Phone results

Local speech-engine measurements on a physical iPhone.

iPhone 17 Pro · 2026-10-23

36 smoke runs · 35 OK · 1 failed · 7 engines

Hardware: iPhone18,1
System: iOS 27.2
App: 0.10.23 · Release
Build number: Unknown
Build commit: Unknown
Tested language: English only
WER: Unknown · no verified reference
Romanian: Untested
Per-model energy: Unknown

These smoke runs check execution and timing. They do not establish transcription accuracy or a model ranking.

Batch speed excludes preparation. Live audio is paced in real time, so live runs show latency instead of throughput. Medians use successful runs only.

First preparation is the first observed attempt for each engine, including failures. It can include download, compilation, and loading. Native load and compile times are reported separately and must not be added to preparation.

RAM is the sampled app-process footprint, not model-only memory. Sampling can miss peaks. CPU includes all app threads and can exceed 100%. System speech services and accelerator work are not measured by these process counters.

Engine comparison. Batch medians are not directly comparable when fixture coverage differs. Open an engine below for every run and its load state.
Engine	Runs	First preparation	First native load	First successful preparation	Native load · first success	Batch speed · median	Live final delay · median	Sampled RAM · max	Transcription CPU · median
Parakeet TDT v3 · FluidAudiofluidaudio-parakeet	6 OK · 0 failed	35.88 s#1 · OK · cold-load	13.71 s	35.88 s#1 · batch · cold-load	13.71 s	107.2×3 successful batch runsBatch fixtures: D01, D02, D03	190.53 ms3 successful live runs	1,168.3 MiB	Batch 79.5%Live 13.8%
Fluid Speech Mini · FluidAudiofluidaudio-fluid-speech-mini	6 OK · 0 failed	25.77 s#3 · OK · cold-load	10.29 s	25.77 s#3 · batch · cold-load	10.29 s	132.9×3 successful batch runsBatch fixtures: D01, D02, D03	180.48 ms3 successful live runs	503.4 MiB	Batch 84.9%Live 12.7%
Phonon-2 · FluidAudiofluidaudio-phonon	6 OK · 0 failed	86.13 s#5 · OK · cold-load	63.35 s	86.13 s#5 · batch · cold-load	63.35 s	118.3×3 successful batch runsBatch fixtures: D01, D02, D03	180.89 ms3 successful live runs	565.8 MiB	Batch 113%Live 17.6%
Apple SpeechTranscriberapple	6 OK · 0 failed	157.58 ms#7 · OK · warm	Unknown	157.58 ms#7 · batch · warm	Unknown	69.9×3 successful batch runsBatch fixtures: D01, D02, D03	119.88 ms3 successful live runs	200 MiB	Batch 8.3%Live 6%
Apple shipped adapterapple-shipped	3 OK · 0 failed	2 ms#9 · OK · warm	Unknown	2 ms#9 · batch · warm	Unknown	77×3 successful batch runsBatch fixtures: D01, D02, D03	Not tested0 successful live runs	200.8 MiB	Batch 7.9%Live Not tested
Underdog Bark 3.0 Clean · FluidAudiofluidaudio-underdog-bark-3.0-clean	3 OK · 0 failed	319.14 s#10 · OK · cold-load	316.08 s	319.14 s#10 · batch · cold-load	316.08 s	16.6×3 successful batch runsBatch fixtures: D01, D02, D03	Not tested0 successful live runs	2,363.9 MiB	Batch 108.1%Live Not tested
Phonon-2 · Fermionphonon	5 OK · 1 failed	13.51 s#11 · Failed · cold-load	Unknown	197.17 s#12 · live · cold-load	187.93 s	344.2×2 successful batch runsBatch fixtures: D02, D03	65.64 ms3 successful live runs	1,190.5 MiB	Batch 94.4%Live 12.4%

Individual runs

Parakeet TDT v3 · FluidAudio 6 runs

fluidaudio@0.17.5+FluidInference/parakeet-tdt-0.6b-v3-coreml@7dd20fe6b1797d35f5e3307e8b1732d9a178edfe

Resource sample interval: 250 ms.

