Report · resources

Long documents

Ornith-1.5-35B-A3B · NVFP4 variant · NVIDIA NGC vLLM 0.29 · with multi-token prediction

Measured results / Long documents / Run fcdbf0c4c05e4432a275612c

Measured 2026-10-07 07:04 UTC · Record date 2026-10-07

Failed. 12/15 Hidden facts found4/5 request checks

First word afterttft_median_s · s
673.54 sper person · median of 5
Slowest first wordttft_max_s · s
676.91 sper person · max of 5
Machine totalaggregate_decode_tps · tok/s
18.10 tok/smachine total · single · sample count not recorded
Machine total including first-word waitaggregate_e2e_tps · tok/s
0.86 tok/smachine total · single · sample count not recorded
Everyone answered inwall_seconds · s
682.35 smachine total · single · sample count not recorded
Peak memory, whole hostpeak_host_unavailable_gib · GiB
94.36 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
5 people · shared · warm
not recorded
85.36 tok/s
14.44 s
Passed
2026-10-07
5 people · shared · cold
not recorded
72.02 tok/s
672.79 s
Passed
2026-10-07
5 people · distinct · cold
not recorded
18.10 tok/s
673.54 s
Failed

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 s12355 s25711 stok/s after first wordfirst word afterPerson 2: 650.29 s · 3.76 tok/sPerson 4: 665.16 s · 6.61 tok/sPerson 1: 673.54 s · 13.54 tok/sPerson 3: 676.91 s · 21.62 tok/sPerson 5: 676.46 s · 20.76 tok/s
Per-person table · all measured request rows
Per-person requests for this run
PersonFilled inputCached tokensOutputFirst wordWriting after first wordTotal timeChecksRun
2
250,011
unknown
119
650.29 s
3.76 tok/sone person · includes waits
681.65 s
3/3 Hidden facts found · completed · checks passed
4
250,011
unknown
110
665.16 s
6.61 tok/sone person · includes waits
681.65 s
3/3 Hidden facts found · completed · checks passed
1
250,011
unknown
118
673.54 s
13.54 tok/sone person · includes waits
682.18 s
3/3 Hidden facts found · completed · checks passed
3
250,011
unknown
115
676.91 s
21.62 tok/sone person · includes waits
682.18 s
3/3 Hidden facts found · completed · checks passed
5
250,011
unknown
123
676.46 s
20.76 tok/sone person · includes waits
682.34 s
0/3 Hidden facts found · completed · checks missed
Per-person table · all measured request rows
Per-person requests for this run
PersonFilled inputCached tokensOutputFirst wordWriting after first wordTotal timeChecksRun
2
250,011
unknown
119
650.29 s
3.76 tok/sone person · includes waits
681.65 s
3/3 Hidden facts found · completed · checks passed
4
250,011
unknown
110
665.16 s
6.61 tok/sone person · includes waits
681.65 s
3/3 Hidden facts found · completed · checks passed
1
250,011
unknown
118
673.54 s
13.54 tok/sone person · includes waits
682.18 s
3/3 Hidden facts found · completed · checks passed
3
250,011
unknown
115
676.91 s
21.62 tok/sone person · includes waits
682.18 s
3/3 Hidden facts found · completed · checks passed
5
250,011
unknown
123
676.46 s
20.76 tok/sone person · includes waits
682.34 s
0/3 Hidden facts found · completed · checks missed

Answer checks

12/15 Hidden facts found4/5 request checks

Method

Synchronized long public-domain prompts, three exact hidden facts per user. Decode rate uses completion tokens minus one over first output through stream completion; effective prefill includes scheduling and first-token latency. Aggregate decode uses the union window from first output to final completion; end-to-end rate includes prefill.

Limits

Lab observation, not customer acceptance. Prompt content length excludes the server chat template; actual per-request usage is recorded. Synthesis is requested and retained but semantic accuracy is unassessed. All requested users must overlap in observed decode to pass.

What did not work

  • One or more requests did not meet completion, exact retrieval, synthesis-presence, reasoning-off or token-accounting requirements.

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 · with multi-token prediction
Hardware
NVIDIA GB10, 128 GB unified memory
Harness / workload
keep-long-context-concurrency-v1 · 1
Workload suite · version
not recorded
Harness SHA-256
not recorded
Fixture SHA-256
1e01cbcbd88ac9878b889cba44f56c74f5ab023f35e2d28a3bbe343923d909e0

Workload shape

cache state
cold
People in this run
5
Configured window (tokens)
262144
Filled content target (tokens)
250000
fixture variant
distinct-v1
KV cache precision
fp8
Output token ceiling
512
scenario
distinct
Feature
with multi-token prediction
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
682.35 s
machine total
single
not recorded
wall_seconds
ttft median s
673.54 s
per person
median
5
ttft_median_s
hidden facts total
15.00 count
machine total
single
not recorded
needle_total
ttft max s
676.91 s
per person
max
5
ttft_max_s
requests ok
4.00 count
machine total
single
not recorded
requests_ok
hidden facts correct
12.00 count
machine total
single
not recorded
needle_correct
latency bar seconds
300.00 s
machine total
single
not recorded
latency_bar_seconds
decode overlap
5.00 count
machine total
single
not recorded
decode_overlap
aggregate concurrency
5.00 count
machine total
single
not recorded
aggregate_concurrency
requests total
5.00 count
machine total
single
not recorded
requests_total
aggregate decode tps
18.10 tok/s
machine total
single
not recorded
aggregate_tps
aggregate e2e tps
0.86 tok/s
machine total
single
not recorded
aggregate_end_to_end_tps
peak host unavailable gib
94.36 GiB
machine total
max
not recorded
peak_host_unavailable_gib

Data and citation

Skilak Keep evidence, run fcdbf0c4c05e4432a275612c, measured 2026-10-07T07:04:09.414932+00:00.
https://skilak.ai/keep/resources/evidence/r/fcdbf0c4c05e4432a275612c

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.