LFM2.5 1.2B Instruct on M5 Max · 64 GiB

Experiment Complete Published

LiteRT-LM 0.19.1 on M5 Max · 64 GiB, GPU

LFM2.5 1.2B Instruct as litert-int4-gpu (4-bit), archive 256 input / 128 output: a 256-token prompt and 128 tokens out. Measured 12 Nov 2026, 21:44 UTC by Glenn Sonna, Xybrid.

Doesn’t count in rankings: Its output check failed: 8/12 of the checks the model passes at BF16; needs 11 and the mandatory ones (needle-start, needle-end, continuation-128). Left out, failed by every BF16 configuration: none.

Experiment id s3-lfm2.5-1.2b-4-bit-litert-gpu-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 better330.8 327.6–340.2 331.9 ± 4.5 329.3–332.6
Time to first token ms, lower is better114 104–117 113 ± 4.9 112–117
Peak memory GiB, lower is better1.82 1.82–1.83 1.82 ± 0.00 1.82–1.83
Prefill speed tok/s, higher is better3,438 3,380–3,497 3,438 ± 51.2 3,394–3,481

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
11123,493256 tok in 73.3 ms, engine332.1127 tok in 382 ms, engine1.83nominal → nominal
21123,497256 tok in 73.2 ms, engine332.8127 tok in 382 ms, engine1.82nominal → nominal
31163,445256 tok in 74.3 ms, engine329.6127 tok in 385 ms, engine1.82nominal → nominal
41173,382256 tok in 75.7 ms, engine329.2127 tok in 386 ms, engine1.82nominal → nominal
51043,380256 tok in 75.7 ms, engine340.2127 tok in 373 ms, engine1.83nominal → nominal
61173,431256 tok in 74.6 ms, engine327.6127 tok in 388 ms, engine1.83nominal → 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
LiteRT-LM
Maintainer
Google AI Edge
License
Apache-2.0, open source
Version
0.19.1
Commit
Not recorded
Binary SHA-256
c61e0faa71026f2875a710a42cbe803226004db811f7ee47ea78198a0499e851
Built from
package
Build flags
None
Backend
GPU
Threads
Not recorded
Interface
python-worker
Settings in effect
backend=gpu max_num_tokens=4096 enable_benchmark=true chat_template=pass-through process=one per generation sampler=top_k=1 temperature=0 stop_token_ids=[[7], [2]] cache_dir=[runner]/engines/litert-lm/cache bos_token_id=1

Model and weights

Base model
LFM2.5 1.2B Instruct (LFM2.5, 1.2B)
Upstream
LiquidAI/LFM2.5-1.2B-Instruct@b9d816ab510f9a28dca37cb182e9da8b0226e13c
Weights
litert-community/LFM2.5-1.2B-Instruct@4a148a1032ce264346c6a2a2931e8d0391dbe5d1
Format
litertlm
File 1
LFM2.5-1.2B-Instruct_int4_gpu.litertlm · 702.1 MiB · sha256 8d3a4d73e121166ad377e9c51148d38ee383c223661f4219e01f22a0fe1d4a1a
File 2
README.md · 20,979 bytes · sha256 531c7c0863478b158f7cdf91c3af2a51b26bcd0081b3c976b9aeabe9280dbab6
Quantization
litert-int4-gpu (4-bit)
Weight bits
4
Group size
Not recorded
KV cache
litert-lm default
Conversion
Published upstream by litert-community
Conversion command
Not recorded

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, 21:44 UTC
Harness
xybrid-bench 0.4.0 (a101f5a635e0541dbb6a1f784441a4634584f414)
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
11
Check arithmetic
Failed · sha256 7902699be42c8a8e46fbbb4501726517e86b22c56a189f7625a6da49081b2451
Check capital
Passed · sha256 5dd272b4f316b776a7b8e3d0894b37e1e42be3d5d3b204b8a5836cc50597a6b1
Check opposite
Passed · sha256 f151cd9e12c8994edfa316e4624f956bc68460454869177f287146f8cc956a12
Check weekday
Passed · sha256 7d8af1de1262f150187d5938bf649d6c35726970685f916d1ec6392a801b9762
Check color
Failed · sha256 ec7d56a01607001e6401366417c5e2eb00ffa0df17ca1a9a831e0b32c8f11bf7
Check sort
Passed · sha256 a386b3c9b4b4786df5bb6474bab0a62b8476e2f3f9c8a6433aca40152840f6b7
Check animal
Passed · sha256 4b227777d4dd1fc61c6f884f48641d02b4d121d3fd328cb08b5531fcacdabf8a
Check word
Passed · sha256 3b635d3b4437e023e72a9df26201b6ab51285b376446446ebb379121994550f6
Check copy
Passed · sha256 d61e6890d0e3571b0c7693612d41f8d059a1fcdef4ba959c0af4144541d41907
Check needle-start
Failed · sha256 f24f1a64b591544a871284bdde332d3c5d2cb109d21c03122c57d768e7c535b1
Check needle-end
Passed · sha256 e6589c736af32f81bf239c5eda6de41959122ccc1d05463b75e0eb0c70fc831c
Check continuation-128
Failed · sha256 23fef0ff1c4f19489c3ceaf745b0158027609b06cd8aa3a8eb2035908be56eab
Detail
8/12 of the checks the model passes at BF16; needs 11 and the mandatory ones (needle-start, needle-end, continuation-128). Left out, failed by every BF16 configuration: none.

Reproduce it

Commands
pnpm bench run --plan plans/s3-lfm2.5-1.2b-4-bit.json

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

Raw data