Report · resources

Rolling chat

Qwen3.6-35B-A3B · NVFP4 variant · NVIDIA NGC vLLM 0.29

Measured results / Rolling chat / Run 5f1b8f8e534b4032a6bb6205

Measured 2026-10-07 08:41 UTC · Record date 2026-10-07

Failed. 59/105 turns

Answer checks

59/105 turns

Rolling-chat turn checks
All rolling-chat turns
Person · turnFilled inputFirst wordRecent · pinned · oldResultRun
1 · 1
20113
4.62 s
true · true · recalled
Passed
1 · 2
40238
5.75 s
true · true · recalled
Passed
1 · 3
60364
6.94 s
true · true · recalled
Passed
1 · 4
80492
8.79 s
true · true · recalled
Passed
1 · 5
100620
10.00 s
true · true · recalled
Passed
1 · 6
120744
10.97 s
true · true · recalled
Passed
1 · 7
140868
15.05 s
true · true · recalled
Passed
1 · 8
160993
18.29 s
true · true · recalled
Passed
1 · 9
181122
21.94 s
true · true · recalled
Passed
1 · 10
201252
24.99 s
true · true · recalled
Passed
1 · 11
221374
26.37 s
true · true · recalled
Passed
1 · 12
241504
32.42 s
true · true · recalled
Passed
1 · 13
181126
119.66 s
true · true · incorrect
Failed
1 · 14
201257
21.21 s
true · true · incorrect
Failed
1 · 15
221379
35.19 s
true · true · incorrect
Failed
1 · 16
241508
59.34 s
true · true · incorrect
Failed
1 · 17
181129
216.08 s
true · true · incorrect
Failed
1 · 18
201255
33.66 s
true · true · incorrect
Failed
1 · 19
221380
65.85 s
true · true · incorrect
Failed
1 · 20
241505
67.60 s
true · true · incorrect
Failed
1 · 21
181116
275.56 s
true · true · incorrect
Failed
2 · 1
20120
11.53 s
true · true · recalled
Passed
2 · 2
40243
18.30 s
true · true · incorrect
Failed
2 · 3
60370
21.92 s
true · true · recalled
Passed
2 · 4
80499
24.55 s
true · true · recalled
Passed
2 · 5
100624
26.52 s
true · true · recalled
Passed
2 · 6
120747
29.09 s
true · true · recalled
Passed
2 · 7
140875
19.78 s
true · true · recalled
Passed
2 · 8
161002
30.13 s
true · true · recalled
Passed
2 · 9
181131
24.17 s
true · true · recalled
Passed
2 · 10
201259
25.84 s
true · true · recalled
Passed
2 · 11
221383
41.40 s
true · true · recalled
Passed
2 · 12
241516
60.02 s
true · true · recalled
Passed
2 · 13
181140
215.31 s
true · true · incorrect
Failed
2 · 14
201268
33.46 s
true · true · incorrect
Failed
2 · 15
221391
63.18 s
true · true · incorrect
Failed
2 · 16
241517
66.50 s
true · true · incorrect
Failed
2 · 17
181141
276.94 s
true · true · incorrect
Failed
2 · 18
201270
65.36 s
true · true · incorrect
Failed
2 · 19
221403
55.90 s
true · true · incorrect
Failed
2 · 20
241531
40.94 s
true · true · incorrect
Failed
2 · 21
181148
228.88 s
true · true · incorrect
Failed
3 · 1
20118
27.37 s
true · true · recalled
Passed
3 · 2
40246
20.19 s
true · true · recalled
Passed
3 · 3
60372
29.64 s
true · true · recalled
Passed
3 · 4
80497
23.42 s
true · true · recalled
Passed
3 · 5
100625
25.99 s
true · true · recalled
Passed
3 · 6
120754
40.51 s
true · true · recalled
Passed
3 · 7
140885
60.09 s
true · true · recalled
Passed
3 · 8
161011
33.26 s
true · true · recalled
Passed
3 · 9
181140
53.18 s
true · true · recalled
Passed
3 · 10
201268
35.08 s
true · true · recalled
Passed
3 · 11
221392
39.30 s
true · true · recalled
Passed
3 · 12
241516
69.03 s
true · true · recalled
Passed
3 · 13
181133
215.44 s
true · true · incorrect
Failed
3 · 14
201261
46.06 s
true · true · incorrect
Failed
3 · 15
221389
64.95 s
true · true · incorrect
Failed
3 · 16
241516
57.47 s
true · true · incorrect
Failed
3 · 17
181128
251.12 s
true · true · incorrect
Failed
3 · 18
201258
62.15 s
true · true · incorrect
Failed
3 · 19
221387
48.99 s
true · true · incorrect
Failed
3 · 20
241516
40.72 s
true · true · incorrect
Failed
3 · 21
181142
211.02 s
true · true · incorrect
Failed
4 · 1
20119
17.67 s
true · true · recalled
Passed
4 · 2
40247
39.03 s
true · true · recalled
Passed
4 · 3
60377
59.30 s
true · true · recalled
Passed
4 · 4
80500
32.47 s
true · true · recalled
Passed
4 · 5
100629
52.05 s
true · true · recalled
Passed
4 · 6
120753
31.70 s
true · true · recalled
Passed
4 · 7
140880
39.83 s
true · true · recalled
Passed
4 · 8
161011
63.29 s
true · true · recalled
Passed
4 · 9
181138
66.80 s
true · true · recalled
Passed
4 · 10
201271
51.48 s
true · true · recalled
Passed
4 · 11
221403
52.65 s
true · true · recalled
Passed
4 · 12
241528
67.30 s
true · true · recalled
Passed
4 · 13
181153
192.69 s
true · true · incorrect
Failed
4 · 14
201282
51.60 s
true · true · incorrect
Failed
4 · 15
221413
59.58 s
true · true · incorrect
Failed
4 · 16
241544
52.31 s
true · true · incorrect
Failed
4 · 17
181165
155.46 s
true · true · incorrect
Failed
4 · 18
201296
44.99 s
true · true · incorrect
Failed
4 · 19
221428
32.99 s
true · true · incorrect
Failed
4 · 20
241559
35.60 s
true · true · incorrect
Failed
4 · 21
181168
121.23 s
true · true · incorrect
Failed
5 · 1
20118
37.70 s
true · true · recalled
Passed
5 · 2
40245
60.05 s
true · true · recalled
Passed
5 · 3
60372
33.31 s
true · true · recalled
Passed
5 · 4
80502
52.87 s
true · true · recalled
Passed
5 · 5
100631
34.96 s
true · true · recalled
Passed
5 · 6
120762
39.33 s
true · true · recalled
Passed
5 · 7
140893
56.77 s
true · true · recalled
Passed
5 · 8
161021
66.22 s
true · true · recalled
Passed
5 · 9
181153
51.70 s
true · true · recalled
Passed
5 · 10
201281
54.78 s
true · true · recalled
Passed
5 · 11
221409
65.36 s
true · true · recalled
Passed
5 · 12
241537
56.99 s
true · true · recalled
Passed
5 · 13
181150
246.41 s
true · true · incorrect
Failed
5 · 14
201284
62.80 s
true · true · incorrect
Failed
5 · 15
221417
51.88 s
true · true · incorrect
Failed
5 · 16
241553
41.32 s
true · true · incorrect
Failed
5 · 17
181163
212.36 s
true · true · incorrect
Failed
5 · 18
201292
33.33 s
true · true · incorrect
Failed
5 · 19
221424
35.20 s
true · true · incorrect
Failed
5 · 20
241552
24.32 s
true · true · incorrect
Failed
5 · 21
181169
128.96 s
true · true · incorrect
Failed

