OpenAI API
Metered. Input, cached input, and output tokens; some tools and media have separate units.
Anthropic Claude API
Metered. Input, output, and prompt-cache tokens.
Deepgram Speech and Voice Agent APIs
Credits. Audio duration, text characters, or Voice Agent connection minutes by product.
AssemblyAI Speech-to-text API
Metered. Audio duration; model and add-ons change the rate.
ElevenLabs ElevenCreative
Credits. Credits consumed by generation and speech features.
GitHub Copilot
Hybrid. User seats plus included AI usage credits.
Cursor AI coding assistant
Hybrid. Seats and model usage.
Replicate Model hosting
Metered. Hardware runtime or model input and output.
Modal Compute
Hybrid. GPU, CPU, and memory time.
Pinecone Database
Hybrid. Read units, write units, storage, and other product meters.
Vercel Managed infrastructure
Hybrid. Resource-specific usage metrics.
How this collection was made
We read each company's public pricing or documentation page on November 3, 2026 and recorded only mechanics stated there. The source link accompanies every example. Public pages may omit negotiated terms, internal billing operations, and recent changes. “Not publicly specified” means the cited page did not establish that field. Lago's tradeoffs and implementation notes are our interpretation, not descriptions of these companies' systems or relationships with Lago.
For model selection, read the usage-based pricing playbook. For voice-specific design, read the voice AI guide. The billing glossary defines the underlying terms.