Computer adds effort mode for model selection
Perplexity introduced "effort controls" (from Light to Ultra) for Perplexity Computer, allowing users to balance model selection, reasoning depth, and cost.
143 ships tracked·1 sources·0 events
Perplexity introduced "effort controls" (from Light to Ultra) for Perplexity Computer, allowing users to balance model selection, reasoning depth, and cost.
Perplexity Computer introduced 'effort controls' (Light, Standard, High, Ultra), allowing users to balance model reasoning depth and cost for specific tasks.
Perplexity Computer introduced 'effort controls' that allow users to adjust the reasoning level and model selection for tasks, ranging from 'Light' to 'Ultra'.
Perplexity maintains a dedicated blog and news hub on its platform for company updates and announcements.
Perplexity partnered with HP to pre-install the Perplexity Windows app on select devices, starting with the HP ZBook Ultra G3a.
Perplexity and HP partnered to pre-install the Perplexity Windows app on new HP devices, starting with the ZBook Ultra G3a.
Perplexity introduced CobbleDB, a high-performance key-value storage layer designed for AI search workloads, promising lower latency and reduced costs.
Perplexity released a version of its 'Portable Computer' tool for Windows, allowing users to keep sensitive work local while automating tasks and long workflows.
Nimbus builds production AI systems — internal tools, customer agents, retrieval pipelines — combining humans and AI end-to-end. From scoped pilot to production in 4–8 weeks.
Perplexity has expanded its 'Portable Computer' feature to Windows, allowing Pro and Max subscribers to run local AI models on compatible PCs with NVIDIA RTX GPUs without cloud dependency.
Perplexity announced CobbleDB, a purpose-built storage layer for AI search designed to reduce latency and infrastructure costs. Initial testing indicates a 5x improvement in batch-read latency and approximately 20%…
Portable Computer, which runs AI models locally, is now available in the Perplexity app for Windows on devices with compatible NVIDIA RTX GPUs.
Perplexity introduced a credit-based system for its new premium product, Computer on Perplexity, offering varying monthly allocations for Pro and Max subscribers.
Perplexity has updated its Research feature with Advanced Deep Research and introduced Learn Mode for educational use, alongside an Email Assistant for Max subscribers.
Perplexity announced that its Sonar chat completions endpoint will be retired on September 27, 2026, with users required to migrate to the Agent API. The Agent API offers enhanced capabilities, including multi-step…
Perplexity introduced Q2D-Web, a private benchmark and public leaderboard designed to evaluate first-stage retrievers in agentic RAG systems at a large, web-scale size.
Perplexity Health is a new personalized health experience for US-based Pro and Max subscribers, integrating data from medical records, wearables, and wellness apps to provide context-aware insights.
Perplexity continues to expand its ecosystem integrations, recently highlighting connectivity with tools like Asana, Confluence, Jira, and Google Drive to streamline research and information gathering.
Perplexity published a blog post discussing how businesses can derive real ROI from AI integration by connecting data, tools, and workflows.
Perplexity published a technical blog post detailing their serving infrastructure for high-performance embeddings on GPUs.
Perplexity updated its Advanced Deep Research feature with improved capabilities, such as an enhanced code sandbox for data analysis, direct document processing, and better web-browsing for hard-to-reach sources.
Perplexity detailed its Lily local inference engine, optimized specifically for Apple silicon to enable high-performance, on-device AI tasks for models like Qwen3.6-35B-A3B.
Perplexity is retiring its organization file repository for Enterprise users on October 1, 2026, requiring users to download or migrate their files to Projects or supported File Connectors like Google Drive or…
Perplexity introduced Hybrid Compute for Mac, allowing tasks to be split between cloud-based AI and local models on Apple Silicon to enhance privacy and performance. An on-device PII classifier has also been…
A custom local inference engine was released for Apple Silicon, aimed at significantly improving prefill and decode throughput for on-device AI tasks.
Nimbus builds production AI systems — internal tools, customer agents, retrieval pipelines — combining humans and AI end-to-end. From scoped pilot to production in 4–8 weeks.
Perplexity launched Portable Computer, a local-first AI agent system for the NVIDIA DGX Spark that runs models locally to ensure data privacy and zero inference costs for high-volume tasks.
An informative article that outlines seven signs of AI hallucinations and provides a practical checklist for verifying AI-generated claims and citations.
Perplexity published blog posts regarding the ROI of enterprise AI integration and a guide on identifying AI hallucinations.
Perplexity released PII-TRACE, a new benchmark for evaluating personally identifiable information (PII) detectors, alongside PII-Tracer, a compact 0.6B model designed to run locally to detect PII before data is sent to…
A guide detailing the capabilities and limitations of AI personal assistants and providing best practices for their use.
Perplexity released the Perplexity Computer MCP Server, enabling AI assistants to delegate complex, multi-step tasks to Perplexity Computer via the Model Context Protocol.