Qwen3.8 27B on M5 Max · 64 GiB

Experiment Complete Published

Ollama 0.40.13 on M5 Max · 64 GiB, Metal

Qwen3.8 27B as Q4_K_M (4-bit), archive 2048 input / 128 output: a 2048-token prompt and 128 tokens out. Measured 10 Nov 2026, 00:08 UTC by Glenn Sonna, Xybrid.

Eligible for ranking when the objective's required metrics are measured and this configuration is selected.Approximate comparison: The experiments use different effective KV cache precision: f16, model-dtype.

Experiment id s2-qwen3.8-27b-4-bit-ollama-2048

Summary

Over 6 valid trials of 6 attempted (6 planned), computed from the raw trials below.

MetricMedianRangeMean ± SDQuartiles
Decode speed tok/s, higher is better25.9 25.8–25.9 25.9 ± 0.0 25.9–25.9
Time to first token ms, lower is better2,823 2,749–3,328 2,897 ± 221 2,758–2,897
Peak memory GiB, lower is better19.07 19.07–19.08 19.07 ± 0.01 19.07–19.08
Prefill speed tok/s, higher is better727.0 616.2–746.3 711.3 ± 49.6 708.0–743.8

Trials, metric by metric

The line runs from the lowest to the highest valid trial, the dot is the median and each tick is one trial.

Every trial

Raw token counts and durations, with the speeds computed from them. Invalid trials stay in the record with their reason; they are left out of the summary, never silently dropped.

Every attempted trial
TrialTime to first token msPrefill tok/sDecode tok/sPeak memory GiBThermalNote
13,328616.22,048 tok in 3,323 ms, engine25.8127 tok in 4,926 ms, engine19.07nominal → nominal
22,749746.32,048 tok in 2,744 ms, engine25.9127 tok in 4,905 ms, engine19.08nominal → nominal
32,878712.92,048 tok in 2,873 ms, engine25.9127 tok in 4,908 ms, engine19.08nominal → nominal
42,904706.42,048 tok in 2,899 ms, engine25.9127 tok in 4,906 ms, engine19.08nominal → nominal
52,755744.72,048 tok in 2,750 ms, engine25.9127 tok in 4,904 ms, engine19.07nominal → nominal
62,768741.22,048 tok in 2,763 ms, engine25.9127 tok in 4,902 ms, engine19.07nominal → nominal

Caveats

  • Publisher cards identify the base model but do not always pin the exact export source checkpoint revision.
  • Engines may differ in quantization scheme, effective KV cache precision, kernels, and activations.

Configuration

Device

Device
Apple Mac17,6
Cohort
Apple silicon Macs (desktop)
SoC
Apple M5 Max
CPU
18-core Apple CPU
GPU
Not recorded
NPU
Not recorded
RAM
64.00 GiB
OS
macOS 27.2.0 (26B123)
Kernel
27.2.0
Power
mains
Low power mode
No
Performance mode
Not recorded

Engine

Engine
Ollama
Maintainer
Ollama
License
MIT, open source
Version
0.40.13
Commit
Not recorded
Binary SHA-256
270089e59e1d411ac4d0987154b4e95f48b29a81c9d274833c6b766d6df88300
Built from
release
Build flags
None
Backend
Metal
Threads
Not recorded
Interface
http-server
Settings in effect
backend=metal model=xybrid-d15d4b6de55f662a blob=sha256-d15d4b6de55f662ac40e867702d31afe738d032fa06b829b2b1a21a49d4a48ee offloaded_layers=65 bos_stripped=false llama_server.host=127.0.0.1 llama_server.no-webui=true llama_server.offline=true llama_server.c=4096 llama_server.np=1 llama_server.log-verbosity=4 llama_server.no-log-prefix=true llama_server.no-log-timestamps=true llama_server.cache-type-k=f16 llama_server.cache-type-v=f16 llama_server.flash-attn=on llama_server.b=512 llama_server.ub=512 llama_server.context-shift=true llama_server.keep=4 env.OLLAMA_KEEP_ALIVE=-1 env.OLLAMA_FLASH_ATTENTION=1 env.OLLAMA_KV_CACHE_TYPE=f16 env.OLLAMA_NOHISTORY=1 env.OLLAMA_NUM_PARALLEL=1 env.OLLAMA_MAX_LOADED_MODELS=1 env.LLAMA_ARG_CACHE_RAM=0

