Perplexity Updates

Everything Perplexity has shipped, in one place.

This Week

Perplexityapi

Link·Sep 9, 2026

OAuth for the Perplexity API MCP Server

Connecting an assistant to Perplexity's search now works by signing in, instead of pasting an API key.

Best for: Wiring live web search into an AI assistant without handling or storing a key
When to use it: You are setting up Perplexity search inside an assistant for a team and need to pick which organization pays
Who benefits: Anyone on a Perplexity team plan who wants search in their assistant without pulling in a developer

Perplexityapi

Link·Sep 9, 2026

Perplexity Search API in Hermes Agent

Hermes Agent can now search the live web through Perplexity's Search API.

Best for: Giving an agent current web results instead of only what its model already knows.
When to use it: The agent's answers depend on information that changes faster than a training cutoff.
Who benefits: Developers building agents that need to check the web mid-task.

Last Week

Perplexitymodel

Link·Sep 4, 2026

GPT-6 Astra in Perplexity Computer

OpenAI's GPT-6 Astra is now selectable in Perplexity Computer, for Pro and Max subscribers.

Best for: Running Perplexity's agent work on OpenAI's newest model without leaving Perplexity.
When to use it: The task is long-running or agentic and you want Astra handling it rather than Perplexity's default.
Who benefits: Perplexity Pro and Max subscribers who use Computer for multi-step work.

Perplexityapi

Link·Sep 3, 2026

Perplexity API in Stripe Projects

You can provision a Perplexity API project, key, and prepaid credits from the Stripe command line.

Best for: getting an API key and its billing set up without leaving the terminal
When to use it: you are starting a new project and don't want to go through a separate signup and billing flow
Who benefits: developers who already manage billing through Stripe

Perplexityfeature

Link·Sep 3, 2026

Perplexity Portable Computer on Linux

The fully local version of Perplexity Computer now runs on ordinary Linux PCs with an NVIDIA RTX graphics card.

Best for: running the whole Perplexity agent on hardware you already own instead of buying a dedicated machine
When to use it: you work on Linux and want the local-first version rather than the cloud one
Who benefits: Linux users with an NVIDIA RTX card carrying 24GB of VRAM or more

Perplexityfeature

Link·Sep 2, 2026

Lily

Perplexity's on-device inference engine for Mac, now open source.

Best for: running a capable model locally on a Mac fast enough that on-device work doesn't become the bottleneck
When to use it: you want steps involving private files or sensitive data to stay on the machine
Who benefits: developers building local-first AI on Apple silicon, and Perplexity Computer users on a Mac

Perplexitymodel

Link·Sep 1, 2026

Claude Fable 5.1 in Perplexity Computer

Perplexity Computer adds Anthropic's Claude Fable 5.1 as a model option for Pro and Max subscribers.

Best for: Running agent-style tasks in Perplexity Computer on Anthropic's newest model
When to use it: You want the model Perplexity's own testing ranked first on both quality and cost for this kind of work
Who benefits: Perplexity Pro and Max subscribers using Computer

Perplexityfeature

Link·Sep 1, 2026

Hybrid Compute in Perplexity Computer

Perplexity Computer can start a task in the cloud and finish the sensitive parts on a model running on your own Mac.

Best for: Work that touches private files you would rather not send to a cloud service
When to use it: Part of a task involves confidential documents or client data but the rest does not
Who benefits: Perplexity Mac users handling sensitive material, and anyone with a compliance reason to keep certain steps local

August 2026

Perplexitymodel

Link·Aug 28, 2026

GLM 5.3 in Perplexity Computer

GLM 5.3 is now one of the models you can run inside Perplexity Computer.

Best for: Long research tasks where the agent has to hold a lot of context and cite what it found.
When to use it: A job runs long, spans text and images, and needs evidence attached to the answer.
Who benefits: Perplexity Computer users who pick a model per task instead of taking the default.

Perplexityapi

Link·Aug 27, 2026

Connectors in the Perplexity Agent API

Agents on the Perplexity Agent API can reach GitHub, Slack, Google Drive, and Datadog without you passing credentials on every call.

Best for: Agents that need to touch the same work tool over and over across many requests.
When to use it: You are passing the same server URL and token on every API call and want to stop.
Who benefits: Developers building on the Agent API, and the API Group admins who set up access for a whole team.