Irys (formerly Iqidis) vs ChatGPTA chatbot sees the current prompt. A matter platform preserves the work.

It is built by OpenAI, it improves quickly, and it is genuinely useful for an enormous range of work, including plenty of tasks inside a law office.

Its business tiers are real workplace products, with administrative and security features and connectors described on OpenAI's business pages.

If the goal is one AI assistant for the whole firm, marketing, operations, finance, and administration included, a general assistant is a sensible default.

Standard business-tier pricing could not be verified on official OpenAI pages at the time of review; confirm current per-seat pricing with OpenAI.

Irys One is publicly listed at $299 per month, with no per-token or per-query overage on the advertised plan. Irys Teams is $329 per seat per month for 2 to 25 attorneys. There is a 7-day trial with no credit card.

On sticker price, the general assistant is cheaper. That is true and worth saying plainly. The real comparison is what each subscription buys for legal work specifically.

OpenAI builds the frontier models itself, and ChatGPT offers them directly.

Breadth. A general assistant is useful across departments, not just legal. If a firm wants one tool for everything from client communications to spreadsheets, this is that tool.

Price for general use. At $20 per month for the individual paid tier, it is an inexpensive way to put strong AI in front of a broad workforce.

Ecosystem and familiarity. A very large user base, extensive documentation, and business-tier connectors into common workplace systems.

The next task picks up where the last one ended, so there is no need to rebuild the matter in every new prompt.

Output is source-linked, and Cite Check reviews the authorities behind it. A lawyer can trace material statements back to the underlying source instead of taking a chat answer on faith.

Research and drafting happen inside the same matter workspace, and a Word add-in brings that work into the document where lawyers actually finish it. Built-in e-signature covers execution without another subscription.

Workbench, currently in beta, connects to applications such as Gmail, Google Drive, Calendar, Sheets, Outlook, SharePoint, OneDrive, and Slack, and can prepare documents and email drafts across them. Workbench requires user approval rather than autonomously sending communications or changing files.

The infrastructure around the models is Irys's own. Irys runs its own agentic orchestration, structured matter intelligence, and evaluation infrastructure around the models it uses, built to plan and carry multi-step legal work. In Irys's internal evaluations on published legal benchmark tasks (Irys internal testing, July 2026), the Irys system scored above leading frontier-model baselines on legal work. Methodology available on request.

None of this is an argument that general models are weak. The difference is the legal structure built around them: matter state, provenance, citation review, and finished work product. As of the review date we found no public documentation of a built-in legal citator for ChatGPT.

Take one real matter and run the same work through both tools, then answer these questions.

That is a reasonable stack, not a contradiction.

Choose ChatGPT when the priority is one broadly capable assistant for the whole organization at the lowest entry price. Choose Irys when the priority is legal work product: matters, citations, drafting, and finished files.

Except where a statement is expressly labeled as Irys internal testing, comparisons on this page are based on publicly available information reviewed August 3, 2026. Pricing, plans, and features change. Confirm current terms with each vendor before you buy.