Evidaxis
Dated record

llm-d

org · Inference & Serving, n = 13 · llm-d/llm-d · paper: not linked · as of

32 commits/week averaged over the trailing window (niche median 32). No paper is linked in OpenAlex. Dependents: no reading in the scored panel. 4,661 GitHub stars, recorded, not scored. Measured for llm-d as of 2026-09-26; sources: GitHub, OpenAlex and deps.dev where linked.

commits/wk 32 averaged, trailing window niche median 32 (n = 13) last commit week 2026-09-20 measured · GitHub ·
citations no paper linked all years, indexed to date niche median 13 (n = 5) no reading · OpenAlex ·
direct dependents not in the scored panel partition 2026-09-26 niche median 8 (n = 8) no reading · deps.dev snapshot ·
stars 4,661 recorded, not scored niche median 14,719 (n = 13) measured · GitHub ·
Download values (CSV) JSON record Follow this system (feed) Report an error

Is llm-d still maintained?

Development velocity is the one measurable axis of three: 32 commits per week averaged over the trailing window (niche median 32/wk, n = 13).

GitHub averaged 32 per week over the trailing window; raw weekly counts below are individual weeks.

Last commit week 2026-09-20.

Commits in the latest captured week vs 4 weeks earlier: 31 · vs 26 weeks earlier: 22 · 52-week range: 0 to 101 commits in an individual week.

line chart: llm-d, weekly commits, 42 in latest week vs niche median 44, 2025-09-28 to 2026-09-20
llm-d averaged 32 commits a week, 1.0 times the niche median of 32. GitHub commit_activity · .
Latest eight weeks
Week startingCommitsNiche P25Niche P75
2026-08-023315128
2026-08-0910120167
2026-08-161919160
2026-08-231111125
2026-08-302925141
2026-09-063824135
2026-09-133523140
2026-09-204215138
heatmap chart: llm-d, weeks with commits, 26 of 26, 2026-03-29 to 2026-09-20
26 of the last 26 captured weeks had commits for llm-d. GitHub commit_activity · .
All 52 weeks
Weekly commits
Week startingCommits
2025-09-2821
2025-10-0517
2025-10-126
2025-10-199
2025-10-2615
2025-11-027
2025-11-092
2025-11-1620
2025-11-2319
2025-11-3011
2025-12-077
2025-12-142
2025-12-210
2025-12-280
2026-01-045
2026-01-113
2026-01-1813
2026-01-2531
2026-02-0123
2026-02-0829
2026-02-1524
2026-02-2226
2026-03-0119
2026-03-087
2026-03-1512
2026-03-2220
2026-03-2936
2026-04-0525
2026-04-1237
2026-04-1924
2026-04-2676
2026-05-0365
2026-05-1033
2026-05-1714
2026-05-2427
2026-05-3133
2026-06-0759
2026-06-1448
2026-06-2155
2026-06-2822
2026-07-0516
2026-07-1215
2026-07-1928
2026-07-2617
2026-08-0233
2026-08-09101
2026-08-1619
2026-08-2311
2026-08-3029
2026-09-0638
2026-09-1335
2026-09-2042

GitHub commit_activity · period 2025-09-28 to 2026-09-20 · n = 52 weeks · method weekly totals; reconstructed dates · updated · version 9d2fe956adbc-b1 · CSV

Readings

Recorded values and sources
ReadingValueΔ 4 weeksΔ 26 weeks52-week rangeInstrumentDateReproducePrevious reading
Commits per week32-3.314.37.4 to 41.3GitHub, trailing 12 weeks2026-09-26Public query30.3
Citing worksno readingno readingno readingno readingOpenAlex, all years2026-09-26Public queryno reading
Direct dependentsno readingno readingno readingno readingDirect dependents2026-09-26Public queryno reading
Daily dependents (unscored)no readingno readingno readingno readingdeps.dev REST, unscored2026-09-26Public queryno reading
Stars4,661347no reading3,674 to 4,661GitHub stars, unscored2026-09-26Public query4,593

Evidaxis snapshot · period 2026-09-26 · n = 1 system · method source linked readings · updated · version 9d2fe956adbc-b1

How to cite llm-d's record?

Evidaxis. llm-d measurement record. Snapshot 2026-09-26. urn:evidaxis:claim:e_KR3M8ESKAEW:m3:2026-09-26. CC0-1.0. https://evidaxis.org/e/e_KR3M8ESKAEW/

Evidaxis · period 2026-09-26 · n = 1 system · method claim URN · updated · version 9d2fe956adbc-b1

How does llm-d compare with its niche?

