Report · resources

Team load

Ornith-1.5-35B-A3B · NVFP4 variant · NVIDIA NGC vLLM 0.29

Measured results / Team load / Run 024e2d0373e84abfb3875bf4

Measured 2026-10-07 05:44 UTC · Record date 2026-10-07

Failed. 39/48 Hidden facts found13/16 probes16/16 request checks

Writing speed, each persondecode_tps · tok/s
7.51 tok/sper person · median of 16
First word afterttft_median_s · s
93.56 sper person · median of 16
Slowest first wordttft_max_s · s
129.45 sper person · max of 16
Machine totalaggregate_decode_tps · tok/s
79.26 tok/smachine total · single · sample count not recorded
Machine total including first-word waitaggregate_e2e_tps · tok/s
50.34 tok/smachine total · single · sample count not recorded
Everyone answered inwall_seconds · s
162.72 smachine total · single · sample count not recorded
Peak memory, whole hostpeak_host_unavailable_gib · GiB
94.55 GiBmachine total · max · sample count not recorded

Runs on this setup

Earlier runs keep their own engine versions and outcomes. Workload shapes stay separate.

History and measured sweep
History of this setup, newest first
DateShapeEach person · medianMachine totalFirst word · medianResultRun
2026-10-07
20 people
5.95 tok/s
71.65 tok/s
113.84 s
Failed
2026-10-07
16 people
7.51 tok/s
79.26 tok/s
93.56 s
Failed
2026-10-07
8 people
16.69 tok/s
125.33 tok/s
54.30 s
Failed
2026-10-07
1 people
67.26 tok/s
67.26 tok/s
10.06 s
Passed

Each person in this run

First-word wait against writing speed after that word, including pauses. Earlier streams can pause while other people’s prompts are read. The table preserves filled input, cached-token counts and each person’s checks.

Each person's first word and writing speed; table below00 s968 s18136 stok/s after first wordfirst word afterPerson 1: 119.68 s · 11.97 tok/sPerson 2: 62.85 s · 5.22 tok/sPerson 3: 62.84 s · 5.22 tok/sPerson 4: 115.76 s · 10.99 tok/sPerson 5: 125.59 s · 13.82 tok/sPerson 6: 106.42 s · 9.20 tok/sPerson 7: 62.84 s · 5.22 tok/sPerson 8: 93.56 s · 7.51 tok/sPerson 9: 118.85 s · 11.75 tok/sPerson 10: 128.75 s · 15.06 tok/sPerson 11: 93.56 s · 7.51 tok/sPerson 12: 62.84 s · 5.22 tok/sPerson 13: 59.56 s · 5.06 tok/sPerson 14: 129.45 s · 15.36 tok/sPerson 15: 92.11 s · 7.36 tok/sPerson 16: 61.29 s · 5.15 tok/s
Per-person table · all measured request rows
Per-person requests for this run
PersonFilled inputCached tokensOutputFirst wordWriting after first wordTotal timeChecksRun
1
40,105
unknown
512
119.68 s
11.97 tok/sone person · includes waits
162.36 s
3/3 Hidden facts found · completed · checks passed
2
40,102
unknown
512
62.85 s
5.22 tok/sone person · includes waits
160.67 s
3/3 Hidden facts found · completed · checks passed
3
40,103
unknown
512
62.84 s
5.22 tok/sone person · includes waits
160.67 s
3/3 Hidden facts found · completed · checks passed
4
40,105
unknown
512
115.76 s
10.99 tok/sone person · includes waits
162.23 s
3/3 Hidden facts found · completed · checks passed
5
40,107
unknown
512
125.59 s
13.82 tok/sone person · includes waits
162.56 s
0/3 Hidden facts found · completed · checks missed
6
40,102
unknown
512
106.42 s
9.20 tok/sone person · includes waits
161.97 s
0/3 Hidden facts found · completed · checks missed
7
40,105
unknown
512
62.84 s
5.22 tok/sone person · includes waits
160.66 s
3/3 Hidden facts found · completed · checks passed
8
40,101
unknown
512
93.56 s
7.51 tok/sone person · includes waits
161.61 s
3/3 Hidden facts found · completed · checks passed
9
40,106
unknown
512
118.85 s
11.75 tok/sone person · includes waits
162.33 s
3/3 Hidden facts found · completed · checks passed
10
40,101
unknown
512
128.75 s
15.06 tok/sone person · includes waits
162.68 s
0/3 Hidden facts found · completed · checks missed
11
40,104
unknown
512
93.56 s
7.51 tok/sone person · includes waits
161.61 s
3/3 Hidden facts found · completed · checks passed
12
40,106
unknown
512
62.84 s
5.22 tok/sone person · includes waits
160.66 s
3/3 Hidden facts found · completed · checks passed
13
40,104
unknown
512
59.56 s
5.06 tok/sone person · includes waits
160.53 s
3/3 Hidden facts found · completed · checks passed
14
40,103
unknown
512
129.45 s
15.36 tok/sone person · includes waits
162.72 s
3/3 Hidden facts found · completed · checks passed
15
40,105
unknown
512
92.11 s
7.36 tok/sone person · includes waits
161.56 s
3/3 Hidden facts found · completed · checks passed
16
40,107
unknown
512
61.29 s
5.15 tok/sone person · includes waits
160.60 s
3/3 Hidden facts found · completed · checks passed
Per-person table · all measured request rows
Per-person requests for this run
PersonFilled inputCached tokensOutputFirst wordWriting after first wordTotal timeChecksRun
1
40,105
unknown
512
119.68 s
11.97 tok/sone person · includes waits
162.36 s
3/3 Hidden facts found · completed · checks passed
2
40,102
unknown
512
62.85 s
5.22 tok/sone person · includes waits
160.67 s
3/3 Hidden facts found · completed · checks passed
3
40,103
unknown
512
62.84 s
5.22 tok/sone person · includes waits
160.67 s
3/3 Hidden facts found · completed · checks passed
4
40,105
unknown
512
115.76 s
10.99 tok/sone person · includes waits
162.23 s
3/3 Hidden facts found · completed · checks passed
5
40,107
unknown
512
125.59 s
13.82 tok/sone person · includes waits
162.56 s
0/3 Hidden facts found · completed · checks missed
6
40,102
unknown
512
106.42 s
9.20 tok/sone person · includes waits
161.97 s
0/3 Hidden facts found · completed · checks missed
7
40,105
unknown
512
62.84 s
5.22 tok/sone person · includes waits
160.66 s
3/3 Hidden facts found · completed · checks passed
8
40,101
unknown
512
93.56 s
7.51 tok/sone person · includes waits
161.61 s
3/3 Hidden facts found · completed · checks passed
9
40,106
unknown
512
118.85 s
11.75 tok/sone person · includes waits
162.33 s
3/3 Hidden facts found · completed · checks passed
10
40,101
unknown
512
128.75 s
15.06 tok/sone person · includes waits
162.68 s
0/3 Hidden facts found · completed · checks missed
11
40,104
unknown
512
93.56 s
7.51 tok/sone person · includes waits
161.61 s
3/3 Hidden facts found · completed · checks passed
12
40,106
unknown
512
62.84 s
5.22 tok/sone person · includes waits
160.66 s
3/3 Hidden facts found · completed · checks passed
13
40,104
unknown
512
59.56 s
5.06 tok/sone person · includes waits
160.53 s
3/3 Hidden facts found · completed · checks passed
14
40,103
unknown
512
129.45 s
15.36 tok/sone person · includes waits
162.72 s
3/3 Hidden facts found · completed · checks passed
15
40,105
unknown
512
92.11 s
7.36 tok/sone person · includes waits
161.56 s
3/3 Hidden facts found · completed · checks passed
16
40,107
unknown
512
61.29 s
5.15 tok/sone person · includes waits
160.60 s
3/3 Hidden facts found · completed · checks passed

