The AI Index
Every major AI assistant, compared.
One independent table for the whole assistant market. Every figure below was checked against the provider's own pricing or model page in June 2026, and each row links to its source so you can verify the live number yourself. Where a provider does not publish a figure, we say so rather than guess.
| Assistant | Paid plan | Per month | Free tier | Flagship model | Context | Standout strength |
|---|---|---|---|---|---|---|
| ChatGPT OpenAI | Plus | $20 | Yes, capped | GPT-5.5 | 1M (API) | Strongest agentic task execution |
| Claude Anthropic | Pro | $20 | Yes, daily cap | Claude Opus 4.8 | 500K chat / 1M API | Best coding and agentic reliability |
| Gemini Google | AI Pro | $19.99 | Yes, capped | Gemini 3.1 Pro | 1M | Deep Google and multimodal integration |
| Microsoft Copilot Microsoft | Pro | $20 | Yes, capped | GPT-5.5 | Unconfirmed | Native Microsoft 365 integration |
| Perplexity Perplexity | Pro | $20 | Yes, limited | Sonar Pro | 200K | Best cited-source answer engine |
| Grok xAI | SuperGrok | $30 | Yes, ~10 / 2h | Grok 4.3 | 1M API / 128K app | Largest real-time and X data access |
| DeepSeek DeepSeek | No paid tier | Free | Yes, fully free | DeepSeek V4 | 1M | Lowest-cost frontier-class model |
| Meta AI Meta | AI+ | $10 | Yes, in WhatsApp/IG | Llama 4 | 10M (Scout, approx) | Largest open-weight context window |
| Mistral Le Chat Mistral | Pro | $14.99 | Yes, limited | Mistral Medium 3.5 | 256K (approx) | Cheapest paid plan, EU-based |
Three things the table shows
- Price has collapsed to one number. The four best-known assistants all land within a rand of each other at about $20 a month, so price is rarely the deciding factor between them. The real spread is at the edges: Meta AI+ at roughly $10 and DeepSeek at zero.
- Free no longer means weak. DeepSeek ships a frontier-class model for free, and Meta AI is free to billions inside the apps they already use. The paid tiers now sell convenience, limits and integration more than raw capability.
- Context windows are converging near 1M. Most flagships now sit around a million tokens, with the practical catch that the in-app chat product often exposes less than the model's API maximum.
How we compiled this
Each price, plan and model was read off the provider's official pricing or model page in June 2026, linked in the table. A few honest caveats, because a data page is only as good as its footnotes:
- Microsoft Copilot's context window is unconfirmed. Microsoft does not publish a token figure for the consumer Copilot app, so we left it blank rather than borrow GitHub Copilot's number, which is a different product.
- In-app and API context windows differ. Where they split, we show both (for example Claude at 500K in chat but 1M on the API and in Claude Code).
- Figures marked approx (Mistral 256K, Llama Maverick) are corroborated across sources but not pinned to a single first-party spec line.
Spotted a number that has moved? That is the nature of this market. Check the linked source for the live figure, and we update the table as plans change. Nothing here is invented; anything we could not verify is flagged, not filled in.
For the verdicts behind the data, see the ChatGPT vs Claude vs Gemini head-to-head or browse all of the rankings.