Report · resources

Answer quality

Qwen3.6-35B-A3B · NVFP4 variant · NVIDIA NGC vLLM 0.29 · with multi-token prediction

Measured results / Answer quality / Run 525c0716346b4c9b81526ec3

Measured 2026-10-19 08:01 UTC · Record date 2026-10-19

Partial. 111/114 tasks109/112 API cases

First word afterttft_median_s · s
0.06 sper person · median · 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
2026-10-19
unknown people
not recorded
not recorded
not recorded
Passed
2026-10-19
unknown people
not recorded
not recorded
not recorded
Passed
2026-10-19
unknown people
not recorded
not recorded
not recorded
Passed
2026-10-19
unknown people
not recorded
not recorded
not recorded
Passed
2026-10-19
unknown people
not recorded
not recorded
not recorded
Passed
2026-10-19
unknown people
not recorded
not recorded
not recorded
Passed
2026-10-19
unknown people
not recorded
not recorded
not recorded
Passed
2026-10-19
unknown people
not recorded
not recorded
not recorded
Failed
2026-10-19
1 people
not recorded
not recorded
0.06 s
Partial
2026-10-19
unknown people
not recorded
not recorded
not recorded
Partial
2026-10-19
unknown people
not recorded
not recorded
not recorded
Failed
2026-10-19
unknown people
not recorded
not recorded
not recorded
Passed
2026-10-19
unknown people
not recorded
not recorded
not recorded
Passed
2026-10-19
64 people · shared · warm
not recorded
not recorded
not recorded
Above confirmed limit — not attempted
2026-10-19
32 people · shared · warm
not recorded
not recorded
not recorded
Above confirmed limit — not attempted
2026-10-19
20 people · shared · warm
not recorded
not recorded
not recorded
Above confirmed limit — not attempted
2026-10-19
16 people · shared · warm
not recorded
not recorded
not recorded
Above confirmed limit — not attempted
2026-10-19
8 people · shared · warm
not recorded
not recorded
not recorded
Capacity limit — confirmed
2026-10-19
64 people · distinct · cold
not recorded
not recorded
not recorded
Above confirmed limit — not attempted
2026-10-19
32 people · distinct · cold
not recorded
not recorded
not recorded
Above confirmed limit — not attempted
2026-10-19
20 people · distinct · cold
not recorded
not recorded
not recorded
Above confirmed limit — not attempted
2026-10-19
16 people · distinct · cold
not recorded
not recorded
not recorded
Above confirmed limit — not attempted
2026-10-19
8 people · distinct · cold
not recorded
not recorded
not recorded
Capacity limit — confirmed
2026-10-19
64 people · distinct · cold
not recorded
not recorded
not recorded
Above confirmed limit — not attempted
2026-10-19
32 people · distinct · cold
not recorded
not recorded
not recorded
Above confirmed limit — not attempted
2026-10-19
20 people · distinct · cold
not recorded
not recorded
not recorded
Above confirmed limit — not attempted
2026-10-19
16 people · distinct · cold
not recorded
not recorded
not recorded
Above confirmed limit — not attempted
2026-10-19
8 people · distinct · cold
not recorded
not recorded
not recorded
Capacity limit — confirmed
2026-10-18
64 people · distinct · cold
not recorded
not recorded
not recorded
Above confirmed limit — not attempted
2026-10-18
32 people · distinct · cold
not recorded
not recorded
not recorded
Above confirmed limit — not attempted
2026-10-18
20 people · distinct · cold
not recorded
not recorded
not recorded
Above confirmed limit — not attempted
2026-10-18
16 people · distinct · cold
not recorded
not recorded
not recorded
Capacity limit — confirmed
2026-10-18
64 people · distinct · cold
not recorded
not recorded
not recorded
Above confirmed limit — not attempted
2026-10-18
32 people · distinct · cold
not recorded
not recorded
not recorded
Above confirmed limit — not attempted
2026-10-18
20 people · distinct · cold
not recorded
not recorded
not recorded
Above confirmed limit — not attempted
2026-10-18
16 people · distinct · cold
not recorded
not recorded
not recorded
Capacity limit — confirmed
2026-10-18
unknown people
not recorded
not recorded
not recorded
Passed
2026-10-18
unknown people
not recorded
not recorded
not recorded
Passed

Answer checks

111/114 tasks109/112 API cases

API suite category outcomes
CategoryPassed / totalSkippedRun
api
7/8
0
code
16/16
0
format
12/12
0
multi_turn
8/8
0
multilingual
6/6
0
numeric
16/16
0
reasoning
0/0
4
refusal
6/6
0
structured_extraction
15/16
0
supplied_context
15/16
0
tools
0/0
8
untrusted_text
8/8
0
vision
0/0
4

Method

Deterministic API suite core-v2; SHA256 0ad0b13a94ad74bcdefe99d12d796b175f56e5777d9642a1dd4c0f170d708398; no model judge; no retries; unchanged endpoint cache; latency excludes local grading; paired calls counted separately. Client concurrency; configured sequence limit 20.

Limits

Bounded fictional tasks only. Unsupported cases are separately counted, never passes. Original publisher name with variant; exact variant repository and revision retained privately. Not customer acceptance.

What did not work

  • 111/114 tasks passed.
  • 109/112 API cases passed.

What ran

Exact identity

Original publisher model
Qwen3.6-35B-A3B
Publisher revision
not recorded
Variant · format · precision
NVFP4 variant · safetensors · NVFP4
Weights SHA-256
not recorded
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-api-suite-run-v1 · 1
Workload suite · version
core-v2 · 2
Harness SHA-256
not recorded
Fixture SHA-256
0ad0b13a94ad74bcdefe99d12d796b175f56e5777d9642a1dd4c0f170d708398

Workload shape

People in this run
1
Configured window (tokens)
262144
KV cache precision
fp8
Feature
with multi-token prediction

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
skipped cases
16.00 count
machine total
single
not recorded
skipped_cases
correctness failures
2.00 count
machine total
single
not recorded
correctness_failures
ttft median s
0.06 s
per person
median
not recorded
ttft_median_s
workload total
114.00 count
machine total
single
not recorded
workload_total
workload passed
111.00 count
machine total
single
not recorded
workload_passed
api cases passed
109.00 count
machine total
single
not recorded
api_cases_passed
ttft p95 s
0.09 s
per person
p95
not recorded
ttft_p95_s
completion p95 s
0.52 s
per person
p95
not recorded
completion_p95_s
api cases total
112.00 count
machine total
single
not recorded
api_cases_total
completion median s
0.19 s
per person
median
not recorded
completion_median_s
completeness failures
1.00 count
machine total
single
not recorded
completeness_failures

Data and citation

Skilak Keep evidence, run 525c0716346b4c9b81526ec3, measured 2026-10-19T08:01:15.141355+00:00.
https://skilak.ai/keep/resources/evidence/r/525c0716346b4c9b81526ec3

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.