Perplexity Pro users hit drastic AI model usage limits as company pushes $2,000 Max tier

At a glance:

• Perplexity Pro subscribers are reporting they hit weekly limits on advanced AI models like Gemini 3.1 Pro or Thinking with far fewer queries than before, sometimes as low as 3-5 requests per day.
• Token limits may have dropped from 200 to 100, and weekly query caps on advanced models appear to have fallen to roughly 100-150, while regular models remain unaffected.
• The company is prompting affected users to upgrade to Perplexity Max, which costs $2,004 per year ($167/month) versus the Pro tier at $204 per year ($17/month).

What users are reporting

Several Perplexity Pro subscribers have taken to Reddit to voice frustration over what they describe as sharply reduced access to advanced AI models. The complaints span multiple threads and paint a consistent picture: limits that were once generous enough for daily research workflows are now being exhausted within hours or days.

Reddit user SaberToaster said they were running roughly 10-20 queries per day and still managed to max out their weekly allowance for advanced models. Another user, SuppliDev, reported hitting the wall with as few as three to five requests in a single day. A third Redditor, KhoaHachiman7, reached their weekly file upload limit after just two uploads. Across these reports, users independently identified what appears to be a systemic change rather than an isolated glitch.

The most specific detail comes from users who claim Perplexity quietly altered its token limits from 200 to 100 and cut the weekly query allowance for advanced models down to approximately 100-150. Regular, non-advanced models are reportedly unaffected, which suggests the restriction targets only the higher-end reasoning and multimodal tiers such as Gemini 3.1 Pro and the Thinking model.

The push toward Perplexity Max

When a Pro subscriber runs into the new ceiling, the on-screen prompt reads: "Get enhanced access to advanced models with Perplexity Max." That tier carries a steep price tag — $2,004 per year when billed annually, or $167 per month. By contrast, the standard Pro plan costs $204 per year ($17 per month when paid annually). The nearly tenfold price jump has not gone unnoticed by the community.

For users who signed up for Perplexity Pro expecting reliable access to its most capable models, the shift feels like a bait-and-switch. The advanced models — Gemini 3.1 Pro and Thinking — are precisely the selling point that differentiates Pro from the free tier. If those models are now gated behind a $2,000-a-year subscription, the value proposition of the existing Pro plan is significantly diminished.

Regional discrepancies and workarounds

One notable pattern in the Reddit complaints is that some users claim the reduced limits are not affecting subscribers in the US. Several affected users have suggested using a US-based VPN as a temporary workaround to bypass the restrictions. This geographic disparity hints that the change may be tied to regional rollout, server load management, or perhaps a staged rollout that hasn't been communicated publicly.

Perplexity has not yet issued an official statement on the allegations, nor has it updated its pricing page to reflect any new limits. Until the company responds, it's difficult to say whether the reported restrictions are intentional policy changes, temporary rate-limiting measures triggered by high demand, or regional outages masquerading as caps.

Why it matters for the broader AI search market

Perplexity has been positioning itself as a serious challenger to Google and traditional search by wrapping large language models in a conversational, citation-heavy interface. The Pro tier has been a key revenue driver, and its subscribers are among the most engaged users of the platform. If the company is indeed tightening access to its most capable models without clear communication, it risks eroding trust among the power users who generate the most traffic and word-of-mouth growth.

The situation also raises questions about how AI search companies will monetize increasingly expensive model inference. Running Gemini 3.1 Pro or a reasoning-heavy model at scale is far costlier than serving a lightweight embedding model, and Perplexity may be recalibrating its economics. Whether that recalibration should be imposed on existing subscribers without notice is a separate — and arguably more important — conversation.

What to watch next

The community will be watching for an official Perplexity response. If the company confirms the limit changes, subscribers will likely demand clearer communication about tier boundaries and a roadmap for what advanced model access will look like at each price point. If the issue turns out to be temporary server-side throttling, Perplexity will still face pressure to be more transparent about capacity constraints so users can plan their workflows accordingly.

For now, affected users outside the US may find short-term relief through VPNs, but that workaround is unlikely to hold up if Perplexity decides to enforce the new limits globally. The coming weeks should clarify whether this is a pricing strategy in motion or a growing pain that will be resolved.