Run timings. Unknown values were not recorded. Native load and compile are parts of preparation.
Run	Mode	State	Outcome	Preparation	Native load	Native compile	Transcription	Batch speed	First partial	Final delay
#1 · D018.98 s audio	Batch	cold-loadNot loaded	OK	35.88 s	13.71 s	Unknown	83.73 ms	107.2×	Not applicable	Not applicable
#2 · D018.98 s audio	Live	warmAlready loaded	OK	0.03 ms	Unknown	Unknown	9.16 s	Not applicable	2.01 s	190.53 ms
#14 · D0210.28 s audio	Batch	cold-loadNot loaded	OK	360.44 ms	358.23 ms	Unknown	96.36 ms	106.7×	Not applicable	Not applicable
#15 · D0210.28 s audio	Live	warmAlready loaded	OK	0.04 ms	Unknown	Unknown	10.44 s	Not applicable	2.04 s	183.42 ms
#33 · D0328.436 s audio	Batch	cold-loadNot loaded	OK	378.54 ms	375.94 ms	Unknown	189.04 ms	150.4×	Not applicable	Not applicable
#34 · D0328.436 s audio	Live	warmAlready loaded	OK	0.04 ms	Unknown	Unknown	28.63 s	Not applicable	2.01 s	226.5 ms

App-process RAM and CPU. Peak RAM is sampled. CPU includes all app threads, but excludes system speech services.
Run	Start RAM	After preparation	After transcription	Sampled peak	Whole-run CPU	Transcription CPU	Thermal maximum	Dropped audio	Committed regressions
#1 · D01	115.5 MiB	575.7 MiB	619.5 MiB	619.5 MiB	18.6%6.7 s CPU in 35.963 s	79.5%0.067 s CPU in 0.084 s	nominal	Not applicable	Not applicable
#2 · D01	617.6 MiB	617.6 MiB	608.2 MiB	621.3 MiB	13%1.188 s CPU in 9.167 s	12.9%1.186 s CPU in 9.166 s	nominal	0 s	0
#14 · D02	680.9 MiB	1,128 MiB	1,166.8 MiB	1,166.8 MiB	38.8%0.182 s CPU in 0.469 s	62.1%0.06 s CPU in 0.096 s	fair	Not applicable	Not applicable
#15 · D02	1,168.3 MiB	1,168.3 MiB	635.9 MiB	1,168.3 MiB	13.8%1.443 s CPU in 10.439 s	13.8%1.441 s CPU in 10.437 s	fair	0 s	0
#33 · D03	681.1 MiB	1,087.3 MiB	1,145.1 MiB	1,146.2 MiB	78.7%0.453 s CPU in 0.575 s	168.6%0.319 s CPU in 0.189 s	fair	Not applicable	Not applicable
#34 · D03	1,146.7 MiB	1,146.7 MiB	675.9 MiB	1,146.7 MiB	16.7%4.774 s CPU in 28.629 s	16.7%4.772 s CPU in 28.627 s	fair	0 s	0

Run IDs and fixture hashes

#1 · 89575f83-1128-4600-a1d1-258a0c0a88a3
D01 SHA-256: c09ae00386e8cb7f9450b2f535ac1b157cb72d1efbc35f8805810d72bc81ab7e
#2 · 713f84ee-270b-409a-a49c-2690cc6f8ad4
D01 SHA-256: c09ae00386e8cb7f9450b2f535ac1b157cb72d1efbc35f8805810d72bc81ab7e
#14 · 88b1f5c4-5e8c-4575-a272-9cd2397aee64
D02 SHA-256: b9432690b3648fba32878af9e4fe51cc1b8fdab7c0df15607f6ed0c145813e15
#15 · 29568d17-c85e-4c66-891d-860e3b8c6623
D02 SHA-256: b9432690b3648fba32878af9e4fe51cc1b8fdab7c0df15607f6ed0c145813e15
#33 · 29f37ee0-4e17-4aa3-95ea-5220d651895d
D03 SHA-256: 0465ae37184793e6a496f9b1b388b18978ff4d835edc444fa06943291bab61c4
#34 · 114094de-1ee6-4c9f-87df-772875def088
D03 SHA-256: 0465ae37184793e6a496f9b1b388b18978ff4d835edc444fa06943291bab61c4

Fluid Speech Mini · FluidAudio 6 runs

fluidaudio@0.17.5+FluidVoice-Labs/fluid-speech-mini-coreml@d81c3160dfe5a5e1f8f7267164a07d30617a3a1a

Resource sample interval: 250 ms.

