Explore / Build the product / AI agents

AI agents · How Founders Use AI

How much does it cost to run AI features, and how do I keep costs sane?

You pay per token (roughly per word) in and out, so cost scales with usage; a small feature can run on tens of dollars a month while a chat-heavy product can eat a large share of revenue. Tiers span a hundredfold: GPT-6 Astra and Claude Fable 5.1 list at $10 in and $50 out per million tokens, GPT-6 Luna and Claude Haiku 5.5 at $0.10 and $0.50. Prompt caching, routing easy requests to cheap models and batching non-urgent work at half price often cut bills by more than half. Track cost per user per feature from day one so pricing follows data, not fear.

On the AI map, 29 October 2026 Frontier modelsOpen-weight modelsDecision models

Go deeper22 sources

  • API pricing (OpenAI docs)

    OpenAI · Tool · Oct 2026

    Current per-token prices for every GPT-6 tier, with cached input and batch rates side by side.

  • Introducing Claude Haiku 5.5

    Anthropic · Article · Oct 2026

    Anthropic's cheap, fast tier for classification and routing, about 75% cheaper to run than Haiku 4.5, plus a cut to cache-read prices.

  • Introducing GPT-6.1 Sol

    OpenAI · Article · Sep 2026

    OpenAI says it nearly matches GPT-6 Astra at a fifth of the price, with cached input at $0.10 per million tokens.

  • Welcome to LLMflation: LLM inference cost is going down fast

    Guido Appenzeller (a16z) · Article · Nov 2024

    The definitive chart: equivalent-quality inference gets 10x cheaper per year, which changes what you can afford to build.

  • a16z on LLMflation (thread)

    a16z · Thread · Nov 2024

    The five-minute version of the cost-decline argument, useful for convincing a worried co-founder.

  • Prompt caching (Claude docs)

    Anthropic · Guide · 2025

    Cached reads cost 10% of normal input price; the single highest-leverage cost lever, from the primary source.

  • Batch API (OpenAI docs)

    OpenAI · Guide · 2025

    A flat 50% discount for anything that can wait up to 24 hours, like digests, enrichment, and backfills.

All 22 sources