently available in research preview on Vercel's AI Gateway. It is designed to accelerate output token generation for complex tasks without sacrificing the model's underlying intelligence or reasoning capabilities, making it ideal for latency-sensitive agentic workflows.
Fast mode is priced at a six-times premium compared to standard Opus 4.7 rates. On the Vercel AI Gateway, this results in a cost of thirty dollars per million input tokens and one hundred and fifty dollars per million output tokens. Standard pricing multipliers, such as those for prompt caching, still apply on top of these rates.
To use this feature within the Vercel AI SDK, developers must specify the speed parameter as fast within the provider options for the Anthropic model. Specifically, when calling the model anthropic/claude-opus-4.7, you include the speed: fast configuration to trigger the high-speed inference tier instead of the standard, slower generation mode.
Yes, Fast mode is compatible with Claude Code when accessed via the Vercel AI Gateway. To enable it, users must set two specific environment variables in their shell configuration or settings file: CLAUDE_CODE_ENABLE_OPUS_4_7_FAST_MODE and CLAUDE_CODE_SKIP_FAST_MODE_ORG_CHECK. This allows the terminal-based agent to utilize the accelerated output speeds for autonomous coding tasks.
Fast mode delivers approximately two and a half times faster output token generation than the standard Claude Opus 4.7 tier. This performance boost targets the latency bottleneck often found in long-running asynchronous tasks and agentic loops, allowing the model to complete multi-step reasoning and code generation significantly quicker than the default inference speed.
