OpenAI Passes Anthropic in Gamma’s AI Spend, 40% to 38%

OpenAI took 40% of Gamma’s AI spend in the last full week of October, against 38% for Anthropic and 9% for Google, Gamma co-founder Jon Noronha wrote in the Gamma Model Index, a monthly post the company shared on X on 16 November. The presentation company says that on 29 October, for the first time since 2023, OpenAI surpassed Anthropic in its usage, after three A/B tests in which GPT-6.1 Terra replaced Claude Haiku 5.5 and Sonnet 5.5 for several key jobs.

Our number: in the data behind the index’s first chart, OpenAI’s share of Gamma’s trailing seven-day AI spend rose from 11.5% in the week ending 7 September to 43.7% in the week ending 6 November, 3.8 times, while Anthropic’s fell from 63.2% to 36.0%, a drop of 27.2 points, by our arithmetic. A 51.7-point Anthropic lead became a 7.7-point OpenAI lead in about two months.

Business Pill · Model Routing Explained

An evergreen Business Engineer explainer on model routing, sending each task to the right AI model. It teaches the general idea only and does not discuss Gamma’s index.

The key insight: As we read Gamma’s index, the lab with the larger share of one heavy buyer’s bill changed because one high-volume job, slide repair, changed hands: OpenAI’s share of Gamma’s spend went from 11.5% to 43.7% in about two months, by our arithmetic, while Anthropic kept 78% of the planning job.

What Gamma Published

Gamma says its agent “handles millions of sessions each week, and we A/B test every new model on real users”. The agent splits every deck into about a dozen jobs, and each one can run on a different model. Gamma says the agent made more than 70 million model calls last week, using more than 200 billion tokens.

Noronha wrote on X: “Last month, our spend on OpenAI surpassed Anthropic for the first time in years. But their lead is not secure.” Gamma co-founder Grant Lee wrote: “Since OpenAI now owns almost all of our repair tasks, it surpassed Anthropic in our usage leaderboard for the top spot.” Spend in the index is measured at list prices, across the language and image models in Gamma’s agent, the page says.

Each lab's share of Gamma's AI spend, language and image models, trailing seven days at list prices, for the s
Each lab’s share of Gamma’s AI spend, language and image models, trailing seven days at list prices, for the seven days ending 7 September to 6 November 2026, from the data behind Fig. 1 of the Gamma Model Index. One company’s spend.

Slide Repair and Images Moved the Bill

GPT-6.1 Terra cost about half as much as the larger models before it, according to Gamma. On slide repair, it halved the share of slides still failing Gamma’s checks after repair, from 9.1% to 4.7%. Repairing slides is 14% of the agent’s spend, and 97% of it goes to OpenAI, the index says. Images are another 14%, and GPT Image has gone from under 1% of Gamma’s images in January to about half.

Anthropic keeps the largest job. Planning and talking with users is about 42% of spend, and 78% of it goes to Anthropic; editing on request is another 9%, 71% of it to Anthropic. Noronha writes that “Anthropic’s models are better at juggling several requirements at once, and at inferring the ones a user never spelled out, where GPT models tend to be too literal”. Writing slides, 17% of spend, splits 61% OpenAI, 30% Google and 10% Anthropic.

Calls Versus Dollars

Counted by model calls, the picture flips, Gamma says. In the week of 27 October to 2 November, OpenAI’s GPT-6 Luna and GPT-5.6 Luna made 54.1% of the agent’s calls for 19.2% of its language-model spend, and Claude Haiku 5.5 made 22.6% of calls for 4.6% of spend, according to the data behind the index’s third chart, by our sums.

Claude Sonnet 5.5 made 9.8% of calls and took 36.1% of language-model spend, about 3.7 times its share of calls, by our arithmetic. GPT-6.1 Terra made 9.7% of calls and took 15.2% of spend. Gamma says OpenAI’s models wrote 2.6 to 4.6 times fewer output tokens than Claude’s for the same work on two jobs it measured side by side.

One Customer Against Two Cross-Sections

Two broader datasets we covered on 16 November point the other way on spend. In our Databricks AI Index piece, Anthropic took 70.1% of spend on proprietary models routed through Databricks’ Unity Gateway in the week of 2 November, by our sums, and OpenAI 27.4%, up from 10.1% in the week of 20 July. In our Ramp AI Index piece, Anthropic took 49.3% of token spending in the week ending 1 November and OpenAI 46.1%.

