Coding agents pivot to pricing war

OpenAI, Cursor and Windsurf have all changed how they sell coding agents in recent weeks, shifting competition toward seat design, token billing and team controls rather than headline model claims. OpenAI said on April 2 that ChatGPT Business and Enterprise customers can now add Codex-only seats with pay-as-you-go pricing instead of a fixed seat fee. Cursor said on May 18 that Composer 2.5 is now available in Cursor, with separate “Standard” and “Fast” token-priced options. Windsurf said on March 18 that it replaced its credit-based system with new pricing plans and set Teams at $40 per user per month. ### Why are pricing changes getting as much attention as model updates? Windsurf said on March 18 that its new plans were designed to replace credits with “industry-standard quotas,” a change that put the company’s packaging closer to rivals selling predictable team subscriptions. Windsurf’s documentation says Teams and Enterprise plans include features such as centralized billing, admin dashboards, single sign-on and role-based access control. (openai.com) OpenAI said on April 2 that Codex-only seats give ChatGPT Business and Enterprise workspaces “full access to Codex without a fixed seat fee.” OpenAI’s help documentation says Business workspaces can mix standard ChatGPT seats and usage-based Codex seats, and Enterprise workspaces received the same seat type update in April. ### What did Cursor actually ship with Composer 2.5? Cursor said on May 18 that Composer 2.5 is “a substantial improvement in intelligence and behavior over Composer 2.” The company said the update is better at long-running tasks, follows complex instructions more reliably and is “more pleasant to collaborate with.” Cursor’s changelog lists pricing for Composer 2.5 at $0.50 per million input tokens and $2.50 per million output tokens for Standard, and $3.00 and $15.00 respectively for Fast. (openai.com) (windsurf.com) Cursor had already been moving toward more explicit usage pricing. Cursor’s March 19 Composer 2 changelog listed separate Standard and Fast token prices, lower than the new Fast tier attached to Composer 2.5. That pricing structure, combined with the May 18 release, shows Cursor tying product updates directly to usage economics. That is an inference from Cursor’s published changelogs and pricing tables. ### What is OpenAI trying to do with Codex-only seats? (cursor.com) OpenAI said the new Codex seat lets “small groups” start pilots inside ChatGPT Business and Enterprise and expand later. OpenAI’s help center says the seat type is codex-only access based on flexible pricing, and that Codex usage is billed through a rate card aligned to token-based pricing. ChatGPT Business also cut the price of subscription-based standard seats by $5 per month as part of the April 2 update, according to OpenAI’s help documentation. (cursor.com) That combination — lower standard seat pricing and a usage-based coding seat — gives buyers two different cost structures inside one workspace. ### Where does Google fit into the same fight? (openai.com) Google said on April 20 that AI Pro and Ultra subscribers now get increased usage limits in Google AI Studio. In separate posts, Google said it added project spend caps, updated usage tiers and launched Flex and Priority inference options in the Gemini API to give developers more control over cost, rate limits, latency and reliability. (help.openai.com) Those Google changes were not framed as a coding-agent launch, but they address the same buyer concerns: how much usage is available, what happens when demand spikes and how billing is controlled. That is an inference from Google’s descriptions of spend caps, usage tiers and inference options. ### What should buyers watch next? May 18 is the latest dated product milestone among the three companies, with Cursor’s Composer 2.5 release now live in the product changelog. (blog.google) OpenAI’s April 2 seat changes are already reflected in Business and Enterprise help pages, and Windsurf’s March 18 pricing reset remains in effect on its pricing materials. The next concrete comparison point will be whether vendors change seat prices, token rates or team controls again as enterprise customers expand pilots into broader deployments. (blog.google) (cursor.com)