OpenAI researchers might be spending >$4m/day on tokens (at API prices)

Based on this quote from the GPT6 announcement: “Valued at API prices, daily token usage has exceeded $600 for the median researcher and $7,000 for researchers at the 90th percentile”.

Prompted by: Abtin Forouzandeh (full log)

Estimated slop ratioWhere 0% is full human with absolutely zero AI involvement and 100% is the cause of great dismay.: 80%

September 22, 2026

HN discussion

An estimate for 1,125 researchers, assuming 25% of 4,500 employees are researchers. This values token usage at API prices; it does not estimate OpenAI’s internal costs.

1Assumptions

OpenAI reports daily token usage above $600 for the median researcher and above $7,000 at the 90th percentile, valued at API prices.[1] The Financial Times reported about 4,500 employees in March 2026, as relayed by Reuters.[2]

Assume 25% are researchers, and their daily usage follows a lognormal distribution. Treat $600 and $7,000 as exact quantiles for this calculation. Neither the 25% share nor the distribution is reported by OpenAI.

N=4,500×0.25=1,125(1)

2Calculation

Let X be a researcher’s daily usage value, measured in USD. For a lognormal distribution:[3]

ln(X)∼N(μ,σ2)(2)
μ=ln(600)≈6.396930σ=ln(7000/600)Φ−1(0.90)≈1.917001(3)

Here, Φ⁻¹ is the inverse standard normal cumulative distribution; Φ⁻¹(0.90) ≈ 1.281552.

The mean—not the median—is the quantity to multiply by headcount:

E[X]=eμ+σ2/2≈3,768.29(4)
T=N·E[X]≈1,125×3,768.29≈4,239,325USD/day(5)

The estimate excludes non-researchers’ usage and training compute. It is an expected daily value under the assumptions, not a measured company total.

