OpenAI launches GPT‑5.3 Instant

OpenAI has paired a model update with a pricing document that gives users a clearer view of how one part of its coding stack is billed. In Help Center release notes, the company said GPT‑5.3 Instant was updated on March 3, 2026 to deliver more accurate answers, richer and better-contextualized web-search results, and fewer “dead ends, caveats, and overly declarative phrasing” that interrupt conversation flow. (help.openai.com) In a separate Help Center article updated on May 22, OpenAI published a Codex rate card covering Plus, Pro, Business, Enterprise, Edu, Health and Gov plans. (help.openai.com) ### What did OpenAI say changed in GPT‑5.3 Instant? (help.openai.com) OpenAI’s model release notes said the March 3 update focused on “tone, relevance, and conversational flow.” The company said GPT‑5.3 Instant now gives more accurate answers, improves how web-search results are contextualized, and cuts back on unnecessary caveats and interruptions. A second entry dated March 16 said OpenAI also rolled out another GPT‑5.3 Instant update aimed at follow-up tone. That change, the company said, reduced teaser-style phrasing in responses, including prompts such as “If you want…” and similar constructions. (help.openai.com) ### Where does GPT‑5.3 Instant sit in OpenAI’s current lineup? OpenAI said on March 11 that GPT‑5.1 Instant, GPT‑5.1 Thinking and GPT‑5.1 Pro were retired from ChatGPT. Existing conversations using GPT‑5.1, the company said, automatically continued on GPT‑5.3 Instant, GPT‑5.4 Thinking or GPT‑5.4 Pro. (help.openai.com) ChatGPT release notes also show OpenAI continuing to adjust adjacent products around that model family. On April 9, the company said GPT‑5.3 Instant Mini replaced GPT‑5 Instant Mini as the fallback model after users hit GPT‑5.3 Instant rate limits. (help.openai.com) ### What does the Codex rate card actually price? OpenAI’s Codex rate card says usage is now priced on API token consumption rather than average per-message estimates. The article says billing is calculated in credits per million input tokens, cached input tokens and output tokens. (help.openai.com) The table lists GPT‑5.3‑Codex at 43.75 credits per million input tokens, 4.375 credits for cached input tokens and 350 credits per million output tokens. The same table lists GPT‑5.5 at 125 input credits and 750 output credits, while GPT‑5.4 is priced at 62.50 input credits and 375 output credits. (help.openai.com) OpenAI also said code review uses GPT‑5.3‑Codex, and that GPT‑5.3‑Codex‑Spark may be available as a research preview with rates that “are not final.” The company said fast mode consumes credits at a higher rate for supported models. ### Which customers are covered by the pricing change? (help.openai.com) The Codex rate card says the token-based structure applies to new and existing ChatGPT Plus and Pro customers, new and existing ChatGPT Business customers, and new and existing Enterprise, Edu, Gov, Health and ChatGPT for Teachers customers. A small subset of Enterprise customers, OpenAI said, remains on a legacy rate card. (help.openai.com) The same article says OpenAI updated Codex pricing on April 2, 2026 for new and existing Plus, Pro, ChatGPT Business and new ChatGPT Enterprise plans. On April 23, it extended that update to existing Enterprise plans, including Edu, Health, Gov and ChatGPT for Teachers. (help.openai.com) ### Where can users see what they are spending next? OpenAI said users can monitor usage limits and remaining credit in Codex settings through the Usage panel. The company also said average Codex spending runs about $100 to $200 per developer per month, though usage varies by model, number of running instances, automations and use of fast mode. (help.openai.com) ChatGPT release notes published in April added that OpenAI was also changing Plus and Pro plan options around Codex usage, including a $100-per-month Pro tier and temporary promotional usage terms that run through May 31. Those plan notes direct users to settings and pricing pages for upgrades or changes. (help.openai.com 1) (help.openai.com 2)