The measures differ. Gamma’s index is one application company’s whole AI bill, including Google and image models. Databricks counts proprietary-model spend across companies on its gateway; Ramp counts businesses using its AI Spend Management product. By our comparison, Gamma shows how fast one heavy buyer that tests every model can move its bill. In our piece on multi-provider use, 71.1% of companies active on the Databricks gateway had used two or more model providers.

What Gamma Says It Is Testing

Gamma says that since 28 October half of its free users’ decks have been planned by Terra and half by Sonnet, and “so far the results are close”. If Terra wins, the page says, OpenAI takes a share of Anthropic’s biggest job. Gamma also says its research team has trained a slide-repair model that fixed 76% of failing slides on the first try, against 57% for the general model it was tested against.

The Structural Read

As we read the job table, Gamma does not buy a lab; it buys about a dozen jobs, and each is re-picked when a new model arrives, which is how a single release such as GPT-6.1 Terra can move several points of the bill in a week.

As we read the call data, volume and money sit in different places: Claude Sonnet 5.5 made 9.8% of calls and took 36.1% of language-model spend, about 3.7 times its call share, by our arithmetic.

Grant Lee, co-founder, Gamma, 16 November 2026

“Since OpenAI now owns almost all of our repair tasks, it surpassed Anthropic in our usage leaderboard for the top spot.”

Three Implications

FOR AI LABS At Gamma, repairing slides is 14% of the agent’s spend and 97% of it goes to OpenAI, according to the index.

FOR ANTHROPIC Planning and talking with users is about 42% of Gamma’s spend, and 78% of it goes to Anthropic, the index says.

FOR APPLICATION BUILDERS Gamma says its agent splits every deck into about a dozen jobs, and each one can run on a different model.

The Business Engineer Lens

This story maps onto the Business Engineer framework The Agentic Harness War.

The framework says: “Foundation models become relatively commoditized as orchestration layers capture differentiation.”

As we read Gamma’s index, multi-model routing makes provider share fluid: an application that re-picks the model for each job moved its spend from Anthropic to OpenAI in about two months, while keeping Claude on planning; the framework predates the index and does not discuss Gamma.

What Is Not Established

The index covers one company’s spend, measured at list prices, so it shows shares of Gamma’s bill, not what Gamma pays after any discounts or the size of the bill. The weekly series values come from the data behind the page’s first chart; the windows overlap, since each is a trailing seven days.

The comparison with Databricks and Ramp is ours; the three sources use different samples, scopes and measures. Gamma’s statements about its tests and its own models are the company’s account.

This Week In The AI Supercycle

AI Lost Its Central Banker, the Lab Oligopoly, and the Future of Open Source

This week’s issue: the Nvidia leadership change, read as the AI build-out losing its central banker: who now finances the GPU landlords, backstops the data centres and sponsors open models, and why two labs are becoming the bank.

Read the full issue →

Business Engineer Framework

The Agentic Harness War

The Business Engineer analysis of why the harness around the model, not the model itself, is where value pools in the AI stack.

Read the Map of AI →

The Bottom Line

Gamma’s index puts OpenAI at 40% of its AI spend and Anthropic at 38% in the last full week of October, after GPT-6.1 Terra took slide repair. In the chart data, OpenAI’s share went from 11.5% to 43.7% between early September and 6 November, by our reading, while Anthropic still holds 78% of the planning job, Gamma’s largest.

95,000+ executives read Business Engineer for the AI strategy frameworks cited by ChatGPT, Claude, and Perplexity.

A note on sourcing. The figures come from the Gamma Model Index on research.gamma.app, written by Gamma co-founder Jon Noronha, including the data embedded behind its charts, and from the X posts by Gamma co-founders Jon Noronha and Grant Lee. We found the story through a relay post on X by @guilleflorvs and checked every figure against Gamma’s page. The Databricks and Ramp figures come from our earlier pieces on those indexes.

Multiples, point changes, the summed call and spend shares and the comparison across the three sources are our arithmetic. We did not contact Gamma for this piece.

Sources: Gamma Research: The Gamma Model Index, by Jon Noronha (Nov 2026) · X: Jon Noronha, Gamma co-founder (16 Nov 2026, 18:03 UTC) · X: Grant Lee, Gamma co-founder (16 Nov 2026, 18:09 UTC)

Scroll to Top
FourWeekMBA