The Question Changed This Month
In August 2026, five of the twenty-one companies that booked a discovery call with us said an AI assistant had recommended OneWave. ChatGPT, Claude, and Grok were all named. A year ago, Grok did not come up on sales calls at all. Now it is the fourth platform on the shortlist, and the question we hear is no longer "Claude or ChatGPT?" but "which one do we buy seats for?"
That is a different question than "which model scores highest." We already published a five-platform capability comparison and a deeper Claude vs ChatGPT breakdown. This post is about the purchasing decision: seats, admin controls, what happens to your data, what the platform plugs into, and the weakness the vendor will not put on the pricing page.
Our position up front. We built our practice on Claude starting in 2024 and we are a Claude Partner Network member. We are also an official OpenAI partner and deploy ChatGPT for clients every month. We have no commercial relationship with xAI or Google. Everything below is as of August 24, 2026.
Buying an AI platform for a team is a procurement decision, not a benchmark decision. The model matters less than who can see the data, who administers the seats, and what the tool connects to.
Grok (xAI)
Best at: price per token and live X data
xAI shipped Grok 4.6 on August 12 with a 500K-token context window at $2 per million input tokens and $6 per million output. Grok 4.1 Fast, the budget tier, runs at $0.20 and $0.50. For high-volume API workloads, that is the cheapest frontier-class pricing on this list. Grok is also the only platform with native search over X, which matters if your business monitors public sentiment, breaking news, or competitor announcements in real time.
On August 11, xAI launched Grok Bot, a beta of always-on agents that each get a cloud computer, sign into your tools with your credentials, and run multi-step jobs unattended. It is the most aggressive agent launch of the summer.
Team plan and data policy
Grok Business is $30 per seat monthly, $25 on annual billing, and includes team management, SOC 2, and a contractual commitment not to train on your data. Grok Enterprise adds SSO, SCIM, and custom retention at negotiated pricing. The important caveat is that free, X Premium, and individual SuperGrok accounts may be used for training. If employees are using Grok on personal accounts today, your data is already in scope.
The honest weakness
The business tooling is young. Grok Bot access is bundled through SuperGrok Heavy at $300 a month and through Cursor Ultra and Cursor Teams Premium tiers, with enterprise access on a waitlist. There is no spend cap on the Bot usage meter yet, all of a user's Bots share one security boundary, and the connector library is thin next to the other three. Grok is a strong API vendor and a compelling research tool. It is not yet a platform we would roll out to a fifty-person team as the default assistant.
ChatGPT (OpenAI)
Best at: breadth and employee familiarity
The GPT-5.6 family reached general availability on July 9, 2026 in three tiers: Sol, Terra, and Luna. Every ChatGPT Business seat gets all three, while Free and Go accounts stay on GPT-5.5 Instant. ChatGPT remains the tool most of your employees have already used at home, which shortens training time more than any feature does. Voice, image generation, Custom GPTs, and the largest third-party integration catalog are all still OpenAI advantages.
Team plan and data policy
ChatGPT Business is $20 per seat on annual billing or $25 monthly, with a two-seat minimum, after OpenAI cut the price in April 2026. Business and Enterprise do not train on your content by default and include SAML SSO, SOC 2 Type 2, and ISO 27001. Free, Go, Plus, and Pro accounts train by default with an opt-out. That split is the same governance trap as Grok: the consumer plan your staff already pay for is the one with the weaker policy.
The honest weakness
Predictability. Model names, defaults, and rate limits change often, and Enterprise pricing is quote-only. For workflow automation that has to behave the same way every day, we still see more drift on GPT-5.6 than on Claude.
Gemini (Google)
Best at: living inside Google Workspace
Google released Gemini 3.1 Pro in February 2026 and it now runs across Gmail, Docs, Sheets, Slides, and Meet with Flash tiers handling the high-volume work. If your company already runs on Workspace, Gemini reads your email, calendar, and Drive without a connector, a permission review, or a second vendor agreement. No other platform on this list can say that.
Team plan and data policy
You do not buy Gemini seats. You buy Workspace: Business Starter at $7, Standard at $14, and Plus at $22 per user per month on annual billing, with full Gemini across the apps from Standard up and an AI Expanded Access add-on for heavier users. Google commits under its Workspace data processing terms that customer content is not used to train Gemini. Admin, SSO, and audit logging are the Workspace controls you already have.
The honest weakness
Outside of Workspace, Gemini is the least consistent of the four at following precise instructions, and Google's agentic coding tooling trails Claude Code and Codex. We wrote a full verdict in Gemini in Google Workspace: is it worth it? The short version: excellent if you are all-in on Google, unremarkable if you are not.
