Cloudflare Updates Clairic and Introduces Clairic-flash for Advanced Data Extraction

Cloudflare has unveiled updated Clairic and a new Clairic-flash model, now open-weight. These AI models enhance data extraction from complex documents,…

Cloudflare's Workers AI Superintelligence team announced on November 18, 2026, the release of an updated Clairic model and a new, lighter version called Clairic-flash. Both models are now open-weight and focus on robust data extraction from complex, unstructured documents, a critical capability for many modern applications. This development underscores Cloudflare's ongoing efforts to push the frontier of AI research, providing advanced tools for its customers and the open-source community.

Significant Enhancements in Extraction Technology

The updated Clairic model, a 27-billion-parameter solution, features a redesigned architecture since its preview. It now employs parallel slot autoregressive generation for text and numerical fields, which has demonstrably improved extraction accuracy from both text and image inputs. A key advancement is its enhanced capability to process images and PDFs, allowing for precise data extraction, such as line items from multi-page documents. Specialized processing heads have also been integrated to handle complex data types like estimations, forecasts, and calculated dates, expanding its utility. Cloudflare's internal evaluations suggest a substantial reduction in extraction errors, particularly for image-based data.

Alongside the updated Clairic, the introduction of Clairic-flash provides a more compact, 9-billion-parameter alternative. This model is engineered for speed and efficiency, offering strong extraction performance in scenarios where a smaller footprint and faster processing are priorities. Both models maintain essential API features, including confidence scores and character spans, to provide comprehensive context with extracted data. Their availability as open-weight models on Hugging Face facilitates broader experimentation and customization by developers and researchers.

Practical Implications for Developers

These new models offer substantial practical benefits for development teams grappling with unstructured data. Use cases such as automating the extraction of key information from receipts, invoices, customer support tickets, or legal documents can see immediate improvements. The enhanced support for PDF documents is particularly valuable for organizations that process large volumes of multi-page forms and reports. The reported sub-second response times for both Clairic and Clairic-flash mark a significant performance advantage over general-purpose large language models for similar extraction tasks, making them suitable for high-throughput environments.

Developers now have the flexibility to choose between the higher accuracy of the 27-billion-parameter Clairic model and the speed and efficiency of the 9-billion-parameter Clairic-flash. While both models charge only for input tokens, providing a predictable cost structure, a temporary promotional offer is available until December 1, 2026, allowing users to experience the more powerful Clairic model at a rate aligned with the faster Clairic-flash.

Considerations for Adoption

While Cloudflare has provided internal evaluation results comparing Clairic against other emerging extraction models, it is important to acknowledge that these benchmarks are self-reported. The absence of standardized public benchmarks for extraction models means that developers should conduct their own evaluations tailored to their specific datasets and use cases. However, the open-weight release of both models on Hugging Face significantly lowers the barrier to entry for testing, integration, and potential fine-tuning, allowing organizations to thoroughly assess their suitability. This open approach fosters community contributions and further innovation in the field of specialized AI.

Conclusion: The updated Clairic and the new Clairic-flash represent a notable advancement in specialized AI for data extraction. Developers seeking to enhance their applications with reliable, fast, and purpose-built capabilities for parsing complex documents, including images and PDFs, are encouraged to explore these open-weight models available via Workers AI.

Source and news date

An updated Clairic and Clairic-flash are now open-weight, from the new Workers AI Superintelligence team — 2026-11-18

Related ServerNet services

ServerNet Support

ServerNet engineering & editorial team — specialists in infrastructure, networking and web hosting.

Linux Hosting
Share:

Comments 0

No comments yet — be the first!

Leave a comment

Related service

Linux Hosting

PHP & MySQL hosting on NVMe RAID-10 with LiteSpeed — the solid base for any website, from personal blogs to enterprise Laravel apps. At a price competitors can't explain.