Model lifecycle

DeepSeek V4 Flash

Released 2026-04-24 · Back to DeepSeek

Exact-version signals observed 2026-06-12–2026-07-31 · n=1,839

No consistent trend Compared with the first sample-sufficient 14-day window after release.

Lifecycle updated daily · latest complete UTC day 2026-07-31

Fixed 14-day baseline
72.1
n=116
Latest 28 days
75.9
n=1,457
Change vs baseline
+3.8
90-day slope
+1.2
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.

14-day rolling experience from 2026-06-19 to 2026-08-01; latest score 78.6, fixed baseline 72.1.4050607080Fixed baseline 72.12026-06-05–2026-06-19 · 72.1 · n=1162026-06-06–2026-06-20 · 71.7 · n=1272026-06-07–2026-06-21 · 71.1 · n=1362026-06-08–2026-06-22 · 72.5 · n=1512026-06-09–2026-06-23 · 73.1 · n=1582026-06-10–2026-06-24 · 72.8 · n=1652026-06-11–2026-06-25 · 72.9 · n=1742026-06-12–2026-06-26 · 73.0 · n=2022026-06-13–2026-06-27 · 73.9 · n=2042026-06-14–2026-06-28 · 74.5 · n=1982026-06-15–2026-06-29 · 75.8 · n=1912026-06-16–2026-06-30 · 75.6 · n=1992026-06-17–2026-07-01 · 75.9 · n=2012026-06-18–2026-07-02 · 74.1 · n=1992026-06-19–2026-07-03 · 73.7 · n=2002026-06-20–2026-07-04 · 72.7 · n=2302026-06-21–2026-07-05 · 72.0 · n=2332026-06-22–2026-07-06 · 68.0 · n=2372026-06-23–2026-07-07 · 69.2 · n=2682026-06-24–2026-07-08 · 67.2 · n=2912026-06-25–2026-07-09 · 65.5 · n=3182026-06-26–2026-07-10 · 65.4 · n=3052026-06-27–2026-07-11 · 66.3 · n=3252026-06-28–2026-07-12 · 65.1 · n=3302026-06-29–2026-07-13 · 64.6 · n=3332026-06-30–2026-07-14 · 64.8 · n=3322026-07-01–2026-07-15 · 64.0 · n=3202026-07-02–2026-07-16 · 65.3 · n=3252026-07-03–2026-07-17 · 65.7 · n=3352026-07-04–2026-07-18 · 65.6 · n=3242026-07-05–2026-07-19 · 66.6 · n=3322026-07-06–2026-07-20 · 69.0 · n=3392026-07-07–2026-07-21 · 67.7 · n=3192026-07-08–2026-07-22 · 70.8 · n=3122026-07-09–2026-07-23 · 74.3 · n=3012026-07-10–2026-07-24 · 74.2 · n=3232026-07-11–2026-07-25 · 74.2 · n=3032026-07-12–2026-07-26 · 75.6 · n=3192026-07-13–2026-07-27 · 75.0 · n=3382026-07-14–2026-07-28 · 74.6 · n=3432026-07-15–2026-07-29 · 74.1 · n=3512026-07-16–2026-07-30 · 73.6 · n=3692026-07-17–2026-07-31 · 74.4 · n=4062026-07-18–2026-08-01 · 78.6 · n=113306-1906-2707-0507-1307-2107-2908-01

4 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: Speed & latency, Local deploy.

Category Baseline → current Category change Weight share Experience contribution Discussion-mix contribution
Speed & latency 79.0 → 75.5n=52 → 475 −3.6 43.7% → 32.5% −1.36 −0.55
Local deploy 67.7 → 65.9n=13 → 64 −1.9 9.4% → 4.0% −0.12 +0.29
Reasoning 70.5 → 85.5n=13 → 271 +15.0 11.6% → 20.5% +2.40 +0.51
General text 57.1 → 74.4n=32 → 473 +17.3 25.8% → 30.3% +4.85 −0.30
Coding not enough datan=9 → 141
Image / vision not enough datan=2 → 23
Roleplay / creative not enough datan=0 → 7
Safety & refusals not enough datan=0 → 9

Unresolved contribution from categories below the sample threshold: −1.92 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.