Model Change Log, September 2026
1 of 9 AI models answered the September 2026 edition with a different model version from August 2026: Grok.
Model versions, August 2026 to September 2026
| AI model | August 2026 | September 2026 | Own top pick changed |
|---|---|---|---|
| ChatGPT | gpt-5.5-2026-04-23 | gpt-5.5-2026-04-23 | 46% of 178 |
| Claude | claude-sonnet-5 | claude-sonnet-5 | 38% of 183 |
| Gemini | gemini-3.5-flash | gemini-3.5-flash | no comparison |
| Perplexity | sonar | sonar | 37% of 189 |
| DeepSeek | deepseek-v4-flash | deepseek-v4-flash | 34% of 155 |
| Grok | grok-4.5 | grok-4.3 | 44% of 183 |
| Copilot | Read from the live product, which reports no model version | 54% of 186 | |
| Google AI | Read from the live product, which reports no model version | no comparison | |
| Google AI Mode | Read from the live product, which reports no model version | no comparison | |
The model each AI model answered with in most of the edition’s categories, as stamped on every frozen answer. “Own top pick changed” is the share of categories, answered in both months, where that AI model’s own top pick is a different brand.
Where the model version changed, the AI model’s own top pick moved in 44% of categories. Where it held, the own top pick moved in 42%.
Boards that moved under a new model
Grok
- AI coding assistants: GitHub Copilot to Cursor
- digital marketing agencies: WebFX to Thrive Internet Marketing Agency
- home security systems: SimpliSafe to Ring
- password managers: Bitwarden to Google
- smartwatches: Apple Watch Series 11 to Apple Watch Ultra 3
Every number on this page is from the September 2026 edition of the CiteHawk AI Index, computed from the frozen record with nothing written by hand. Rankings are never for sale. Browse the September 2026 edition → Read the methodology →