Experiment Complete Published
uzu 0.6.17 on iPhone 12 · 4 GB, Metal
LFM2.5 350M as mirai-l (8-bit), archive 2048 input / 128 output: a 2048-token prompt and 128 tokens out. Measured 12 Nov 2026, 22:41 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, uzu default. The experiments here use different 8-bit schemes: Q8_0, mirai-l.
Experiment id s4-iphone12-lfm2.5-350m-8-bit-uzu-2048
Summary
Over 6 valid trials of 6 attempted (6 planned), computed from the raw trials below.
| Metric | Median | Range | Mean ± SD | Quartiles |
|---|---|---|---|---|
| Decode speed tok/s, higher is better | 70.2 | 68.8–71.0 | 70.1 ± 0.8 | 69.8–70.8 |
| Time to first token ms, lower is better | 2,356 | 2,346–2,395 | 2,364 ± 18.6 | 2,353–2,373 |
| Peak memory GiB, lower is better | Not measured | |||
| Prefill speed tok/s, higher is better | 895.5 | 894.2–897.8 | 895.9 ± 1.5 | 894.8–897.1 |
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.
| Trial | Time to first token ms | Prefill tok/s | Decode tok/s | Peak memory GiB | Thermal | Note |
|---|---|---|---|---|---|---|
| 1 | 2,346 | 897.42,048 tok in 2,282 ms, engine | 71.0127 tok in 1,789 ms, engine | not recorded | nominal → nominal | |
| 2 | 2,359 | 894.92,048 tok in 2,289 ms, engine | 70.2127 tok in 1,810 ms, engine | not recorded | nominal → nominal | |
| 3 | 2,354 | 894.82,048 tok in 2,289 ms, engine | 71.0127 tok in 1,790 ms, engine | not recorded | nominal → nominal | |
| 4 | 2,395 | 894.22,048 tok in 2,290 ms, engine | 68.8127 tok in 1,846 ms, engine | not recorded | nominal → nominal | |
| 5 | 2,352 | 897.82,048 tok in 2,281 ms, engine | 70.2127 tok in 1,809 ms, engine | not recorded | nominal → nominal | |
| 6 | 2,377 | 896.22,048 tok in 2,285 ms, engine | 69.6127 tok in 1,824 ms, engine | not recorded | nominal → 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 iPhone13,2
- Cohort
- iPhones (mobile)
- SoC
- Apple A14 Bionic
- CPU
- 6-core CPU (2 performance, 4 efficiency)
- GPU
- 4-core Apple GPU
- NPU
- 16-core Neural Engine
- RAM
- 4.00 GiB
- OS
- iOS 27.2.1 (24B156)
- Kernel
- Darwin Kernel Version 27.2.0: Tue Oct 13 21:44:36 PDT 2026; root:xnu-13432.42.3~28/RELEASE_ARM64_T8101
- Power
- usb
- Low power mode
- No
- App state
- active
Engine
- Engine
- uzu
- Maintainer
- Mirai
- License
- MIT, open source
- Version
- 0.6.17
- Commit
- Not recorded
- Binary SHA-256
- c2ad72f6c7e394f758cbc77dc71c868a9c3d52371bc1379a9e49e3940bc22586
- Built from
- package
- Build flags
- None
- Backend
- Metal
- Threads
- Not recorded
- Interface
- ios-app
- Settings in effect
- backend=metal app=XybridBench process=one app launch per generation threads=engine default session=new per generation sampling=greedy prompt=user content of the frozen chat envelope uzu_identifier=liquidai:lfm2.5:350m:mirai:mirai-l:8
Model and weights
- Base model
- LFM2.5 350M (LFM2.5, 350M)
- Upstream
- LiquidAI/LFM2.5-350M@9921cd4d82c0e3986887aa4b376f9446729ef8a3
- Weights
- trymirai/LFM2.5-350M-L@9825a0acca9efaa1c7e51fe5032a527aedace312
- Format
- mirai
- File 1
- README.md · 3,251 bytes · sha256 94617c9da79daf7253aee79a58cbaaf6ed7b63af67470a29a70107095908cfcf
- File 2
- config.json · 37,022 bytes · sha256 e9f941c136b12c2e1b848d8266f1e94c0a547ce61ec130013cdbb18447a375a7
- File 3
- model.safetensors · 350.2 MiB · sha256 7368e286e5d7370ae2ce3fdbd1592bc0740c9f318417ce202c0ab3fa8711cfc4
- File 4
- tokenizer.json · 4.5 MiB · sha256 df1d8d5ec5d091b460562ffd545e4a5e91d17d4a0db7ebe733be34ed374377bd
- Quantization
- mirai-l (8-bit)
- Weight bits
- 8
- Group size
- Not recorded
- KV cache
- uzu default
- Conversion
- Published upstream by trymirai
- 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
- 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
- 60 s between trials
Conditions
- Ambient
- Not recorded
- Thermal at start
- nominal
- Thermal at end
- nominal
- Thermal gate
- iOS thermal state nominal
- Throttling detected
- No
Provenance
- Measured by
- Glenn Sonna, Xybrid
- Operator affiliation
- Xybrid
- Run at
- 12 Nov 2026, 22:41 UTC
- Harness
- xybrid-bench 0.4.0 (94dc4661ed5b4b9c30dad9b9da3cb91d2f474255)
- 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
- 7
- Check arithmetic
- Failed · sha256 4a44dc15364204a80fe80e9039455cc1608281820fe2b24f1e5233ade6af1dd5
- Check capital
- Passed · sha256 5dd272b4f316b776a7b8e3d0894b37e1e42be3d5d3b204b8a5836cc50597a6b1
- Check opposite
- Passed · sha256 f151cd9e12c8994edfa316e4624f956bc68460454869177f287146f8cc956a12
- Check weekday
- Passed · sha256 7d8af1de1262f150187d5938bf649d6c35726970685f916d1ec6392a801b9762
- Check color
- Passed · sha256 d486dfbd5fb578340bccbdd0a662527eab38550648d5f44517e5ac71b8824703
- Check sort
- Failed · sha256 edb794d615a63f5eec436d83bed7299d6af36f508c0bf38a3c1cd85d6e4ebd56
- Check animal
- Passed · sha256 4b227777d4dd1fc61c6f884f48641d02b4d121d3fd328cb08b5531fcacdabf8a
- Check word
- Failed · sha256 df7e70e5021544f4834bbee64a9e3789febc4be81470df629cad6ddb03320a5c
- Check copy
- Passed · sha256 d61e6890d0e3571b0c7693612d41f8d059a1fcdef4ba959c0af4144541d41907
- Check needle-start
- Passed · sha256 8b9af2e5b8b4cfddb87d171997b95c471f8d68a68e30fb5177abeb3332372edd
- Check needle-end
- Passed · sha256 e6589c736af32f81bf239c5eda6de41959122ccc1d05463b75e0eb0c70fc831c
- Check continuation-128
- Failed · sha256 d7f917cdf362c04c176faac2a3a39c878a91838b9152de3ff5b965c02ccc8763
- Detail
- 8/8 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
pnpm bench run --plan plans/s4-iphone12-lfm2.5-350m-8-bit.json
Three palindromic blocks per quantization class and workload; load, two warmups, one trial, unload per turn.
Raw data
- This record, as committed JSON, schema v0.3.0
- Complete finalized session evidence raw ·
sha256 6b875b35fb78…