Claude (Anthropic)
Best at: coding, long-horizon agents, and reliability
Anthropic released Claude Fable 5 on June 9, 2026 as its flagship, then filled in the tiers below it with Sonnet 5 on June 30 at $2 per million input tokens and $10 output, and Opus 5 on July 24 at $5 and $25, with Haiku 4.5 as the budget tier. Fable 5 leads Cognition's FrontierCode evaluation and is the model we hand the longest, highest-stakes agent runs to; Opus 5 covers most of that ground at half the price. Claude Code is the strongest agentic coding tool on the market, and Claude Cowork brings the same file-and-tool access to non-engineers on the desktop. We have published 204 open-source Claude skills because that ecosystem is where our own work happens.
Team plan and data policy
Claude Team is $20 per standard seat on annual billing, $25 monthly, for two to 150 people, with a $100 to $125 Premium seat for heavy Claude Code users. Team includes Claude Code, Cowork, SSO, connector admin controls, and Microsoft 365 connectivity. Enterprise is $20 per seat plus usage and adds SCIM, audit logs, and a compliance API. Under Anthropic's commercial terms your content is never used for training and there is no toggle to get wrong.
The honest weakness
Cost predictability at the top end. Fable 5 is priced at $10 per million input tokens and $50 output, and heavy Fable use on Team and Enterprise plans draws down usage credits. Claude also does not generate images, and fewer employees arrive already knowing it. Our guide to Claude Chat, Cowork, and Code covers which surface each role actually needs so you are not buying Premium seats for people who need standard ones.
Update, September 2026: Anthropic released Claude Fable 5.1 and Claude Mythos 5.1 on September 1, 2026. Fable 5.1 replaces Fable 5 as the generally available flagship on Team and Enterprise plans. Nothing about the seat pricing or data policy above changed.
The Decision Table
| Platform | Team price per seat | Trains on team data | Admin depth | Best fit |
|---|---|---|---|---|
| Grok Business | $30 monthly, $25 annual | No on Business; yes on consumer plans | SOC 2; SSO and SCIM on Enterprise only | Cheap API volume, live X monitoring |
| ChatGPT Business | $25 monthly, $20 annual | No on Business; yes by default on consumer plans | SAML SSO, SOC 2, ISO 27001 | Broad rollouts, Microsoft shops, employee familiarity |
| Gemini via Workspace | $14 to $22 annual, Gemini included | No, under Workspace data terms | Inherits Workspace admin and audit | Companies already running on Google Workspace |
| Claude Team | $25 monthly, $20 annual; Premium $100 to $125 | No, on every paid business plan | SSO and connector controls; SCIM and audit on Enterprise | Engineering, agents, regulated document work |
What We Recommend by Company Type
- Google Workspace shop. Start with the Gemini you already pay for on Business Standard, and add Claude Team for the people who build, analyze contracts, or run agents. Do not buy a third platform until those two are saturated.
- Microsoft shop. ChatGPT Business is the fastest team-wide rollout because adoption is nearly free, and Claude connects to Microsoft 365 for the operations and engineering work that needs consistency. Copilot alone rarely satisfies either group.
- Engineering-heavy company. Claude Team with Premium seats for engineers and standard seats for everyone else. Claude Code is the reason. Add Grok API access if you have a high-volume, cost-sensitive inference workload where the cheaper token price compounds.
- Regulated industry. Claude Enterprise or ChatGPT Enterprise, chosen on the DPA, audit log, and retention terms your compliance team signs off on. Rule out any consumer plan on day one, and treat Grok Bot as a pilot with its own risk review, not a team deployment.
How We Would Run the Purchase
Inventory what people are already using on personal accounts. In every audit we have run this year, that list includes at least one consumer plan that trains on inputs. Move those users to a business plan first. Then pick one default platform for the whole company and one specialist platform for the teams that need it. Two platforms is normal. Four is a support burden with no upside.
If you want the decision made with your data, your stack, and your compliance constraints in the room, that is the first session of our team AI training.
Sources
- Models and Pricing - xAI Developer Docs
- xAI Launches Grok Bot, Always-On AI Teammates With Their Own Cloud Computers - Unite.AI
- Grok Pricing in 2026: Plans, Free Tier, and Business Seats - GeoToolbox
- GPT-5.6: Frontier intelligence that scales with your ambition - OpenAI
- ChatGPT pricing in 2026: every plan and what it costs - CloudZero
- Google announces Gemini 3.1 Pro - 9to5Google
- Google Workspace Plans and Pricing - Google
- Generative AI Security, Compliance and Privacy - Google Workspace
- Claude Fable 5 and Claude Mythos 5 - Anthropic
- Introducing Claude Sonnet 5 - Anthropic
- Introducing Claude Opus 5 - Anthropic
- Plans and Pricing - Claude by Anthropic
- Is my data used for model training? - Anthropic Privacy Center