AI Platform Pricing: Examples & Companies

What is an AI platform?

AI platforms are software products that give developers, businesses, or end users the capabilities of AI models — whether via API, a hosted interface, or a workflow builder. The category spans frontier model APIs (OpenAI, Anthropic, Mistral), inference platforms that serve open-weight models (Fireworks, Together AI, Groq), application builders and orchestration layers (Dify, Relevance AI, Gumloop), and vertical-specific platforms that embed AI into a workflow (Intercom for customer support, Clay for outbound sales).

Seventy-three companies in the corpus are tagged ai-platforms — the largest single product-segment category. The pricing diversity across this segment is the widest in the corpus: the same category contains companies billing per token, per seat, per credit, per action, and per resolution.

The five pricing archetypes in AI platforms

1. Pure-usage API (tokens or requests). Frontier labs and inference platforms bill per input/output token or per API call. Rate cards are public, metered at consumption, with free tiers and volume discounts. Examples: OpenAI ($2.50-$5/1M input), Anthropic ($3-$15/1M), Groq ($0.59-$0.89/1M), DeepSeek ($0.07/1M cached, $0.27 standard).

2. Freemium subscription + usage overages. A free tier with a monthly credit or request allotment, then paid plans with larger pools and overage rates. The most common structure in the category (85/158 corpus companies carry the freemium tag). Examples: Cartesia (Starter free + Creator $24 + credit overages), Gumloop (Free + Pro tiers with credit pool).

3. Seat-plus-credits hybrid. A per-user seat anchors predictable revenue; a credit or request pool meters AI-intensive work. Most common in AI coding tools and outbound platforms that embed within a team workflow. Examples: Cursor, GitHub Copilot, Codeium, Clay, Apollo.

4. Platform fee + usage (enterprise). A base platform fee (often sales-quoted) plus metered usage above a committed minimum. Used for high-volume or dedicated-deployment buyers. Examples: Anyscale, Baseten enterprise, Fireworks AI enterprise.

5. Outcome-based. You pay per resolved task, not per token or seat. Rare (2/158 in corpus) but growing: Intercom Fin ($0.99/resolution) and Rox (per Agent Action). Requires attributable, discrete task completion.

Pricing transparency in the category

127 of 158 corpus companies publish public pricing — an 80% transparency rate. The exceptions cluster in two places: (a) vertical enterprise-knowledge platforms (legal, enterprise search) that sell top-down into regulated buyers, and (b) billing-infrastructure vendors. Standard AI platforms are overwhelmingly public.

Go-to-market patterns

AI platform companies almost universally combine motions: 140/158 are sales-led, 109/158 use PLG, and 137/158 are self-serve. The dominant pattern is PLG-first (free tier + self-serve) with a sales overlay for enterprise commitments. Of 109 PLG companies, 105 publish public pricing — the PLG→public-pricing correlation is the strongest structural signal in the corpus.

The $200 prosumer ceiling

Six consumer-facing AI platforms have added a ~$200/month “Max” or “Ultra” tier above their $20 Pro tier: OpenAI (ChatGPT Pro), Anthropic (Claude Max), Perplexity (Max), You.com (Max), Cursor (Ultra), Google (AI Ultra). This tier monetises power users who would otherwise cap against lower plan limits, without raising the mass-market price.

What to watch

The AI platform market is consolidating at the tooling layer: 8 corpus companies were acquired in 18 months (Voyage→MongoDB, Jina→Elastic, OpenPipe→CoreWeave, OpenMeter→Kong, Togai→Zuora, Metronome→Stripe, Tavily→Nebius, Galileo→Cisco). Post-acquisition pricing has mostly held, but gating risk is real. If your AI platform vendor is a small specialist, model the acquisition scenario into procurement planning.