3Usage and concentration

Cumulative distribution of daily researcher usageA fitted lognormal distribution. Half of researchers are below $600 per day; 90 percent are below $7,000 per day. All figures are valued at API prices.Researchers at or below this amount0%25%50%75%100%$1$10$100$1k$10k$100k$1MMedian: $600P90: $7,000USD per researcher per day · logarithmic scale+3σ in log usage: $188,703.63/day (99.865th percentile)+3σ: $188k
Cumulative distribution of daily researcher usageA fitted lognormal distribution. Half of researchers are below $600 per day; 90 percent are below $7,000 per day. All figures are valued at API prices.Researchers at or below this amount0%25%50%75%100%$1$100$10k$1MMedian: $600P90: $7,000USD per researcher per day · logarithmic scale+3σ in log usage: $188,703.63/day (99.865th percentile)+3σ: $188k
Figure 1. Daily usage per researcher. The fitted curve passes through the assumed median ($600) and 90th percentile ($7,000). The mean is $3,768, or 6.28 times the median.
Normal versus power usersThe bottom 90 percent of researchers account for 26.3 percent of daily value. The top 10 percent account for 73.7 percent. Within-group mean daily values are $1,099 and $27,789 respectively.Normal: bottom 90%Power: top 10%Share of researchers90%10%Share of daily usage value26.3%73.7%
Normal versus power usersThe bottom 90 percent of researchers account for 26.3 percent of daily value. The top 10 percent account for 73.7 percent. Within-group mean daily values are $1,099 and $27,789 respectively.Normal: bottom 90%Power: top 10%Share of researchers90%10%Share of daily usage value26.3%73.7%
Figure 2. Normal users and power users. Define power users as the top 10%. They account for 73.7% of daily value and average $27,789 each, versus $1,099 for the bottom 90%. These are two groups within the same distribution.
Effect of daily usage caps, with ±1σ, ±2σ, ±3σ, and ±4σ calloutsAll amounts are USD valued at API prices. Standard deviations refer to the natural logarithm of per-researcher usage, not dollar deviations from the arithmetic mean. The cap at k standard deviations is 600 times exp(k times 1.917001). Bars show 1,125 times the expected capped usage. The eight requested callouts and a zero-sigma median reference are added to the existing $7,000, $25,000, $100,000 and no-cap scenarios. Small circles mark the exact endpoints of the sigma-scenario bars; bar widths are not exaggerated.Log offsetDaily capTotal / day$0$2M$4M−4σ: daily cap $0.28052995; total $315.59 per day−4σ$0.28$316−3σ: daily cap $1.90775341; total $2,145.19 per day−3σ$1.91$2,145−2σ: daily cap $12.97374158; total $14,453.45 per day−2σ$12.97$14,453−1σ: daily cap $88.22836817; total $91,000.48 per day−1σ$88.23$91,0000σ: daily cap $600.00000000; total $454,585.47 per day; median reference0σ$600$454,585+1σ: daily cap $4,080.32028096; total $1,489,544.72 per day+1σ$4,080$1.49Mdaily cap $7,000.00000000; total $1,900,609.51 per day$7,000$1.90Mdaily cap $25,000.00000000; total $2,895,094.35 per day$25,000$2.90M+2σ: daily cap $27,748.35599205; total $2,970,061.60 per day+2σ$27,748$2.97Mdaily cap $100,000.00000000; total $3,709,063.42 per day$100,000$3.71M+3σ: daily cap $188,703.63286282; total $3,934,916.12 per day+3σ$188,704$3.93M+4σ: daily cap $1,283,285.43376885; total $4,206,088.33 per day+4σ$1,283,285$4.21MNo cap; total $4,239,324.92 per dayNo cap$4.24MTotal daily value for 1,125 researchers
Effect of daily usage caps, with ±1σ, ±2σ, ±3σ, and ±4σ calloutsAll amounts are USD valued at API prices. Standard deviations refer to the natural logarithm of per-researcher usage, not dollar deviations from the arithmetic mean. The cap at k standard deviations is 600 times exp(k times 1.917001). Bars show 1,125 times the expected capped usage. The eight requested callouts and a zero-sigma median reference are added to the existing $7,000, $25,000, $100,000 and no-cap scenarios. Small circles mark the exact endpoints of the sigma-scenario bars; bar widths are not exaggerated.OffsetDaily capTotal / day$0$2M$4M−4σ: daily cap $0.28052995; total $315.59 per day−4σ$0.28$316−3σ: daily cap $1.90775341; total $2,145.19 per day−3σ$1.91$2,145−2σ: daily cap $12.97374158; total $14,453.45 per day−2σ$12.97$14,453−1σ: daily cap $88.22836817; total $91,000.48 per day−1σ$88.23$91,0000σ: daily cap $600.00000000; total $454,585.47 per day; median reference0σ$600$454,585+1σ: daily cap $4,080.32028096; total $1,489,544.72 per day+1σ$4,080$1.49Mdaily cap $7,000.00000000; total $1,900,609.51 per day$7,000$1.90Mdaily cap $25,000.00000000; total $2,895,094.35 per day$25,000$2.90M+2σ: daily cap $27,748.35599205; total $2,970,061.60 per day+2σ$27,748$2.97Mdaily cap $100,000.00000000; total $3,709,063.42 per day$100,000$3.71M+3σ: daily cap $188,703.63286282; total $3,934,916.12 per day+3σ$188,704$3.93M+4σ: daily cap $1,283,285.43376885; total $4,206,088.33 per day+4σ$1,283,285$4.21MNo cap; total $4,239,324.92 per dayNo cap$4.24MTotal daily value for 1,125 researchers
Figure 3. Dependence on the upper tail. Bars show total daily value after capping each researcher’s usage. The σ callouts use cap = $600 × ekσ, where σ = 1.917001 is the standard deviation of ln(X) and k ranges from −4 to +4. Zero marks the $600 median. Only caps of $7,000 or more preserve both the median and P90; capped values are no longer lognormal.

Limits. The source says “exceeded,” not “equal to.” Its two thresholds do not establish the mean or the upper tail, so $4.24M is not a proven lower bound. The March headcount and later usage observations are from different dates. All three charts are calculations, not employee data.

References

[1]

OpenAI. Research acceleration: The view inside OpenAI. September 6, 2026, §1. Median observation: mid-August. Repeated in Introducing GPT-6 Sol and Luna, September 22, 2026.

[2]

Reuters. OpenAI to nearly double workforce to 8,000 by end-2026, FT reports. March 21, 2026. Reported headcount: 4,500; 8,000 was a hiring target. Reuters did not independently verify the FT report.

[3]

NIST/SEMATECH. e-Handbook of Statistical Methods, §1.3.6.6.9: Lognormal Distribution. Distribution, quantiles, and mean. Parameters and results calculated here.