13 systems in Inference & Serving; niche median 32 commits per week. The reader chooses the sort order.

dot plot chart: llm-d, averaged commits per week, 32 vs niche median 32, 2026-09-26
llm-d at 32 commits a week among 13 systems; niche median 32. GitHub, current cohort · .
All 13 systems, alphabetical; choose a heading to sort
AIBrix22.2no readingno reading5,113
KTransformers4.39119,537
llama.cpp132.6no reading68129,606
llm-d32no readingno reading4,661
LMCache40.22311,912
LMDeploy13.1no readingno reading8,099
Mooncake64.31306,664
Nano-vLLM0no readingno reading15,631
Ollama23.6no reading1,078181,771
SGLang499.51021336,455
TensorRT-LLM191.2no reading014,719
vLLM536.91,58715092,723
Xinference29.2no readingno reading9,592

Evidaxis snapshot · period 2026-09-26 · n = 13 systems · method medians of measured values · updated · version 9d2fe956adbc-b1 · CSV for this niche

What changed since the last snapshot?

2 published readings changed between snapshots fcfd3a7ccdcb and 9d2fe956adbc.

slope chart: llm-d, changed published values, 2, 2026-09-19 to 2026-09-26
2 published values changed for llm-d since 2026-09-19. Evidaxis snapshots · .
Snapshot values
MetricPreviousCurrentDelta
Commits per week30.3321.7
Citing worksno readingno readingno reading
Direct dependentsno readingno readingno reading
Stars4,5934,66168

Evidaxis snapshots · period 2026-09-19 to 2026-09-26 · n = 15 partitions · method paired snapshot values · updated · version 9d2fe956adbc-b1 · History CSV

Take the data or follow changes

Values and captured history are available under CC0-1.0. The system feed records changes to published numbers.

Values CSVHistory CSVJSON recordAtom feedJSON FeedAdd Evidaxis as a preferred source on GoogleReport an error

Evidaxis · period 2026-09-26 · n = 1 system · method CC0-1.0 open export · updated · version 9d2fe956adbc-b1

For teams tracking many systems

Use the cohort CSV, snapshot archive and membership manifest to reproduce observations across systems.

Cohort CSV · Snapshot archive · Membership manifest

Paper and links

Source records identify the system, repository and linked works; no paper authors are published.

What is not measured for llm-d, and why

Missing readings describe instrument coverage. A missing value is not a measured zero.

Unmeasured slots
ReadingReason
Citing worksnot linked
Direct dependentsoutside panel
Daily dependents (unscored)no reading captured

Dependents: package outside the frozen m3 panel (2026-07-21), a coverage state, not a measured zero. Unscored daily count: no reading (2026-09-26).

Gate ETA: never published; Evidaxis makes no public actionable prediction.

Dependents ecosystem composition: no reading; this field is not present in the snapshot.

Evidaxis · period 2026-09-26 · n = 1 system · method measurement coverage · updated · version 9d2fe956adbc-b1

Measurement detail

Gate, gauges, percentile, derived signals and snapshot receipt

History: 13 weekly observations; minimum 4. Axis coverage: 1 of 3 axes measurable. Gate: not computable - 1 of 3 axes measurable

Gate: not computable - 1 of 3 axes measurable

llm-d

org
Inference & Serving · n = 13 · llm-d/llm-d
1 of 3 axes measurable · conf low
convergence gate
Convergence gate: Gate: not computable - 1 of 3 axes measurable Development velocity: measured × Citation momentum: not measurable ∅ Direct-dependents momentum (deps.dev): not measurable ∅
gate
0 of 3 axes

