# A new line of advanced processors: Ultra2x, Ultra4x, and Ultra8x

Starting today, three new Deep Research processors are generally available in the Parallel Task API[Task API]($https://parallel.ai/blog/parallel-task-api), giving you precise control over research depth and accuracy for any query complexity.

## **State-of-the-Art Performance, Now Scalable**

Last month, our Ultra processor achieved 27% accuracy on BrowseComp[27% accuracy on BrowseComp]($https://parallel.ai/blog/deep-research) - outperforming human experts given two hours per problem and establishing new state-of-the-art results on this challenging deep research benchmark. Building on this foundation, we tested scaled versions with 2x and 4x compute allocation, demonstrating consistent performance improvements: our 4x processor reached 48% accuracy, demonstrating that increased budget translates directly to enhanced research capabilities.

Today, these processors are publicly available as Ultra2x and Ultra4x, joined by our most advanced offering yet: Ultra8x.

## **Flexible Compute for Every Research Challenge**

Each Deep Research processor represents a different compute allocation, allowing you to match processing power to task complexity:

• - **Pro**: Advanced research for sophisticated queries requiring multi-source synthesis
• - **Ultra**: Deep research for complex multi-hop queries
• - **Ultra2x**: Enhanced deep research with 2x compute for challenging investigations
• - **Ultra4x**: Advanced deep research capabilities with 4x compute for comprehensive analysis
• - **Ultra8x**: Maximum deep research with 8x compute for the most demanding queries

With each step up, Parallel considers more web pages, reads more comprehensively, and reasons more deeply across scattered sources. This architecture gives you the flexibility to dial up performance for critical tasks or optimize costs for routine queries - a level of adaptability unavailable in specialized systems.

## **Start Building**

**Get started with Ultra2x, Ultra4x, and Ultra8x today. **Try out the processors in our Developer Platform[Developer Platform]($https://platform.parallel.ai/play/deep-research) or dive directly into our documentation[documentation]($https://docs.parallel.ai/core-concepts/processors).