Cloudflare Clairic: System One Extraction Model Launches

Cloudflare has introduced Clairic, a specialized System One extraction model now available on its Workers AI platform. The model delivers typed values with built-in confidence scores in a single forward pass, addressing a critical pain point for developers building structured data extraction workflows. Officially launched on October 20, 2026, Clairic combines extraction accuracy with an unconventional pricing model that charges $0.10 per million input tokens while eliminating output token costs entirely.

What Makes Clairic Different

Unlike traditional large language models that require careful prompt engineering and post-processing to extract structured data, Clairic operates as a System One model purpose-built for extraction tasks. The architecture returns strongly typed values (strings, numbers, booleans, dates) alongside confidence scores in one inference cycle. This design eliminates the need for validation layers or secondary parsing steps that typically add latency and complexity to extraction pipelines.

The model processes unstructured input data and maps it to predefined schemas without requiring extensive few-shot examples or chain-of-thought reasoning. Developers define their target data structure, and Clairic returns conforming JSON objects with per-field confidence metrics. This approach reduces the token overhead associated with verbose prompting strategies common in general-purpose models.

Pricing Structure and Workers AI Integration

Clairic's pricing model breaks from industry norms by charging exclusively for input tokens at $0.10 per million tokens processed. Output tokens, which typically represent 30-50% of extraction workload costs in competing services, carry no charge. For high-volume extraction scenarios processing millions of documents, this structure can deliver substantial cost savings compared to traditional API-based inference services.

The model runs natively on Cloudflare's Workers AI infrastructure, enabling edge deployment across Cloudflare's global network. Developers can invoke Clairic through standard Workers AI API calls with sub-50ms cold start times in most regions. The integration supports batch processing, streaming extraction, and real-time document parsing without requiring separate infrastructure management.

Technical Capabilities and Use Cases

Clairic handles common extraction scenarios including:

• Invoice and receipt data extraction with line-item parsing
• Resume and job posting structured field extraction
• Contract and legal document clause identification
• Product catalog normalization from unstructured listings
• Customer support ticket categorization and entity extraction

The confidence scoring mechanism allows developers to implement graduated trust models, routing high-confidence extractions to automated workflows while flagging low-confidence cases for human review. This capability proves particularly valuable in regulatory contexts where extraction accuracy directly impacts compliance.

Availability and Access

Clairic is available immediately to all Workers AI users through Cloudflare's dashboard and API interfaces. The model requires no waitlist or special provisioning, and developers can begin testing extraction workflows with existing Workers AI accounts. Cloudflare provides schema definition templates and extraction validation tools in the Workers AI documentation to accelerate integration.

Early adopters report that Clairic's single-pass architecture reduces total extraction pipeline complexity by eliminating retry logic and validation steps previously required with general-purpose models. The combination of edge deployment, simplified pricing, and native confidence scoring positions the model as a pragmatic alternative for production extraction workloads.

What This Means

Cloudflare's launch of Clairic signals a maturation of specialized AI models optimized for narrow, high-value tasks rather than general intelligence. By focusing on extraction accuracy, operational simplicity, and cost transparency, Clairic addresses practical friction points that have limited AI adoption in document processing and data normalization workflows. The zero-cost output token model may pressure other providers to reconsider pricing structures for task-specific inference services, particularly as enterprises demand predictable costs for high-volume AI operations.