Report · resources

Chat and code

Qwen3.8-27B · GSQ-RCO IQ3_S GGUF variant · llama.cpp · with multi-token prediction

Measured results / Chat and code / Run 08b6b6698d9424baa272c369

Measured 2026-10-05 18:20 UTC · Record date 2026-10-05

Not graded. 6/6 request checks

Code writing speedcode_decode_tps · tok/s: 33.64 tok/sper person · mean · sample count not recorded
Prose writing speedchat_decode_tps · tok/s: 26.99 tok/sper person · mean · sample count not recorded
Machine totalaggregate_e2e_tps · tok/s: 53.59 tok/smachine total · single · sample count not recorded
Peak memory, whole hostpeak_host_unavailable_gib · GiB: 27.52 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
Date	Shape	Each person · median	Machine total	First word · median	Result	Run
2026-10-05	2 people	not recorded	53.59 tok/s	not recorded	Partial	Run d92bd079 →
2026-10-05	2 people	not recorded	53.59 tok/s	not recorded	Not graded	Run 08b6b669 →
2026-10-05	1 people	not recorded	not recorded	not recorded	Not graded	Run 71cdb415 →
2026-10-05	unknown people	not recorded	not recorded	not recorded	Failed	Run ee7ff82a →
2026-10-05	unknown people	not recorded	not recorded	not recorded	Failed	Run e07e13ba →

Answer checks

6/6 request checks

Method

One warmup then three measured requests each for code and chat; first-content-to-DONE decode; aggregate uses actual paired request windows including TTFT.

Limits

Short repeated prompts; prefix-slot reuse may improve TTFT. Task correctness and long-context quality unassessed; whole-host memory includes baseline.

What ran

Exact identity

Original publisher model: Qwen3.8-27B
Publisher revision: 1d4bf0f2ff6012fd82039f2fa52739d0dd7c60c0
Variant · format · precision: GSQ-RCO IQ3_S GGUF variant · GGUF · GSQ-RCO IQ3_S
Weights SHA-256: not recorded
Engine · version · feature: llama.cpp · native CUDA 13, sm_121a · 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

People in this run: 2
Configured window (tokens): 65536
Output token ceiling: 512

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
Metric	Value	Scope	Statistic	Sample count	Basis / source field	Run
chat decode tps	26.99 tok/s	per person	mean	not recorded	chat_decode_tps	Run 08b6b669 →
requests ok	6.00 count	machine total	single	not recorded	requests_ok	Run 08b6b669 →
code ttft mean s	0.30 s	per person	mean	not recorded	code_ttft_mean_s	Run 08b6b669 →
aggregate concurrency	2.00 count	machine total	single	not recorded	aggregate_concurrency	Run 08b6b669 →
requests total	6.00 count	machine total	single	not recorded	requests_total	Run 08b6b669 →
aggregate e2e tps	53.59 tok/s	machine total	single	not recorded	aggregate_tps	Run 08b6b669 →
code decode tps	33.64 tok/s	per person	mean	not recorded	code_decode_tps	Run 08b6b669 →
chat ttft mean s	0.25 s	per person	mean	not recorded	chat_ttft_mean_s	Run 08b6b669 →
peak host unavailable gib	27.52 GiB	machine total	max	not recorded	peak_host_unavailable_gib	Run 08b6b669 →

Data and citation

Download this run · OLIR JSON All numbers · CSV OLIR schema

Skilak Keep evidence, run 08b6b6698d9424baa272c369, measured 2026-10-05T18:20:22.433708+00:00.
https://skilak.ai/keep/resources/evidence/r/08b6b6698d9424baa272c369

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.

Request a pilot