On February 5, 2026, Perplexity launched Model Council. Send a single query to Claude Opus 4.6, GPT-5.2, and Gemini 3.0 simultaneously, and a synthesis model distills the three responses — highlighting where they converge and where they diverge. While the rest of the AI search market races to build a better single model, Perplexity went the other direction.
How It Works
The flow is straightforward.
Same query → 3 models in parallel → synthesis analysis → unified output
Disagreement between models isn't hidden — it's surfaced. The divergence itself becomes useful information.
For questions where models routinely give different answers — investment analysis, medical information, legal gray areas — knowing where they diverge can be more actionable than picking a single answer and trusting it. Model Council is built around that premise: uncertainty is data, not a failure state.
Where It's Actually Useful
Investment Research
Compare three models' takes on a stock or sector. Shared risk factors get surfaced; points of disagreement flag where judgment calls live.
Decision-Making
Present a strategic choice and see what reasoning each model brings — and where their conclusions split. Useful for stress-testing assumptions.
Fact-Checking
When three models agree on a sensitive claim, confidence goes up. When they don't, that's your signal to verify further before acting.
Brainstorming
Three models trained on different data and objectives will approach the same problem differently. That structural diversity is the feature.
Pricing: $200/mo, Max Only
Model Council is locked to Perplexity Max. It's not in the $20/mo Pro plan.
With ChatGPT Pro at $200/mo and Claude Max at $100/mo, the premium AI market has clearly settled into a high-stakes tier. For legal, financial, or medical professionals whose work is research-intensive, the math can pencil out. For most users, this remains firmly out of reach — which is exactly what defines who Model Council is actually for.
Perplexity February 2026: Full Changelog
Model Council wasn't the only thing Perplexity shipped in February.
-
Deep Research
79.5% accuracy On complex multi-step research tasks. Benchmark methodology not publicly disclosed.
-
Sonar API
1,200 tok/s inference speed Enterprise-focused update targeting real-time application developers.
-
Memory
95% long-term memory recall Improved context retention for returning users. Increasingly a real differentiator in the assistant wars.
-
Comet
AI-native browser — global launch An attempt to own the AI layer at the browser level rather than the search box.
-
No Ads
Ad-based revenue model officially abandoned Full commitment to subscription. A clean break from Google's ad-dependent search model.
AI Search Market: Where Perplexity Sits
Current AI search market share, per FirstPageSage's 2026 report.
$20B valuation · $1.22B raised · $656M ARR target for 2026. 2% market share in absolute terms, but +370% year-over-year growth.
AI Service Comparison
| Feature | ChatGPT | Claude | Gemini | Perplexity |
|---|---|---|---|---|
| Provider | OpenAI | Anthropic | Perplexity AI | |
| Free plan | Y | Y | Y | Y |
| Entry paid | $20/mo (Plus) | $20/mo (Pro) | $19.99/mo | $20/mo (Pro) |
| Premium paid | $200/mo (Pro) | $100/mo (Max) | $249.99/mo | $200/mo (Max) |
| Web search | Y | Y | Y | Y (default) |
| Multi-model consensus | N | N | N | Y (Max only) |
| Long-term memory | Y | Y | Y | Y |
| Proprietary browser | N | N | N | Y (Comet) |
| Ad model | N | N | N | N (dropped) |
Takeaway
What makes Model Council interesting isn't the technical execution — it's the underlying bet. Perplexity is explicitly saying that no single model wins on every question, and that surfacing disagreement is more honest and more useful than pretending otherwise. That's a meaningful design stance in a market full of services competing to be your one AI.
Whether it justifies $200/mo depends entirely on who's paying. For researchers, analysts, or professionals who use AI to inform high-stakes decisions, eliminating even one bad confident answer per week could be worth it. For everyone else, it's a premium feature on a premium tier — and that's a strategic choice, not an oversight.
References
- Perplexity Hub Blog — Introducing Model Council (2026.02.05)
- Perplexity Changelog — February 2026
- TechCrunch — Perplexity hits $20B valuation
- MacRumors — Perplexity discontinues advertising
- FirstPageSage — AI Search Market Share Report 2026