Gate: not computable - 1 of 3 axes measurable

· Evidaxis snapshot · 2026-09-26
Gate: not computable - 1 of 3 axes measurable
position in cohort distribution one of 13 in cohort
percentile 17 within cohort p25 p50 p75 p0 p100
percentile
p17
· Evidaxis snapshot · 2026-09-26
momentum averages present-axis z-scores; two- and three-axis scores are not strictly comparable
velocity z
velocity z gaugevelocity z-1.08 within cohort, log-slope -0.009/wk.0-1.08
velocity z
-1.08 σ
slope
-0.009/wk
· Evidaxis snapshot · 2026-09-26
log-slope -0.009/wk
citation z Citation source: not measured.
direct-dependents z Direct-dependents axis: outside the frozen panel; not measured.
cohort
13
systems measured
weekly commits
32
Development velocity axis: measured.
total citations
no reading
Citation source: not measured.
momentum
36.5
percentile p17
commit heatmap, last 26 weeks32/wk
weekly commits, 26 weeks26 weeks of commits, peak 101 in a week, 26 active weeks, 938 total.week 1: 36 commitsweek 2: 25 commitsweek 3: 37 commitsweek 4: 24 commitsweek 5: 76 commitsweek 6: 65 commitsweek 7: 33 commitsweek 8: 14 commitsweek 9: 27 commitsweek 10: 33 commitsweek 11: 59 commitsweek 12: 48 commitsweek 13: 55 commitsweek 14: 22 commitsweek 15: 16 commitsweek 16: 15 commitsweek 17: 28 commitsweek 18: 17 commitsweek 19: 33 commitsweek 20: 101 commitsweek 21: 19 commitsweek 22: 11 commitsweek 23: 29 commitsweek 24: 38 commitsweek 25: 35 commitsweek 26: 42 commits
weeks
26
peak/wk
101
active
26
total
938
· Evidaxis snapshot · 2026-09-26
citations by yearCitation source: not measured.
recency-weighted swing trailing 13-week vs prior 39-week log-slope ratio +1.03, regime steady. prior slope +0.052, recent slope +0.054 per week. +0.052prior 39w+0.054recent 13w+1.03×
swing
+1.03×
regime
steady
windows
13w vs 39w
snapshot 1 of 1 · · m3
commit-axis Sharpe log-linear slope over residual std, value +0.042 with 95% CI +0.008 to +0.077, 50 degrees of freedom. CI excludes zero.0+0.04295% CI +0.008, +0.077+0.008+0.077
sharpe
+0.042
95% ci
+0.008, +0.077
dof
50
snapshot 1 of 1 · · m3
commit changepoint52 weekly commit points. Changepoint at week 4 (2025-10-25), slope change +8.221 commits per week.2025-10-25
date
week
4 of 52
Δslope
+8.221/wk
snapshot 1 of 1 · · m3
code-leads-citation lag reserved

reserved: citation series too short / axis absent for cross-correlation

status
reserved
value
insufficient data
snapshot 1 of 1 · · m3
axis alpha vs cohort reserved

reserved: cohort N too small for regression (power ~0)

status
reserved
value
insufficient data
snapshot 1 of 1 · · m3
methodology gate snapshot 1 of 1 · · m3
commit weeks
52
citation years
0
shipped
3
reserved
3

shipped

  • recency-swing
  • commit-sharpe
  • commit-changepoint

reserved

  • code-leads-citation-lag
  • axis-alpha
  • gate-eta
snapshot 15 of 15 · · m3

Direct-dependents momentum (deps.dev)

Direct-dependents axis: outside the frozen panel; not measured.

Coverage state
out_of_panel
Cohort z
no axis
OLS slope
no reading
Theil-Sen slope
no reading
Latest direct dependents
no reading
Confirmed-clean points
no reading
Reconstructed history points
no reading
As-of partition
no reading
Fragility veto
no reading
Rising vote
no

Source: deps.dev (Google), CC-BY 4.0. Modified: aggregated to system-level counts and momentum scores.

Axis definition · Snapshot axis correlations

Reconstructed history

Reconstructed from git history, not a point-in-time capture. This series is separate from captured observations.

sparkline chart: llm-d, reconstructed commits, 35, 2025-w39 to 2026-w37
llm-d has 35 reconstructed commits in the latest week; reconstructed niche median not measured. Git history reconstruction · .
Reconstructed weekly commits
PeriodCommits
2025-w3921
2025-w4017
2025-w416
2025-w429
2025-w4315
2025-w447
2025-w452
2025-w4620
2025-w4719
2025-w4811
2025-w497
2025-w502
2026-w015
2026-w023
2026-w0313
2026-w0431
2026-w0523
2026-w0629
2026-w0724
2026-w0826
2026-w0919
2026-w107
2026-w1112
2026-w1220
2026-w1336
2026-w1425
2026-w1537
2026-w1624
2026-w1776
2026-w1865
2026-w1933
2026-w2014
2026-w2127
2026-w2233
2026-w2359
2026-w2448
2026-w2555
2026-w2622
2026-w2716
2026-w2815
2026-w2928
2026-w3017
2026-w3133
2026-w32101
2026-w3319
2026-w3411
2026-w3529
2026-w3638
2026-w3735
Peer signals
SystemMomentumPercentile
AIBrix58.883
KTransformers30.78
llama.cpp50.367
llm-d36.517
LMCache50.158
LMDeploy5050
Mooncake69.9100
Nano-vLLM24.50
Ollama49.542
SGLang49.133
TensorRT-LLM5675
vLLM37.725
Xinference61.592

Evidaxis · period 2026-09-26 · n = 1 system · method measurement detail · updated · version 9d2fe956adbc-b1 · record JSON · snapshot receipt