ChatGPT vs Perplexity
How ChatGPT and Perplexity compare on constraint stability — usage limits, pricing, model access and transparency over time.
Constraint stability only — pricing, limits, model availability, transparency. Not a capability comparison.
01 / SIDE-BY-SIDE
| Tool | Verdict | Pricing | Limits | Models | Transparency | 90d | Last change |
|---|---|---|---|---|---|---|---|
| ChatGPT | Stable | Stable100 | Stable99 | Stable97 | Stable100 | 2 | Increases model availability for free tier users, removing previous access restrictions. (Aug 2026) |
| Perplexity | Stable | Stable100 | Shifting72 | Stable100 | Shifting72 | 2 | Adds enterprise role-based access controls, a custom API credential vault, Brain access for Max subscribers and Opus 5. (Jul 2026) |
02 / WHICH CHANGES THE RULES LESS?
Disclosure is the fault line. ChatGPT's changes arrive labeled and often get softened: the 2025-08-08 GPT-5 launch imposed 80 messages/3h on Plus plus a 200/week Thinking cap, and by 2025-09-09 backlash had doubled that to 160/3h with Thinking raised to 3,000/week. Even the harder moves were public — contextual ads on Free and Go in a US test on 2026-02-09, legacy models (GPT-4o, 4.1, o4-mini) retired from the picker on 2026-02-13. Perplexity's most damaging event was the inverse in style: on 2026-02-06 Pro Deep Research was quietly cut from roughly 500 queries/day to about 20/month — a ~95% reduction framed as 'allocating more compute per session' — the same silent stroke halving Pro Labs (50 to 25/month) and capping previously unlimited uploads at 50/week, prompting a user to say an annual plan's limits had been 'quietly gutted.' Perplexity does loosen elsewhere (Max launched 2025-07-17 with uncapped research, Comet went free worldwide 2025-10-02, Deep Research upgraded to Opus 4.6), but its signature tightenings were flagged silent while ChatGPT's were announced. ChatGPT is the safer commitment for a stability-minded buyer — you can watch caps move and expect walk-backs, its friction mostly per-model message rationing. Perplexity pays off only if your workload fits inside the post-cut allowances; heavy Deep Research or upload users are the ones it has burned, making a long annual prepayment there riskier than ChatGPT's monthly quota churn.