Model lifecycle
Grok 4.5
Released 2026-07-08 · Back to Grok
Exact-version signals observed 2026-08-06–2026-09-03 · n=61
Data window: 2026-08-08 00:00:00 to 2026-09-05 00:00:00 UTC · Scoring method: experience_score_v2.
Lifecycle updated daily · latest complete UTC day 2026-09-04
- Fixed 14-day baseline
- 73.1 n=34
- Latest 28 days
- 60.5 n=59
- Change vs baseline
- −12.6
- 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.
1 of 4 independent 14-day segments ready for trend judgment.
What explains the change
Experience change and discussion-mix change are separated. These are observational contributions, not proof of cause.
Largest negative experience contributions: General text, Coding.
| Category | Baseline → current | Category change | Weight share | Experience contribution | Discussion-mix contribution |
|---|---|---|---|---|---|
| General text | Baseline → current54.0 → 46.2n=14 → 24 | Category change−7.8 | Weight share40.8% → 39.4% | Experience contribution−3.13 | Discussion-mix contribution+0.14 |
| Coding | Baseline → current70.7 → 66.9n=10 → 11 | Category change−3.8 | Weight share28.1% → 17.4% | Experience contribution−0.87 | Discussion-mix contribution−0.91 |
| 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=6 → 7 | Category change– | Weight share– | Experience contribution– | Discussion-mix contribution– |
| Roleplay / creative | Baseline → currentnot enough datan=0 → 5 | Category change– | Weight share– | Experience contribution– | Discussion-mix contribution– |
| Safety & refusals | Baseline → currentnot enough datan=0 → 1 | Category change– | Weight share– | Experience contribution– | Discussion-mix contribution– |
| Speed & latency | Baseline → currentnot enough datan=4 → 11 | 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– |
Unresolved contribution from categories below the sample threshold: −7.83 pts.
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.