Anthropic has unveiled a new high-velocity research preview for its premier model, Claude Opus 4.6, introducing a "fast mode" designed to drastically reduce response times for demanding workflows. Accessible via the Claude Code interface and specialized API endpoints, this mode reportedly operates 2.5 times faster than standard versions, catering to developers and power users who prioritize immediate execution over cost efficiency.

The performance boost carries a substantial premium, with pricing set at six times the standard rate—climbing to $30 per million input tokens and $150 per million output tokens. To encourage early adoption, a 50% discount is available through mid-February, though the long-term cost remains significant for high-volume users. This tiered pricing model reflects the immense computational power required to maintain such high throughput for a top-tier Large Language Model (LLM).

In addition to speed, the update expands the context window—the total amount of text data the model can process in a single session (context window)—from 200,000 to 1,000,000 tokens. While this massive expansion allows for the analysis of entire codebases or dense technical manuals, it introduces additional surcharges once the initial token threshold is exceeded. This release signals a strategic shift toward providing specialized, high-performance tiers for professional AI integration.

Anthropic has unveiled a new high-velocity research preview for its premier model, Claude Opus 4.6, introducing a "fast mode" designed to drastically reduce response times for demanding workflows. Accessible via the Claude Code interface and specialized API endpoints, this mode reportedly operates 2.5 times faster than standard versions, catering to developers and power users who prioritize immediate execution over cost efficiency.

The performance boost carries a substantial premium, with pricing set at six times the standard rate—climbing to $30 per million input tokens and $150 per million output tokens. To encourage early adoption, a 50% discount is available through mid-February, though the long-term cost remains significant for high-volume users. This tiered pricing model reflects the immense computational power required to maintain such high throughput for a top-tier Large Language Model (LLM).

In addition to speed, the update expands the context window—the total amount of text data the model can process in a single session (context window)—from 200,000 to 1,000,000 tokens. While this massive expansion allows for the analysis of entire codebases or dense technical manuals, it introduces additional surcharges once the initial token threshold is exceeded. This release signals a strategic shift toward providing specialized, high-performance tiers for professional AI integration.