LFM2.5 350M on M5 Max · 64 GiB

Experiment Complete Published

ExecuTorch 1.6.0 on M5 Max · 64 GiB, Metal

LFM2.5 350M as mlx-4w-g64 (4-bit), archive 256 input / 128 output: a 256-token prompt and 128 tokens out. Measured 12 Nov 2026, 20:40 UTC by Glenn Sonna, Xybrid.

Doesn’t count in rankings: Its output check failed: 6/8 of the checks the model passes at BF16; needs 7 and the mandatory ones (needle-start, needle-end). Left out, failed by every BF16 configuration: arithmetic, sort, word, continuation-128.

Experiment id s3-lfm2.5-350m-4-bit-et-mlx-256

Summary

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

MetricMedianRangeMean ± SDQuartiles
Decode speed tok/s, higher is better786.4 779.1–788.8 785.6 ± 4.0 784.0–788.8
Time to first token ms, lower is better23.4 23.1–23.5 23.4 ± 0.1 23.3–23.4
Peak memory GiB, lower is betterNot measured
Prefill speed tok/s, higher is better11,130 10,667–11,130 11,053 ± 189.3 11,130–11,130

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
123.311,130256 tok in 23.0 ms, engine788.8127 tok in 161 ms, enginenot recordednominal → nominal
223.410,667256 tok in 24.0 ms, engine788.8127 tok in 161 ms, enginenot recordednominal → nominal
323.111,130256 tok in 23.0 ms, engine784.0127 tok in 162 ms, enginenot recordednominal → nominal
423.411,130256 tok in 23.0 ms, engine788.8127 tok in 161 ms, enginenot recordednominal → nominal
523.411,130256 tok in 23.0 ms, engine784.0127 tok in 162 ms, enginenot recordednominal → nominal
623.511,130256 tok in 23.0 ms, engine779.1127 tok in 163 ms, enginenot recordednominal → 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
ExecuTorch
Maintainer
PyTorch (Meta)
License
BSD-3-Clause, open source
Version
1.6.0
Commit
Not recorded
Binary SHA-256
5ffb7850c43a0ee6d679ba2d94973c6060256a9f05ac14067496498b24ff9095
Built from
package
Build flags
None
Backend
Metal
Threads
Not recorded
Interface
python-worker
Settings in effect
backend=metal recipe=lfm2.5-350m-et-mlx-4w,lfm2_mlx_4w.yaml,lfm2.5-350m-et-mlx-4w.pte,mlx-4w-g64 delegates=MLXBackend echo=false temperature=0 num_bos=0 num_eos=0 reset_before_generation=true eos_ignorable=false

Model and weights

Base model
LFM2.5 350M (LFM2.5, 350M)
Upstream
LiquidAI/LFM2.5-350M@9921cd4d82c0e3986887aa4b376f9446729ef8a3
Weights
LiquidAI/LFM2.5-350M@9921cd4d82c0e3986887aa4b376f9446729ef8a3
Format
pte
File 1
lfm2.5-350m-et-mlx-4w.pte · 307.8 MiB · sha256 43c671ba6bde0a9ff826bbbcac508a061a65e22a19fb7c2cd9e7587e98b7d368
File 2
tokenizer.json · 4.5 MiB · sha256 df1d8d5ec5d091b460562ffd545e4a5e91d17d4a0db7ebe733be34ed374377bd
Quantization
mlx-4w-g64 (4-bit)
Weight bits
4
Group size
64
KV cache
bf16
Conversion
Converted by Xybrid with executorch export_llm 1.6.0
Conversion command
cd "$ROOT/engines/executorch/build/lfm2.5-350m-et-mlx-4w" && PYTHONPATH="$ROOT/engines/executorch/.venv/lib/python3.12/site-packages/executorch" "$ROOT/engines/executorch/.venv/bin/python" -m executorch.extension.llm.export.export_llm --config "$ROOT/engines/executorch/.venv/lib/python3.12/site-packages/executorch/examples/models/lfm2/config/lfm2_mlx_4w.yaml" +base.model_class=lfm2_5_350m "+base.checkpoint=$ROOT/engines/executorch/build/lfm2.5-350m-canonical.pth" "+base.params=$ROOT/downloads/executorch/4776f71b6f0daa5dfc7d758d5836fc1cc0d40154/examples/models/lfm2/config/lfm2_5_350m_config.json" +export.max_seq_length=4096 +export.max_context_length=4096 +export.output_name=lfm2.5-350m-et-mlx-4w.pte

Workload

Workload
Archive 256 input / 128 output
Prompt
256 tokens, from prompts/source.txt
Prompt SHA-256
c10d1b7bef62d95322f8f8d229dc66118d9375107bdc49b06f76287d058b2f6c
Chat template
ba551d58630afa3190b1be3602e28301f3d2e9bbac978dfc49d6d825171648b6
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
12 Nov 2026, 20:40 UTC
Harness
xybrid-bench 0.4.0 (f67b01e5a1a77cc89f5984f23e35dd4590fa1bec)
Methodology
v0.6
Publication
published
Published
16 Nov 2026, 10:23 UTC
Reviewed by
Glenn Sonna

Output check

Result
Failed
Method
sanity-v1-bf16-reference
Reference
checks/sanity-v1.json
Reference SHA-256
a702dee079525ae0ec90465b1c9e26052c99cb3e1c53e83dea05be4c194fc7ca
Required passing checks
7
Check arithmetic
Failed · sha256 9f14025af0065b30e47e23ebb3b491d39ae8ed17d33739e5ff3827ffb3634953
Check capital
Passed · sha256 5dd272b4f316b776a7b8e3d0894b37e1e42be3d5d3b204b8a5836cc50597a6b1
Check opposite
Passed · sha256 f151cd9e12c8994edfa316e4624f956bc68460454869177f287146f8cc956a12
Check weekday
Passed · sha256 7d8af1de1262f150187d5938bf649d6c35726970685f916d1ec6392a801b9762
Check color
Failed · sha256 ec7d56a01607001e6401366417c5e2eb00ffa0df17ca1a9a831e0b32c8f11bf7
Check sort
Failed · sha256 edb794d615a63f5eec436d83bed7299d6af36f508c0bf38a3c1cd85d6e4ebd56
Check animal
Failed · sha256 d4735e3a265e16eee03f59718b9b5d03019c07d8b6c51f90da3a666eec13ab35
Check word
Failed · sha256 f61e68691bdd0f5ebd3252de0396569d624bd2fba240d66fae90edb456a9f48f
Check copy
Passed · sha256 d61e6890d0e3571b0c7693612d41f8d059a1fcdef4ba959c0af4144541d41907
Check needle-start
Passed · sha256 8b9af2e5b8b4cfddb87d171997b95c471f8d68a68e30fb5177abeb3332372edd
Check needle-end
Passed · sha256 e6589c736af32f81bf239c5eda6de41959122ccc1d05463b75e0eb0c70fc831c
Check continuation-128
Failed · sha256 d7f917cdf362c04c176faac2a3a39c878a91838b9152de3ff5b965c02ccc8763
Detail
6/8 of the checks the model passes at BF16; needs 7 and the mandatory ones (needle-start, needle-end). Left out, failed by every BF16 configuration: arithmetic, sort, word, continuation-128.

Reproduce it

Commands
pnpm bench run --plan plans/s3-lfm2.5-350m-4-bit.json

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

Raw data