The Latent
EN▾
EnglishEspañol中文PortuguêsFrançaisالعربية日本語한국어
Sign Up
NEWSDATARESEARCHPODCASTNEWSLETTERPRESSPARTNER WITH US →
Data/Capability

Capability

Historical Token Efficiency Across Frontier Models

Historical Token Efficiency Across Frontier Models. Demo series compiled by The Latent research desk.

Historical Token Efficiency Across Frontier Models

  • GPT-5.5
  • Claude 4.5
  • Gemini 3
Historical Token Efficiency Across Frontier ModelsDemo preview through July 2026 (benchmark points per 1M tokens).6420Aug '25Oct '25Dec '25Feb '26Apr '26Jun '26GPT-5.5: 4.4 on Jul '26. Claude 4.5: 5 on Jul '26. Gemini 3: 4 on Jul '26
SOURCE: The LatentUPDATED: Jul 1, 2026
ZOOMALLYTD12M3M1M
CSV
View data
DateGPT-5.5Claude 4.5Gemini 3
2025-08-013.273.742.96
2025-09-013.33.772.99
2025-10-013.74.233.35
2025-11-013.754.283.39
2025-12-013.74.233.35
2026-01-013.814.353.44
2026-02-013.564.073.22
2026-03-013.694.223.34
2026-04-0144.573.62
2026-05-014.044.623.66
2026-06-014.385.013.97
2026-07-014.3853.95

API access · Pro coming soon

Demo preview through July 2026 (benchmark points per 1M tokens).

Download CSVPro API coming soon

Methodology

Synthetic demo series generated to preview the data product layout.

Related charts

  • Historical Token Throughput per GPU
  • Historical Inference Speeds
  • Epoch Capabilities Index (ECI)

The Latent

AI industry news. A sister publication to The Block.

Editorial

  • Standards
  • Corrections
  • Commercial policy
  • Contact

Company

  • About
  • Privacy Policy
  • Terms of Service
  • Security
  • The Block