Reasoning effort mix · May 2026
13.4% of chat requests set a reasoning effort in May 2026. Of those, "none" was the most requested level at 57.1%.
Share of requests that set a reasoning effort, and the mix of effort levels among those requests.
Data through May 31, 2026Updated November 11, 2026
This is the permanent snapshot for May 2026. See the latest month (October 2026)
Share of requests that set effort, May 2026
Source: Requesty production gateway- none57.1%
- medium6.6%
- high5.0%
- low3.8%
- max0.2%
Key findings
What to do about it
- Set effort per task. Most traffic runs fine at low; keep high for planning and hard debugging.
- Measure cost per task before and after changing effort. Higher effort raises output tokens.
Set reasoning effort once and Requesty maps it to each provider's format.
Read about reasoningData
| Effort | Share of requests that set effort | Organizations (band) |
|---|---|---|
| none | 57.08% | 25-99 |
| medium | 6.63% | 100-499 |
| high | 4.99% | 100-499 |
| low | 3.78% | 100-499 |
| max | 0.21% | 25-99 |
5 rows published. Rows below the floors are withheld.
Questions this answers
- What reasoning_effort do developers use most?
- 13.4% of chat requests set a reasoning effort in May 2026. Of those, "none" was the most requested level at 57.1%.
- How many API requests set a reasoning effort?
- 13.4% of chat requests set a reasoning effort in May 2026. Of those, "none" was the most requested level at 57.1%.
- Do developers use high reasoning effort?
- 13.4% of chat requests set a reasoning effort in May 2026. Of those, "none" was the most requested level at 57.1%.
Method and definitions
- Effort is read from the reasoning.effort or reasoning_effort request parameter. Numeric thinking budgets are grouped as token budget. Requests without the parameter count as not set.
- A row is published only when it has at least 20 requests, at least 5 distinct organizations, and no single organization above 80% of its traffic. These are the same floors /rankings uses. Rows below the floors are withheld, not estimated.
- Only shares, rates, medians and organization bands are published. Absolute request, token and spend totals stay private.
- Set effort:
- The request carried an explicit reasoning effort or thinking budget.
Caveats
- Requests that do not set an effort run at the model default, which differs by model.
- Some clients set an effort on every request even for models that ignore it.
Revision history and other months
- October 2026first published later
- September 2026first published later
- August 2026first published later
- July 2026first published later
- June 2026first published later
- May 2026 (this page)first published 2026-11-11
Cite and reuse
Free to reuse under CC BY 4.0. Credit "Requesty Data" and link to this page. Cite the month URL so the numbers you quote stay fixed.
APA
Click to copy
BibTeX
Click to copy
CSV
Click to copy