OpenAI pricing and channel tensions
OpenAI’s internal memo flagged distribution tensions—praising an Amazon alliance while saying Microsoft has limited its ability to reach customers—just as the firm and rivals make pricing structures explicit. OpenAI published which plans include Codex and released usage rate cards, and Anthropic likewise set out model and feature pricing, signaling a shift toward metered enterprise buying. (cnbc.com) (help.openai.com) (platform.claude.com)
OpenAI is making its pricing more explicit as it pushes beyond Microsoft’s sales channel and deeper into Amazon’s cloud marketplace. (cnbc.com) In a memo viewed by CNBC and published April 13, 2026, OpenAI revenue chief Denise Dresser said Amazon Web Services would be a key route for enterprise growth and said Microsoft had “limited our ability” to meet customers where they buy artificial intelligence services. CNBC said the memo came less than two months after Amazon announced plans to invest up to $50 billion in OpenAI. (cnbc.com) At the same time, OpenAI updated how it sells Codex, its coding agent. OpenAI’s help center says Codex is included with ChatGPT Plus, Pro, Business, and Enterprise and Edu plans, with temporary access on Free and Go, and says the tool can edit files, run commands, and execute tests across code repositories. (help.openai.com) The bigger change landed on April 2, 2026, when OpenAI shifted Codex billing from per-message charges to token-based usage for Plus, Pro, ChatGPT Business, and new ChatGPT Enterprise customers. OpenAI said existing Enterprise customers would stay on a legacy per-message style rate card until they are migrated in coming weeks. (help.openai.com) OpenAI also split workplace access into two seat types for ChatGPT Business and ChatGPT Enterprise: a standard ChatGPT seat and a Codex-only seat. OpenAI says the Codex-only seat is usage-based with no minimum, while standard Business seats remain fixed-price and now cost $25 per user monthly or $20 per user monthly on annual billing in most countries. (help.openai.com 1) (help.openai.com 2) Anthropic is making a similar pitch to buyers who want a menu, not a bundle. Its pricing page lists token rates by model, including Claude Opus 4.6 at $5 per million input tokens and $25 per million output tokens, Claude Sonnet 4.6 at $3 and $15, and Claude Haiku 4.5 at $1 and $5. (platform.claude.com) Anthropic also breaks out extra charges for prompt caching and regional routing, and says Claude is sold directly and through Amazon Web Services Bedrock, Google Vertex Artificial Intelligence, and Microsoft Foundry. That gives enterprise customers a way to compare not just model quality, but where they want billing, compliance, and infrastructure to sit. (platform.claude.com 1) (platform.claude.com 2) OpenAI’s own consumer tiers now read more like a usage ladder than a flat subscription. Its Pro documentation says Plus costs $20, one Pro tier costs $100, and the top Pro tier stays at $200, with higher Codex and research allowances at each step. (help.openai.com) The result is a market where model makers are no longer hiding the meter. OpenAI’s memo points to a fight over who sells the models, and its new rate cards show how those models will increasingly be bought. (cnbc.com) (help.openai.com)
