Key points
- Passing a page URL with web grounding mode fetches text before searching, costing $0.001 per successful fetch.
- The high effort level now uses openai/gpt-6.1-terra and allows up to 40 steps, while xhigh supports 300 steps.
- The Analytics API credit-usage responses currently omit the Credit Source breakdown, requiring integration updates.
The Perplexity API changelog page now includes details on URL fetching for web grounding and updated effort levels for the Agent API. These changes affect how developers handle page freshness, model fallbacks and credit usage tracking in their integrations.These changes add costs for URL fetching and require updates to how effort levels and analytics responses are handled.The Agent API effort levels now pin specific model versions and step limits, while web grounding introduces fetch_url pricing and staleness controls.
The evidence
The changelog states that passing a page URL as state with mode set to web grounding fetches its text before searching. The fetched text replaces the URL before the search runs, with results returned in fetched_results and search evidence in search_results. Setting max_staleness defines the maximum page age in seconds since its last crawl, triggering a fresh crawl if an older stored copy exists. Each successful URL fetch costs $0.001 in addition to searches and input tokens.reported
The decision
The Agent API effort levels have been updated with new version pins and model assignments. The fast level falls back to openai/gpt-5.6-luna when openai/gpt-6-luna is unavailable.reported The low and medium levels raise max output tokens to 128,000 and increase web_search max_tokens to 4,000, using openai/gpt-6-luna for fetch_url content extraction.reported The high level uses openai/gpt-6.1-terra instead of openai/gpt-6-sol, raising max steps to 40.reported The xhigh level stays on anthropic/claude-opus-5-5 with openai/gpt-6.1-sol as fallback, raising max steps to 300.reported
What to make of it
A new ultra-preview effort level is available for broad research tasks where coverage matters most, though runs can take a long time. The undocumented ultra effort name is removed and now returns an error, while the wide-research alias now selects xhigh. The max_steps parameter now accepts values up to 500, up from 100.reported Developers using frozen configurations must copy updated values to keep matching each effort level's public API fields.
In practice
Developers must update integrations expecting a fixed set of breakdowns in the Analytics API credit-usage responses. The Credit Source breakdown is currently omitted, so systems should read the returned categories list and use count for total credits. The Perplexity Python library and TypeScript SDK do not yet have a mode parameter for web grounding, so it must be sent in the request body. The preset parameter is deprecated in favour of effort, though responses return both fields.
Why it matters
- Startups using the Agent API must update frozen configurations to match new effort level public API fields, or risk mismatches.
- Integrations relying on Credit Source breakdowns in Analytics API responses must adapt to reading the returned categories list instead.
- Developers using web grounding must send the mode parameter in the request body until SDKs add support.
- The removal of the undocumented ultra effort name means code using it will now return an error.
Topics: APIs, Perplexity, Agent API, web grounding, effort levels
Updated
Then / now / next
The Analytics API returned Credit Source breakdowns, and the undocumented ultra effort name was available.
Credit Source breakdowns are omitted, ultra returns an error, and URL fetching is documented with pricing.
Watch for SDK updates adding the mode parameter for web grounding and further effort level changes.
Story timeline
- Then The Analytics API returned Credit Source breakdowns, and the undocumented ultra effort name was available.
- 04 Nov 2026 Perplexity API Changelog page updated:
- 05 Nov 2026 Perplexity API Changelog page updated: Filters
- 05 Nov 2026 Perplexity API Changelog page updated:
- 05 Nov 2026 Perplexity launches its own take on AI agents for work
- 06 Nov 2026 Perplexity rolls out Teammates, an always-on AI agent system that integrates with Computer to handle multiple roles and execute large-scale tasks simultaneously (Emma Roth/The Verge)
- 06 Nov 2026 Claude API Release Notes page updated: You can now sign a member of your Claude Enterprise (claude.ai) organization out of Claude
- 06 Nov 2026 AI for scientific discovery: How researchers use Claude
- 06 Nov 2026 Perplexity's 'teammates' are the company's AI agents for daily work tasks
- 06 Nov 2026 Anthropic adds bank account access to Claude so users can track their spending
- 06 Nov 2026 Rogue AI tasked with renumbering pages accidentally deletes law professor's entire syllabus, one year of lecture notes
- 06 Nov 2026 Teachers and parents: what are your experiences of AI-generated emails?
- 06 Nov 2026 Perplexity now has its own always-on AI agent
- 06 Nov 2026 v2.1.319
- 06 Nov 2026 Claude API Release Notes page updated: We’ve launched Claude Opus 4.1, an incremental update to Claude Opus 4 with enhanced capab
- 06 Nov 2026 The Guardian view on the AI backlash: American voters want a say in their future | Editorial
- 06 Nov 2026 Perplexity launches AI agents for work
- 06 Nov 2026 Sources: Anthropic is continuing with plans to list before the Thanksgiving holiday this month, expecting to raise $80B-$100B at a valuation of $1.25T to $1.75T (Financial Times)
- Now Credit Source breakdowns are omitted, ultra returns an error, and URL fetching is documented with pricing.
- Next Watch for SDK updates adding the mode parameter for web grounding and further effort level changes.
The strongest doubt
Counterpoint from the sources
The Perplexity Python library and TypeScript SDK do not yet support the mode parameter for web grounding, requiring manual request body updates.
Ask the article
Local keyword search of this article; no generated answers.Did this article give you the evidence you needed?
Your choice stays on this device. No public tally.Questions readers ask
How much does URL fetching cost in the Perplexity API?
Each successful URL fetch costs $0.001 in addition to searches and input tokens.
What happens if I use the undocumented ultra effort name?
It is removed and now returns an error; use ultra-preview or xhigh instead.
Does the Analytics API still return Credit Source breakdowns?
No, responses currently omit the Credit Source breakdown, though Total credit usage and Model breakdown remain.
Connected stories
Reading receipt
- Passing a page URL with web grounding mode fetches text before searching, costing $0.001 per successful fetch.
- The high effort level now uses openai/gpt-6.1-terra and allows up to 40 steps, while xhigh supports 300 steps.
- The Analytics API credit-usage responses currently omit the Credit Source breakdown, requiring integration updates.
Still open: When will the Python and TypeScript SDKs add support for the mode parameter?
Still open: Will the Credit Source breakdown return to Analytics API responses, and if so, when?