Pricing Transparency Rising

OpenAI and Anthropic are publishing more detailed price sheets for coding and application workloads, giving buyers a clearer way to estimate artificial intelligence spend. (help.openai.com) (platform.claude.com) OpenAI’s Codex rate card was updated within the past day and says that, as of April 2, 2026, pricing moved from per-message charges to token-based pricing for new and existing ChatGPT Plus, Pro, and Business customers, plus new Enterprise customers. The company says customers on existing Enterprise plans still use a legacy rate card until migration in the coming weeks. (help.openai.com) That OpenAI table breaks usage into input, cached input, and output tokens, with different credit charges for each. It lists GPT-5.4 at 62.50 credits per 1 million input tokens and 375 credits per 1 million output tokens, and says Fast mode uses twice as many credits. (help.openai.com) Anthropic’s Claude API pricing page uses the same kind of itemized format, but in dollars instead of credits. Its current table lists Claude Opus 4.7 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 1) (platform.claude.com 2) Anthropic also prices prompt caching separately, which is the practice of reusing repeated context instead of paying full price each time. Its pricing page lists cache hits and refreshes for Claude Opus 4.7 at $0.50 per million tokens, compared with $5 per million for standard input. (platform.claude.com) The practical change is that buyers can now model costs by workload shape instead of rough message counts. OpenAI says the new format replaces average per-message estimates with a direct mapping between token usage and credits, and Anthropic publishes the same breakdown across its model lineup. (help.openai.com) (platform.claude.com) That is useful in coding tools because the bill depends on what the system does: reading a large codebase, reusing cached context, or generating long patches all consume different token categories. OpenAI says actual Codex credit usage depends on the mix of input, cached input, and output tokens in each task. (help.openai.com 1) (help.openai.com 2) OpenAI is also tying those prices more directly to ChatGPT plan economics. Its help page says Codex is included with ChatGPT Plus, Pro, Business, and Enterprise or Edu plans, and estimates average Codex cost at about $100 to $200 per developer per month, with wide variation by model, automation volume, and Fast mode use. (help.openai.com 1) (help.openai.com 2) Anthropic’s pricing page goes a step further on deployment geography, noting that Google Vertex AI regional and multi-region endpoints carry a 10% premium over global endpoints for Claude 4.5 models and later. That gives procurement teams another line item to compare when they weigh residency requirements against cost. (platform.claude.com) The short version is that artificial intelligence vendors are starting to show their math in public tables. As coding agents move from pilot projects into contracted software budgets, those tables give finance teams something more concrete than “usage may vary.” (help.openai.com) (platform.claude.com)