TechniqueRAG5 recent entries20 May 2026Context compression isn't new in RAG. Our contribution is making it query-aware, citation-preserving, and fast enough for orchestration. Rea…Context compression isn't new in RAG. Our contribution is making it query-aware, citation-preserving, and fast enough for orchestration. Read the full research blog: https://research.perplexity.ai/art→4 Sept 2026ROSE is the model engine. It reuses the same kernels for LLMs and embeddings. For embeddings, it skips the KV cache and uses ragged attentio…ROSE is the model engine. It reuses the same kernels for LLMs and embeddings. For embeddings, it skips the KV cache and uses ragged attention instead of paged attention. ROSE supports multiple attenti
TechniqueAgents8 recent entries2 Sept 2026Today we’re open-sourcing Lily, the local inference engine we built for hybrid compute in Perplexity Computer. Lily is specialized for Qwen3…Today we’re open-sourcing Lily, the local inference engine we built for hybrid compute in Perplexity Computer. Lily is specialized for Qwen3.6-35B-A3B on Apple silicon, built so on-device compute does→3 Sept 2026Portable Computer is now available on Linux for @NVIDIA RTX GPUs with 24GB of VRAM or higher.Portable Computer is now available on Linux for @NVIDIA RTX GPUs with 24GB of VRAM or higher. Today we’re launching Portable Computer on @NVIDIA DGX Spark. Portable Computer is a fully local version o→9 Sept 2026We're introducing Q2D-Web (Query2Doc-Web), a benchmark and public leaderboard for evaluating retrieval in agentic RAG systems. Q2D-Web tests…We're introducing Q2D-Web (Query2Doc-Web), a benchmark and public leaderboard for evaluating retrieval in agentic RAG systems. Q2D-Web tests how embedding models perform on large-scale web search usin→9 Sept 2026Q2D-Web uses agent citations, production web rankings, and a combined set expanded with LLM judgments. These three relevance sets reduce fal…Q2D-Web uses agent citations, production web rankings, and a combined set expanded with LLM judgments. These three relevance sets reduce false negatives and reliance on a single labeling pipeline, whi→28 Sept 2026We’re partnering with Nvidia and 100+ industry partners to build infrastructure that contains rogue AI agents. In this research, we gave 9 A…We’re partnering with Nvidia and 100+ industry partners to build infrastructure that contains rogue AI agents. In this research, we gave 9 AI models root access inside SPACE and told them to break out→29 Sept 2026Agent governance is an engineering problem. We’ve built safeguards into Perplexity’s infrastructure, harnesses, and tools, and put them to w…Agent governance is an engineering problem. We’ve built safeguards into Perplexity’s infrastructure, harnesses, and tools, and put them to work across our products. Today we're sharing how we engineer→7 Oct 2026In agentic search, a GPT-OSS-120B agent using the 9B model answers 64.0% of BrowseComp+ questions correctly, 4.9 points ahead of the next Co…In agentic search, a GPT‑OSS‑120B agent employing the 9B model achieves a 64.0 % accuracy on BrowseComp+ questions. This performance is 4.9 points higher than the next best ColBERT model and requires →12 Oct 2026This is how we built the filesystem behind Perplexity Projects as a Git forge for agents. Each agent works in its own sandbox and commits to…This is how we built the filesystem behind Perplexity Projects as a Git forge for agents. Each agent works in its own sandbox and commits to a shared repository where people and other agents can build
TechniqueMultimodal2 recent entries28 Aug 2026GLM 5.3 is now available in Perplexity Computer. Built for long-context, multimodal agent workloads, it beat GLM 5.2 on WANDR, our benchmark…Perplexity AI has released its new language model GLM 5.3 on Perplexity Computer. Designed for long‑context, multimodal agent workloads, GLM 5.3 outperformed the previous GLM 5.2 on the company’s WAND→6 Oct 2026pplx-decider-v1.1-27b, our updated open weights multimodal decision model, is now available. It scores the highest on the new @huggingface D…pplx-decider-v1.1-27b, our updated open weights multimodal decision model, is now available. It scores the highest on the new @huggingface Decision Index 0.3 benchmark. pplx-decider-v1.1-27b costs hal
TechniqueSafety2 recent entries30 Jul 2026“AI Meltdown” is the scenario where the agent goes off the rails and forgets/ignores all previous instructions and soft guardrails. No promp…“AI Meltdown” is the scenario where the agent goes off the rails and forgets/ignores all previous instructions and soft guardrails. No prompt injection or malicious actor is needed for this to happen.→29 Sept 2026Agent governance is an engineering problem. We’ve built safeguards into Perplexity’s infrastructure, harnesses, and tools, and put them to w…Agent governance is an engineering problem. We’ve built safeguards into Perplexity’s infrastructure, harnesses, and tools, and put them to work across our products. Today we're sharing how we engineer