Answer checks

39/48 Hidden facts found13/16 probes16/16 request checks

Method

team-load-v1; distinct generated 40000-token coding contexts plus fresh nonce prefixes; synchronized streams; temperature zero; thinking off; min_tokens=max_tokens=512; separate cold hidden facts probes; no retries.

Limits

Finite lab cohort; configured 262144 context differs from measured inputs. Aggregate uses first-output to final-completion window. Whole-host memory sampled during speed only; missing counters remain unassessed. Broader coding quality and delivery acceptance unassessed.

What did not work

  • 39/48 Hidden facts found.
  • 13/16 probes passed.

What ran

Exact identity

Original publisher model
Ornith-1.5-35B-A3B
Publisher revision
10fbf86fed7ecee4a061f8b499a618f46001cac1
Variant · format · precision
NVFP4 variant · safetensors · nvfp4
Weights SHA-256
648fe83171485f4be03ea58d86089e088c94c2a3dc88922233f9002f30d8bc8b
8303cef7d375021d69dcafc60d3fad5f8186995b18b0db5e2a609aadeb8a4f51
cb5c5978721a1e87406ae68f1cd5a855326e40ff8bc13832644c392115336987
Engine · version · feature
NVIDIA NGC vLLM · 0.29.0+5013de39.dev · plain
Hardware
NVIDIA GB10, 128 GB unified memory
Harness / workload
keep-team-load-v1 · 1
Workload suite · version
not recorded
Harness SHA-256
not recorded
Fixture SHA-256
e2097959b1ca0142f2b93edc3ec411e474c0a6622800dd247af7a71bd9c39e83

Workload shape

People in this run
16
Configured window (tokens)
262144
Configured window (tokens)
262144
KV cache precision
fp8
Output token ceiling
512
Feature
plain
Thinking
off

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
wall seconds
162.72 s
machine total
single
not recorded
wall_seconds
ttft median s
93.56 s
per person
median
16
ttft_median_s
hidden facts total
48.00 count
machine total
single
not recorded
needle_total
ttft max s
129.45 s
per person
max
16
ttft_max_s
requests ok
16.00 count
machine total
single
not recorded
requests_ok
hidden facts correct
39.00 count
machine total
single
not recorded
needle_correct
preemptions
0.00 count
machine total
single
not recorded
preemptions
decode overlap
16.00 count
machine total
single
not recorded
decode_overlap
aggregate concurrency
16.00 count
machine total
single
not recorded
aggregate_concurrency
requests total
16.00 count
machine total
single
not recorded
requests_total
correctness probes passed
13.00 count
machine total
single
not recorded
correctness_probes_passed
aggregate decode tps
79.26 tok/s
machine total
single
not recorded
aggregate_tps
aggregate e2e tps
50.34 tok/s
machine total
single
not recorded
aggregate_end_to_end_tps
correctness probes total
16.00 count
machine total
single
not recorded
correctness_probes_total
decode tps
8.85 tok/s
per person
mean
16
per_request_decode_tps
forced output tokens
512.00 tokens
machine total
single
not recorded
forced_output_tokens
peak host unavailable gib
94.55 GiB
machine total
max
not recorded
peak_host_unavailable_gib
decode tps
7.51 tok/s
per person
median
16
per_person_decode_median_tps

Data and citation

Skilak Keep evidence, run 024e2d0373e84abfb3875bf4, measured 2026-10-07T05:44:16.608851+00:00.
https://skilak.ai/keep/resources/evidence/r/024e2d0373e84abfb3875bf4

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.