Method

Staggered multi-turn fictional archive, exact engine-tokenized rolling gateway; independent recent, pinned and old probes each turn.

Limits

Lab retrieval and rolling-history evidence only. No summaries. Shared engine cache intervals include overlapping users and are not per-user hits. Gateway path, not authenticated document ingestion or customer acceptance.

What did not work

  • 59/105 requests passed.

What ran

Exact identity

Original publisher model
Qwen3.6-35B-A3B
Publisher revision
995ad96eacd98c81ed38be0c5b274b04031597b0
Variant · format · precision
NVFP4 variant · safetensors · NVFP4
Weights SHA-256
not recorded
Engine · version · feature
NVIDIA NGC vLLM · 0.29.0+5013de39.dev · plain
Hardware
NVIDIA GB10, 128 GB unified memory
Harness / workload
keep-rolling-context-v1 · 1
Workload suite · version
not recorded
Harness SHA-256
not recorded
Fixture SHA-256
not recorded

Workload shape

People in this run
5
Configured window (tokens)
262144
Configured window (tokens)
262144
Output token ceiling
256

Full reproduction pack available to customers and evaluators on request

All recorded numbers

Metric definitions, scope and statistics
All aggregate numeric result fields in this run
MetricValueScopeStatisticSample countBasis / source fieldRun
roll ttft max seconds
276.94 s
per person
max
not recorded
roll_ttft_max_seconds
requests ok
59.00 count
machine total
single
not recorded
requests_ok
roll ttft p50 seconds
215.31 s
per person
median
not recorded
roll_ttft_p50_seconds
normal ttft max seconds
69.03 s
per person
max
not recorded
normal_ttft_max_seconds
normal ttft p50 seconds
39.16 s
per person
median
not recorded
normal_ttft_p50_seconds
requests total
105.00 count
machine total
single
not recorded
requests_total

Data and citation

Skilak Keep evidence, run 5f1b8f8e534b4032a6bb6205, measured 2026-10-07T08:41:29.987232+00:00.
https://skilak.ai/keep/resources/evidence/r/5f1b8f8e534b4032a6bb6205

Ready to scope it?

Start with one workload.

Tell us who uses it, what it can read, and what must stay inside. We will map the build.