Model lifecycle
GPT-5.6 Terra
Released 2026-07-10 · Back to GPT
Exact-version signals observed 2026-08-13–2026-08-24 · n=4
Lifecycle updated daily · latest complete UTC day 2026-08-27
- Fixed 14-day baseline
- – –
- Latest 28 days
- – n=4
- Change vs baseline
- –
- 90-day slope
- – pts / 30 days
14-day rolling experience
Each daily point summarizes the previous 14 complete UTC days. The dashed line is the fixed release baseline; the verdict still uses independent non-overlapping 14-day segments.
A complete window needs n=30 across at least 7 active days · current signals n=4
What explains the change
Experience change and discussion-mix change are separated. These are observational contributions, not proof of cause.
No category has enough samples in both windows for a contribution conclusion.
| Category | Baseline → current | Category change | Weight share | Experience contribution | Discussion-mix contribution |
|---|---|---|---|---|---|
| Coding | Baseline → currentnot enough datan=0 → 0 | Category change– | Weight share– | Experience contribution– | Discussion-mix contribution– |
| Image / vision | Baseline → currentnot enough datan=0 → 0 | Category change– | Weight share– | Experience contribution– | Discussion-mix contribution– |
| Local deploy | Baseline → currentnot enough datan=0 → 0 | Category change– | Weight share– | Experience contribution– | Discussion-mix contribution– |
| Reasoning | Baseline → currentnot enough datan=0 → 1 | Category change– | Weight share– | Experience contribution– | Discussion-mix contribution– |
| Roleplay / creative | Baseline → currentnot enough datan=0 → 0 | Category change– | Weight share– | Experience contribution– | Discussion-mix contribution– |
| Safety & refusals | Baseline → currentnot enough datan=0 → 0 | Category change– | Weight share– | Experience contribution– | Discussion-mix contribution– |
| Speed & latency | Baseline → currentnot enough datan=0 → 1 | Category change– | Weight share– | Experience contribution– | Discussion-mix contribution– |
| General text | Baseline → currentnot enough datan=0 → 2 | Category change– | Weight share– | Experience contribution– | Discussion-mix contribution– |
| Video generation | Baseline → currentnot enough datan=0 → 0 | Category change– | Weight share– | Experience contribution– | Discussion-mix contribution– |
Lifecycle scores use the same experience-signal weights as the main index, without sample shrinkage after n=30. They measure public user perception, not model capability or backend causes.