# OpenAI’s top researchers now burn $23,000 in AI tokens a day

> Source: <https://cryptobriefing.com/openai-researchers-23000-daily-token-usage/>
> Published: 2026-11-19 00:06:06+00:00

**Recommendation:** Budget for coding-agent token spend at the 90th percentile, not the median: OpenAI's heaviest researchers burned $23,000 a day at list prices, 105% more than in September, versus $2,400 at the median.
More in the playbook "Controlling LLM cost": <https://wpnews.pro/playbooks/llm-cost>

# OpenAI’s top researchers now burn $23,000 in AI tokens a day

OpenAI official logo (public domain, Wikimedia Commons) — CryptoBriefing brand treatment

An internal OpenAI report shows heavy-use researchers more than doubled their coding agent consumption since September

The most agent-hungry researchers at [OpenAI](https://cryptobriefing.com/markets/openai/) now run through about **$23,000 worth of tokens per day**. That figure comes from an internal report dated November 18, 2026, and it is up 105% from September.

The report, titled “Research acceleration: The jagged frontier,” tracks how deeply coding agents have embedded themselves in OpenAI’s own research staff.

## The numbers behind the surge

The headline metric covers the 90th percentile of OpenAI researchers. These are the heaviest users, but not outliers at the extreme edge.

Their average daily consumption hit $23,000 in tokens, valued at API list prices. Back in September, the same group was at $11,000 per day.

The “API list prices” qualifier matters. The dollar amounts reflect what this usage would cost an outside customer paying the public rate.

The middle of the pack is moving too, just less dramatically. Median researcher usage climbed to $2,400 per day, a 62% jump from $1,500 in September.

Perhaps the most striking figure is about parallelism. The report puts the average at **5.7 agent workdays per researcher per calendar day**.

### AI, tech, and the markets they move—in one daily briefing.

Daily. Free. Join 34,000+ readers across crypto, finance, and policy.

## What the agents are actually doing

According to the findings, agents are now completing almost half of research tasks that would typically take a human two to four hours.

The researchers themselves appear to value the tools enormously. According to the report, staff said they would give up 30% of their team’s compute, or 40% of their team’s headcount, to keep access to frontier agents for three months.

## A trend that has been building all year

This is not OpenAI’s first look at the phenomenon. A September 2026 report found that output tokens grew 124-fold between November 2025 and August 2026.

The pattern also tells a story about adoption. Growth at the 90th percentile outpaced growth at the median by a wide margin, 105% versus 62%.

## What this means for the AI industry

The most immediate implication is about how AI labs build AI. OpenAI is using its own agents to accelerate its own research, and the internal usage curve is steep.

The headcount trade-off finding deserves attention from anyone watching the labor side of AI. When researchers say they would accept 40% fewer teammates to keep their agents, it reframes how organizations might think about staffing versus tooling.

That does not mean OpenAI plans to cut staff. The report frames it as a hypothetical preference, and nothing in the findings indicates a hiring change.

They have privileged access to the newest models, deep expertise in prompting them, and no meaningful cost barrier to running them. The average enterprise developer paying list price faces a very different calculation.

If $23,000 a day at list prices is what peak productivity looks like, the economics for outside customers depend heavily on how token prices move from here.

**Disclosure:** This article was edited by Diego Almada Lopez. For more information on how we create and review content, see our [Editorial Policy](https://cryptobriefing.com/editorial-policy/).
