Product family
OpenAI, product by product
OpenAI's product surface mapped by capability: the judged four plus honest cards for images, Work, ChatKit, and the platform — and the lines that shut down, documented, not hidden.
4 of 10 lines judged in an arena · flagship: ChatGPT in AI Assistants
ChatGPT
JudgedAsk, research the web, and delegate whole tasks that come back as reviewable docs, decks, sites, and spreadsheets — one app across web, desktop, and mobile with projects, memory, and scheduled automations.
ChatGPT Work
Not yet judgedDefine an outcome and ChatGPT plans the work, browses in a cloud browser, uses your files and plugins, and returns a finished deliverable you review.
learn.chatgpt.com/docs/get-started-with-work ↗
A mode of ChatGPT (it replaced agent mode in July 2026), sold on the same plans — judged inside the chatgpt entry's long-running-work and agents stories; a second row would double-list one product.
Image generation (GPT-Image-2.5)
Not yet judgedGenerate and edit images in conversation — text-to-image, precise edits from reference images — and the same capability via the API (GPT-Image-2.5 Sunburst and Flare).
developers.openai.com/api/docs/guides/image-generation ↗
A capability of ChatGPT and the API, not a separately adopted product — and no image-generation arena exists yet (roadmap: planned). Its in-app stories are judged inside the chatgpt entry.
Codex
JudgedA coding agent that writes, tests, reviews, and ships code — CLI, IDE extension, cloud sandboxes, Slack/Linear/GitHub integrations, and Codex Remote from your phone.
Codex Security
Not yet judgedScan repositories for vulnerabilities from a plugin, CLI, SDK, or the cloud, with adversarially-checked findings your team triages.
A Codex sibling launched September 2026; security-scanners ranks established GA scanners — bring-up candidate once it has a record beyond launch week.
Agents SDK
JudgedBuild agent loops with handoffs, guardrails, and sessions in TypeScript or Python — the open framework track alongside the hosted Agents API.
Plugins
JudgedPackage skills, MCP servers, and UI into plugins that run in both ChatGPT and Codex, distributed through a reviewed public directory with checkout-based monetization.
ChatKit
Not yet judgedEmbed a production chat UI over your own agent backend: a widget with streaming, attachments, and theming, shipped from npm.
developers.openai.com/api/docs/guides/agents ↗
A developer component (beta API), not a product users adopt on its own; no embeddable-AI-UI arena exists.
Agents API
Not yet judgedRun agents on OpenAI-hosted infrastructure — the managed Codex harness with sessions and sandboxes, callable from your backend.
developers.openai.com/api/docs/guides/agents ↗
Beta hosted-agents surface of the API platform; proprietary model platforms aren't an arena we rank yet (roadmap: frontier-model-apis).
API Platform
Not yet judgedThe developer platform: Responses API, Realtime voice with SIP calls, hosted tools, containers, and fine-tuning.
developers.openai.com/api/docs ↗
Proprietary model APIs are not an arena we rank — inference-providers ranks hosts of open models, a different job (roadmap: frontier-model-apis).
A judged line competes in its arena on the same stories as every rival — family membership never affects scoring. Lines without a fitting arena stay unscored until one exists. See the methodology.