Every score is derived from a 70+ parameter matrix – not from opinion numbers. Capability parameters score yes = 1, partial = 0.5, no = 0; quality parameters use a 4-dot rating (●●●●…); numeric specs score by tiers. A category score is the average of its parameters (0–100), and the final score is the weighted sum of the eight categories. See the full matrix and methodology.

ChatGPT scores highest at 86.2/100 in our weighted model. The best for you still depends on what you do most – check the category winners.

MultipleChat wins the value category (price, free tier, annual discount and how many models you get). It is computed straight from those facts.

Yes – every subscription here has iOS and Android apps. Desktop apps and a public API vary; see the Apps & ecosystem rows in the matrix.

Approximate per-file upload limits are in the matrix (e.g. ChatGPT ~512 MB, MultipleChat up to 200 MB on higher tiers, Perplexity ~50 MB, Claude ~30 MB). These change often, so confirm on the official page before relying on them.

Copilot scores highest on our privacy category (training opt-out, retention controls, certifications and data region). MultipleChat is close behind and notable for not training on customer data. Always check the data region and training settings for any provider.

Only one subscription here does that natively: MultipleChat runs several models side by side so you can pick the best answer, and its AI Collab modes (debate, ensemble, co-operative, expert) let the models work together on one task. The single-vendor apps give you one company's models.

No. Scores follow the published methodology applied to the matrix, not payment. Referral links never change a score, and we keep the data current.

We review prices, features and standings regularly and date the page. The AI market moves fast, so always confirm current details on the official provider page before subscribing.

We focus on the most widely used consumer AI subscriptions. We may add more (e.g. DeepSeek, Mistral, Meta AI) over time. Tell us what you'd like compared.

This is an independent, experimental project. Where links exist they go to official provider pages; they never change a score.

Quality parameters are rated ●●●● (excellent) down to ●○○○ (basic), based on recognized benchmarks and general standing. Capabilities use ✓ / ~ / ✗.

Yes — the full 70+ parameter matrix is public on the matrix page, including every rating and the per-category scores.

For most individuals, yes. The ~$20 standard tiers unlock the strongest models and core features; the $100–250 top tiers are for heavy or professional use.