Model and weights

Base model
Qwen3.8 27B (Qwen3.8, 27B)
Upstream
Qwen/Qwen3.8-27B@37561816e66b5b9153b1b64c10968c18c1e6f4a0
Weights
ggml-org/Qwen3.8-27B-GGUF@8ccda149192dec5f41450a402e66889e5fd25a24
Format
gguf
File 1
Qwen3.8-27B-Q4_K_M.gguf · 18.07 GiB · sha256 d15d4b6de55f662ac40e867702d31afe738d032fa06b829b2b1a21a49d4a48ee
File 2
README.md · 488 bytes · sha256 7ea038c050ddf1d40bbff2b74de4e58cd480616ec9c6db1e3a0e07a0feb631d5
Quantization
Q4_K_M (4-bit)
Weight bits
4
Group size
Not recorded
KV cache
f16
Conversion
Published upstream by ggml-org
Conversion command
Not recorded

Workload

Workload
Archive 2048 input / 128 output
Prompt
2,048 tokens, from prompts/source.txt
Prompt SHA-256
c10d1b7bef62d95322f8f8d229dc66118d9375107bdc49b06f76287d058b2f6c
Chat template
af16f654968d47d09a6c604e00370b12dbc21f76456f59c9e98b82e84d9f6375
Output
128 tokens, ignoring end of sequence
Context length
4,096
Batch size
1
Sampling
Greedy
Timing
One streamed request per trial
Warmup
2 runs, not counted
Trials planned
6
Cooldown
15 s between trials

Conditions

Ambient
Not recorded
Thermal at start
nominal
Thermal at end
nominal
Thermal gate
macOS thermal state nominal
Throttling detected
No

Provenance

Measured by
Glenn Sonna, Xybrid
Operator affiliation
Xybrid
Run at
10 Nov 2026, 00:08 UTC
Harness
xybrid-bench 0.4.0 (ccb0a8ac606e19454c16dd0b625e29509c726b72)
Methodology
v0.6
Publication
published
Published
16 Nov 2026, 10:23 UTC
Reviewed by
Glenn Sonna

Output check

Result
Passed
Method
sanity-v1-bf16-reference
Reference
checks/sanity-v1.json
Reference SHA-256
a702dee079525ae0ec90465b1c9e26052c99cb3e1c53e83dea05be4c194fc7ca
Required passing checks
11
Check arithmetic
Passed · sha256 6b51d431df5d7f141cbececcf79edf3dd861c3b4069f0b11661a3eefacbba918
Check capital
Passed · sha256 5dd272b4f316b776a7b8e3d0894b37e1e42be3d5d3b204b8a5836cc50597a6b1
Check opposite
Passed · sha256 f151cd9e12c8994edfa316e4624f956bc68460454869177f287146f8cc956a12
Check weekday
Passed · sha256 7d8af1de1262f150187d5938bf649d6c35726970685f916d1ec6392a801b9762
Check color
Passed · sha256 d486dfbd5fb578340bccbdd0a662527eab38550648d5f44517e5ac71b8824703
Check sort
Passed · sha256 a386b3c9b4b4786df5bb6474bab0a62b8476e2f3f9c8a6433aca40152840f6b7
Check animal
Passed · sha256 4b227777d4dd1fc61c6f884f48641d02b4d121d3fd328cb08b5531fcacdabf8a
Check word
Passed · sha256 3b635d3b4437e023e72a9df26201b6ab51285b376446446ebb379121994550f6
Check copy
Passed · sha256 d61e6890d0e3571b0c7693612d41f8d059a1fcdef4ba959c0af4144541d41907
Check needle-start
Passed · sha256 8b9af2e5b8b4cfddb87d171997b95c471f8d68a68e30fb5177abeb3332372edd
Check needle-end
Passed · sha256 e6589c736af32f81bf239c5eda6de41959122ccc1d05463b75e0eb0c70fc831c
Check continuation-128
Passed · sha256 ed26731fe8d39dacac77703998aa4bb784317ae565cb21a58f83f75d7e458142
Detail
12/12; needs 11 and every mandatory check (needle-start, needle-end, continuation-128); the model has no BF16 reference.

Reproduce it

Commands
pnpm bench run --plan plans/s2-qwen3.8-27b-4-bit.json

Three palindromic blocks per quantization class and workload; load, two warmups, one trial, unload per turn.

Raw data