Run timings. Unknown values were not recorded. Native load and compile are parts of preparation.
Run	Mode	State	Outcome	Preparation	Native load	Native compile	Transcription	Batch speed	First partial	Final delay
#3 · D018.98 s audio	Batch	cold-loadNot loaded	OK	25.77 s	10.29 s	Unknown	67.57 ms	132.9×	Not applicable	Not applicable
#4 · D018.98 s audio	Live	warmAlready loaded	OK	0.03 ms	Unknown	Unknown	9.15 s	Not applicable	2 s	180.48 ms
#18 · D0210.28 s audio	Batch	cold-loadNot loaded	OK	235.21 ms	232.91 ms	Unknown	88.53 ms	116.1×	Not applicable	Not applicable
#19 · D0210.28 s audio	Live	warmAlready loaded	OK	0.03 ms	Unknown	Unknown	10.43 s	Not applicable	2.01 s	175.24 ms
#25 · D0328.436 s audio	Batch	cold-loadNot loaded	OK	116.06 ms	113.4 ms	Unknown	171.99 ms	165.3×	Not applicable	Not applicable
#26 · D0328.436 s audio	Live	warmAlready loaded	OK	0.04 ms	Unknown	Unknown	28.6 s	Not applicable	2 s	202.33 ms

App-process RAM and CPU. Peak RAM is sampled. CPU includes all app threads, but excludes system speech services.
Run	Start RAM	After preparation	After transcription	Sampled peak	Whole-run CPU	Transcription CPU	Thermal maximum	Dropped audio	Committed regressions
#3 · D01	134.4 MiB	417.7 MiB	440.1 MiB	440.1 MiB	14.1%3.655 s CPU in 25.844 s	84.9%0.057 s CPU in 0.068 s	nominal	Not applicable	Not applicable
#4 · D01	437.7 MiB	437.7 MiB	399.7 MiB	437.7 MiB	12.6%1.153 s CPU in 9.156 s	12.6%1.151 s CPU in 9.154 s	nominal	0 s	0
#18 · D02	168.5 MiB	428.1 MiB	447.2 MiB	447.2 MiB	52.1%0.171 s CPU in 0.329 s	68.5%0.061 s CPU in 0.089 s	fair	Not applicable	Not applicable
#19 · D02	449.2 MiB	449.2 MiB	434.1 MiB	449.2 MiB	12.7%1.329 s CPU in 10.43 s	12.7%1.328 s CPU in 10.429 s	fair	0 s	0
#25 · D03	185.3 MiB	439 MiB	500.6 MiB	500.6 MiB	125.3%0.367 s CPU in 0.293 s	135%0.232 s CPU in 0.172 s	fair	Not applicable	Not applicable
#26 · D03	503.4 MiB	503.4 MiB	476 MiB	503.4 MiB	15.4%4.408 s CPU in 28.607 s	15.4%4.407 s CPU in 28.605 s	fair	0 s	0

Run IDs and fixture hashes

#3 · cde8f490-6c1b-4b4b-bc05-fc98ddf8e179
D01 SHA-256: c09ae00386e8cb7f9450b2f535ac1b157cb72d1efbc35f8805810d72bc81ab7e
#4 · 39517153-b80a-4933-b7c5-b18adb458a82
D01 SHA-256: c09ae00386e8cb7f9450b2f535ac1b157cb72d1efbc35f8805810d72bc81ab7e
#18 · 55348948-42d8-4f5b-8dc2-e2a17e0c7c14
D02 SHA-256: b9432690b3648fba32878af9e4fe51cc1b8fdab7c0df15607f6ed0c145813e15
#19 · a1debb22-63fc-491d-863b-eaadb248b5b8
D02 SHA-256: b9432690b3648fba32878af9e4fe51cc1b8fdab7c0df15607f6ed0c145813e15
#25 · b686b57d-be30-4ef9-9734-4495e806d1dd
D03 SHA-256: 0465ae37184793e6a496f9b1b388b18978ff4d835edc444fa06943291bab61c4
#26 · e8e5ea3b-da3f-4771-a917-a80a34dfad08
D03 SHA-256: 0465ae37184793e6a496f9b1b388b18978ff4d835edc444fa06943291bab61c4

Phonon-2 · FluidAudio 6 runs

fluidaudio@0.17.5+FluidInference/phonon-2-coreml@a812a0dfef205660787ef6234317ab79acf4d5d6

Resource sample interval: 250 ms.

