--- title: "Model Rates, ProxyAI" description: "Compare ProxyAI model rates for AI chat, voice and Bot Knowledge. See usage prices and choose a model for your chatbot." canonical: "https://www.proxyai.app/pricing" language: "en" --- > For the complete documentation index, see [llms.txt](https://www.proxyai.app/llms.txt). Request public content pages with `Accept: text/markdown`. Try ProxyAI free for 7 days Every new account unlocks full access to all ProxyAI features, letting you explore and see if we're the right fit for your needs. [Start free trial](https://www.proxyai.app/signup) # Model Rates Current pricing per million tokens for paid Chat models. New accounts can also select Free models for seven days after creating their first chatbot. Estimate your monthly cost Choose a model and expected context size to estimate AI replies together with Bot Knowledge. LLM model Gemini 3.5 Flash Lite Active models from Model Selection Ask and reply words per message 800 words 100 words 1,000 words Average conversation turns 3 turns 1 turns 20 turns No. of messages per month 1,000 messages 100 messages 100,000 messages Knowledge document size 10 MB 1 MB 100 MB ≈ 3,239 chunks across all documents Context window 5,605 of 32k tokens · 17% 32k 16k 0 - AI response 853 80% of the words - User input 213 20% of the words - Conversation history 1,067 1.0 earlier exchanges on average - Knowledge retrieval 3,072 3 chunks of 1,024 tokens - System prompt 400 Platform instructions 4,752 in, 853 out per message 80% cached from turn 2 | Documents implant | $0.31 | one-off · 6,479 requests | | --- | --- | --- | | Document storage | $0.04 | monthly · 10 MB documents | | Knowledge retrieval | $0.10 | monthly · 1,000 hybrid requests | | AI cost | $8.39 | monthly · 1,000 messages | | Monthly total | $8.53 | AI + storage + retrieval, excluding implant | AI estimate converts every 0.75 words to 1 token. From the second turn of a conversation, 80% of the input is assumed to be served from the provider's prompt cache at the model's cache price; a model without cache pricing is charged the normal input price. Each message is charged for its whole context as input (system prompt, tools, retrieved chunks, earlier turns of the conversation and the customer's words) and the reply as output. Actual usage varies by conversation. Actual chunk counts depend on document structure. Implant is charged again when a document is re-uploaded. Bot Knowledge Charged on vector operations and stored size rather than tokens. Embedding is included, documents are indexed for search inside the vector database, with no separate embedding charge. | Item | Rate | Unit | | --- | --- | --- | | Implant & query | $0.96 | per 10,000 requests | | Stored documents | $0.36 | per 100MB / month | Search is hybrid, so one document chunk and one retrieval each count as two requests, once for semantic matching and once for keyword matching. Storage is settled monthly. [Bot Knowledge guide](https://www.proxyai.app/help/add-ons/bot-knowledge-guide) Built-in helpdesk Running the ticket board costs nothing. Ticket email sends through the support mailbox you already own — an app password, no new accounts, no DNS. Opening ticketseach ticket the assistant opens counts as one store operation, same rate as cart and order lookups Store action rate Built-in desk Free Connecting another helpdeskGorgias, Zendesk, Freshdesk, Help Scout — you pay them, not us Free Connecting your mailboxreplies land back on your ticket board automatically Free Per 100 emails we send $1 Per 100 emails we receivereplies and forwarded emails that land on your ticket board $1 Estimate your monthly ticket cost Choose a model and expected email size to estimate AI draft replies on the built-in helpdesk. LLM model Gemini Pro Latest Models curated for the ticket agent Ask and reply words per message 800 words 200 words 2,000 words Average emails to resolve 3 emails 1 emails 10 emails No. of tickets per month 200 tickets 100 tickets 1,000 tickets Context window 5,755 of 32k tokens · 18% 32k 16k 0 - AI response 853 80% of the words - User input 213 20% of the words - Email thread 1,067 1.0 earlier exchanges on average - Knowledge retrieval 3,072 3 chunks of 1,024 tokens - System prompt 550 Helpdesk instructions 4,902 in, 853 out per email | AI cost | $33.01 | monthly · 200 tickets · 600 drafts | | --- | --- | --- | | Emails received | $6.00 | monthly · 600 emails | | Emails sent | $6.00 | monthly · 600 emails | | Monthly total | $45.01 | AI + emails received + emails sent | AI estimate converts every 0.75 words to 1 token. Each customer email gets one draft reply, charged for the whole thread so far as input (the helpdesk instructions, three 1,024-token knowledge chunks, the earlier emails and the customer's words) and the draft as output. The knowledge search itself, storage and prompt caching are not included. Every customer email is counted as received and every draft as sent. Actual usage varies by ticket. Agent inbox Taking over conversations is included with Human Handoff. Linking the phone app is the only thing with a price on it. First device paired Free Each device after that $1 Sentry visitor analytics Billed per visitor session, only while tracking is switched on. The dashboard costs nothing to use. Per 100 visitor sessionscharged as each 100 completes; one session is one visitor tab, however many pages they view $0.05 Analytics dashboard Free Visitors over your daily limitnot tracked, not billed Free