Back to Blog
Industry UpdatesOctober 18, 2026Industry Updates

Perplexity Updates Its Agent API Model Lineup

Share:

TL;DR

In its October 2026 changelog, Perplexity updated its Agent API model documentation, pricing calculator and examples. It removed several older Anthropic Claude, Google Gemini and xAI Grok models and now recommends claude-opus-5-5, claude-sonnet-5-5, claude-haiku-5-5, gemini-4-argon, gemini-3.8-flash or grok-4.7 for new integrations. The docs also added two new models, GPT-6 Astra and DeepSeek V4 Pro, each with its own per-token pricing, with DeepSeek V4 Pro also available through the Orchestrator API.

Key takeaways

  • Perplexity's Agent API no longer lists several older Anthropic Claude, Google Gemini and xAI Grok models in its docs, pricing calculator or examples.
  • For new integrations, Perplexity recommends claude-opus-5-5, claude-sonnet-5-5, claude-haiku-5-5, gemini-4-argon, gemini-3.8-flash or grok-4.7.
  • The docs added GPT-6 Astra, priced at $10.00 per million input tokens and $50.00 per million output tokens for up to 272k input tokens, with flex and priority tiers.
  • The docs also added DeepSeek V4 Pro, priced at $1.32 per million input tokens, $0.044 per million cached input tokens and $3.96 per million output tokens, and it's available through the Orchestrator API too.

Perplexity: Perplexity Updates Agent Models

Perplexity removed several older models from its Agent API model docs, pricing calculator, and examples, including multiple Anthropic Claude Opus, Sonnet, and Haiku versions, Google Gemini versions, and xAI Grok versions. For new integrations, it now recommends anthropic/claude-opus-5-5, anthropic/claude-sonnet-5-5, anthropic/claude-haiku-5-5, google/gemini-4-argon, google/gemini-3.8-flash, or xai/grok-4.7. The model docs also added openai/gpt-6-astra, priced at $10.00 per million input tokens and $50.00 per million output tokens up to 272k input tokens (with flex and priority tiers), and perplexity/deepseek-v4-pro-0813, priced at $1.32 per million input tokens, $0.044 per million cached input tokens, and $3.96 per million output tokens, which is also available through the Orchestrator API.

  • Removed from Agent API docs, pricing calculator, and examples: older Anthropic Claude Opus/Sonnet/Haiku models, Google Gemini models, and xAI Grok models.
  • Recommended for new integrations: claude-opus-5-5, claude-sonnet-5-5, claude-haiku-5-5, gemini-4-argon, gemini-3.8-flash, or grok-4.7.
  • Added openai/gpt-6-astra: $10.00 per million input tokens, $50.00 per million output tokens, up to 272k input tokens, with flex and priority tiers.
  • Added perplexity/deepseek-v4-pro-0813: $1.32 per million input tokens, $0.044 per million cached input tokens, $3.96 per million output tokens, also available through the Orchestrator API.

Why it matters: Anyone building workflows on Perplexity's Agent API, including AI-visibility monitoring tools, should check whether they reference any of the removed model names before an integration quietly breaks. The new pricing also gives a basis for comparing cost when choosing a model for high-volume monitoring tasks.

Source: Perplexity: Changelog - Perplexity: new entries (2026-10-17)

What this means for your brand's AI visibility

At Sophyx, we watch which models power the AI agents and answer engines that mention brands, and a lineup change like this is a reminder that the model running behind any assistant can shift without much notice. If a brand-monitoring or AI-visibility workflow is built on one of Perplexity's Agent API models, it's worth checking whether that model is still listed before assuming it will keep working the same way.

We also think the new pricing is worth a look if you're running high-volume prompts to see how your brand shows up across models. Choosing between GPT-6 Astra, DeepSeek V4 Pro, or one of the newer Claude, Gemini or Grok options now means comparing their per-token costs directly, since Perplexity lists them side by side in the same docs and pricing calculator. If your brand relies on Perplexity for AI-driven discovery, you can track how Perplexity mentions your brand across the specific models it serves answers from. Our Perplexity rank tracker shows exactly where your brand lands when Perplexity cites sources in response to shopping and research queries. As Perplexity rotates in new models like GPT-6 Astra and DeepSeek V4 Pro, now is a good time to improve your brand's visibility in AI search engines before answer quality shifts again.

Frequently asked questions

What models did Perplexity remove from its Agent API?

Perplexity removed several older Anthropic Claude models (including claude-opus-4-8, claude-opus-4-7, claude-opus-4-6, claude-opus-4-5, claude-sonnet-4-6, claude-sonnet-4-5 and claude-haiku-4-5), Google Gemini models (gemini-3-flash-preview, gemini-3.1-pro-preview, gemini-3.1-flash-lite, gemini-3.5-flash and gemini-3.5-flash-lite), and xAI Grok models (grok-4.3, grok-4.20-reasoning, grok-4.20-non-reasoning and grok-4.20-multi-agent) from its model docs, pricing calculator and examples.

Which models does Perplexity recommend for new Agent API integrations?

Perplexity recommends anthropic/claude-opus-5-5, anthropic/claude-sonnet-5-5, anthropic/claude-haiku-5-5, google/gemini-4-argon, google/gemini-3.8-flash, or xai/grok-4.7 for new integrations.

What is GPT-6 Astra and how much does it cost on Perplexity's Agent API?

GPT-6 Astra is a new model listed in Perplexity's Agent API model docs, priced at $10.00 per million input tokens and $50.00 per million output tokens for up to 272k input tokens, with flex and priority tiers also available.

Is DeepSeek V4 Pro available on Perplexity?

Yes, perplexity/deepseek-v4-pro-0813 is listed in the Agent API model docs at $1.32 per million input tokens, $0.044 per million cached input tokens and $3.96 per million output tokens, and it's also available through the Orchestrator API.

Sources are linked above. Written by Reza Kashi from each company's own announcement; we update this brief every weekday morning.

Comments

Loading comments…

Related articles

Book a Demo