Qwen3.8 27B on M5 Max · 64 GiB

Experiment Unsupported Published

uzu 0.6.17 on M5 Max · 64 GiB, Metal

Qwen3.8 27B as mirai-m (4-bit), archive 256 input / 128 output: a 256-token prompt and 128 tokens out. Measured 9 Nov 2026, 23:53 UTC by Glenn Sonna, Xybrid.

Doesn’t count in rankings: uzu 0.6.17 renders its own Qwen3.8 chat template: 67 input tokens for the 27-token canonical check prompt, so it can't run the frozen prompts (measured in qualification).

Experiment id s2-qwen3.8-27b-4-bit-uzu-256

Unsupported configuration

uzu 0.6.17 renders its own Qwen3.8 chat template: 67 input tokens for the 27-token canonical check prompt, so it can't run the frozen prompts (measured in qualification).

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
uzu
Maintainer
Mirai
License
MIT, open source
Version
0.6.17
Commit
Not recorded
Binary SHA-256
7829db228dc9c1a80eb71f2146724ab39c37a2dea37aa760b0cb1558b88e4a7c
Built from
package
Build flags
None
Backend
Metal
Threads
Not recorded
Interface
python-worker
Settings in effect
unsupported=uzu 0.6.17 renders its own Qwen3.8 chat template: 67 input tokens for the 27-token canonical check prompt, so it can't run the frozen prompts (measured in qualification).

Model and weights

Base model
Qwen3.8 27B (Qwen3.8, 27B)
Upstream
Qwen/Qwen3.8-27B@37561816e66b5b9153b1b64c10968c18c1e6f4a0
Weights
trymirai/Qwen3.8-27B-M@c991a5c5cd99d9528e9f3e83302e82c40f1436ab
Format
mirai
File 1
README.md · 3,444 bytes · sha256 b6955d23742ad1fe12c852cef7124e52ab0e490d5d7155a1aef56898b705c01d
File 2
config.json · 140,613 bytes · sha256 60759c66454c7f48c53bdc7e932d685a80f846e3f956aecbb5ac9374ef447500
File 3
model.safetensors · 13.53 GiB · sha256 79e170fdb07a5410e4e0859ddd302cd76acecec91142030bfeb1eb533192f9a3
File 4
tokenizer.json · 19.1 MiB · sha256 87a7830d63fcf43bf241c3c5242e96e62dd3fdc29224ca26fed8ea333db72de4
Quantization
mirai-m (4-bit)
Weight bits
4
Group size
Not recorded
KV cache
uzu default
Conversion
Published upstream by trymirai
Conversion command
Not recorded

Workload

Workload
Archive 256 input / 128 output
Prompt
256 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
Not recorded

Provenance

Measured by
Glenn Sonna, Xybrid
Operator affiliation
Xybrid
Run at
9 Nov 2026, 23:53 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
Not run
Method
Not recorded
Reference
Not recorded
Reference SHA-256
Not recorded
Required passing checks
Not recorded
Detail
Not recorded

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