Report · resources

Structured tasks

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

Measured results / Structured tasks / Run 2f65fb170ba9426d809523e3

Measured 2026-10-07 06:49 UTC · Record date 2026-10-07

Partial. 9/32 tasks

Answer checks

9/32 tasks

Method

Synthetic structured-context tasks; no retries, no warmup; unchanged endpoint cache; all planned tasks remain in denominator.

Limits

Operator-supplied runtime manifest; the endpoint runner cannot attest hardware or model pins. Bounded task correctness is not customer acceptance.

What did not work

  • 9/32 tasks passed.

What ran

Exact identity

Original publisher model
Ornith-1.5-35B-A3B
Publisher revision
10fbf86fed7ecee4a061f8b499a618f46001cac1
Variant · format · precision
NVFP4 variant · NVFP4
Weights SHA-256
not recorded
Engine · version · feature
NVIDIA NGC vLLM · version not recorded · with multi-token prediction
Hardware
NVIDIA GB10, 128 GB unified memory
Harness / workload
not recorded
Workload suite · version
not recorded
Harness SHA-256
not recorded
Fixture SHA-256
not recorded

Workload shape

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
workload total
32.00 count
machine total
single
not recorded
workload_total
workload passed
9.00 count
machine total
single
not recorded
workload_passed

Data and citation

Skilak Keep evidence, run 2f65fb170ba9426d809523e3, measured 2026-10-07T06:49:08.449660+00:00.
https://skilak.ai/keep/resources/evidence/r/2f65fb170ba9426d809523e3

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.