Anthropic research: data analysis leads usage, not proof of efficiency
The November 5 study covers 207,049 research sessions from August 2 to September 30.
Consumer-plan Chat/Cowork and Claude Code sessions form the sample. Data analysis accounts for 33.7% and 48.1%, respectively. Usage shares neither represent all researchers nor prove efficiency gains.
Output-growth projections are scenarios. The study does not establish HK, Macau or Taiwan eligibility, Web/app/API features, Chinese/Cantonese support, prices or quotas. No hands-on testing was performed.
Output-growth projections are scenarios. The study does not establish HK, Macau or Taiwan eligibility, Web/app/API features, Chinese/Cantonese support, prices or quotas. No hands-on testing was performed.
Usage conditions
- Hong Kong
- Research report, not a local launch. It does not establish product eligibility, language support or pricing.
- Macau
- Research report, not a local launch. It does not establish product eligibility, language support or pricing.
- Taiwan
- Research report, not a local launch. It does not establish product eligibility, language support or pricing.
Official source: AI for scientific discovery: How researchers use Claude