Published rates

The meter, in the open.

Most companies run one AI. An AI Partners workspace runs more than a dozen, across five providers, each picked for its job: the cheapest model that does the work, the genius model when it counts. This is the rate card behind every usage statement. Your monthly bill itemizes exactly which model ran, for whom, on what, at these rates.

Rates as of July 2026, refreshed as provider prices move. A token is roughly a word. Most everyday tasks cost fractions of a cent: a 500-word assistant answer on our workhorse model costs about 1.2 cents.

Language models

Matched to the job, not the hype.

ModelWhat we use it forInput / 1M tokensOutput / 1M tokens
Gemini FlashFast research, image work, high-volume tasks$0.36$3.00
Claude HaikuEmail scanning, quick classification$1.20$6.00
Gemini ProDeep research with live grounding$2.40$14.40
Claude SonnetThe everyday assistant workhorse$3.60$18.00
Claude OpusAgents and complex multi-step work$6.00$30.00
GPT-5.5Second opinions, specialty tasks$6.00$36.00
Claude FableBusiness-advisor grade reasoning$12.00$60.00
Voice & media

Talking, listening, and transcribing.

ServiceWhat it powersRate
Realtime voiceLive voice conversation with your assistant$0.36 / minute
Live dictationStreaming speech-to-text as you talk$0.012 / minute
Meeting transcriptionRecorded meetings into notes and memory$0.45 / hour
Text-to-speech, standardSpoken replies and walkthroughs$0.018 / 1k characters
Text-to-speech, premium voiceNarration and client-facing audio$0.36 / 1k characters
Web, browsing & search

When your AI goes out and looks.

ServiceWhat it powersRate
Web searchAssistant and agent web lookups$0.012 / search
Grounded researchResearch answers cited against live sources$0.042 / request
Agent browser timeAn agent driving a real browser for you$0.15 / hour
Document embeddingsMaking your documents searchable by meaning$0.024 / 1M tokens
How billing works

Itemized, capped, and never a surprise.