Run timings. Unknown values were not recorded. Native load and compile are parts of preparation.
Run	Mode	State	Outcome	Preparation	Native load	Native compile	Transcription	Batch speed	First partial	Final delay
#5 · D018.98 s audio	Batch	cold-loadNot loaded	OK	86.13 s	63.35 s	Unknown	79.35 ms	113.2×	Not applicable	Not applicable
#6 · D018.98 s audio	Live	warmAlready loaded	OK	0.03 ms	Unknown	Unknown	9.08 s	Not applicable	1.53 s	117.36 ms
#16 · D0210.28 s audio	Batch	cold-loadNot loaded	OK	298.73 ms	296.48 ms	Unknown	86.92 ms	118.3×	Not applicable	Not applicable
#17 · D0210.28 s audio	Live	warmAlready loaded	OK	0.04 ms	Unknown	Unknown	10.43 s	Not applicable	1.53 s	180.89 ms
#30 · D0328.436 s audio	Batch	cold-loadNot loaded	OK	309.6 ms	307.21 ms	Unknown	160.95 ms	176.7×	Not applicable	Not applicable
#31 · D0328.436 s audio	Live	warmAlready loaded	OK	0.04 ms	Unknown	Unknown	28.64 s	Not applicable	1.51 s	223.19 ms

App-process RAM and CPU. Peak RAM is sampled. CPU includes all app threads, but excludes system speech services.
Run	Start RAM	After preparation	After transcription	Sampled peak	Whole-run CPU	Transcription CPU	Thermal maximum	Dropped audio	Committed regressions
#5 · D01	157 MiB	486.5 MiB	531.3 MiB	531.3 MiB	6.8%5.824 s CPU in 86.217 s	113%0.09 s CPU in 0.08 s	nominal	Not applicable	Not applicable
#6 · D01	529.7 MiB	531.8 MiB	492.1 MiB	532.2 MiB	33%3.002 s CPU in 9.089 s	33%2.996 s CPU in 9.083 s	nominal	0 s	0
#16 · D02	164.5 MiB	485.8 MiB	522.7 MiB	522.7 MiB	48.2%0.189 s CPU in 0.391 s	82.6%0.072 s CPU in 0.087 s	fair	Not applicable	Not applicable
#17 · D02	520.6 MiB	520.6 MiB	511.9 MiB	520.6 MiB	14.1%1.467 s CPU in 10.436 s	14%1.466 s CPU in 10.435 s	fair	0 s	0
#30 · D03	184.3 MiB	500.9 MiB	559.4 MiB	559.4 MiB	101.8%0.484 s CPU in 0.476 s	212.1%0.342 s CPU in 0.161 s	fair	Not applicable	Not applicable
#31 · D03	565.8 MiB	565.8 MiB	539.3 MiB	565.8 MiB	17.6%5.04 s CPU in 28.64 s	17.6%5.038 s CPU in 28.639 s	fair	0 s	0

Run IDs and fixture hashes

#5 · a2127158-5844-4b62-8777-52b42cd982c5
D01 SHA-256: c09ae00386e8cb7f9450b2f535ac1b157cb72d1efbc35f8805810d72bc81ab7e
#6 · 2d93fa58-0f7e-4fa0-b752-babdceb5c551
D01 SHA-256: c09ae00386e8cb7f9450b2f535ac1b157cb72d1efbc35f8805810d72bc81ab7e
#16 · da6a2621-b309-457e-93d6-9966f5308c57
D02 SHA-256: b9432690b3648fba32878af9e4fe51cc1b8fdab7c0df15607f6ed0c145813e15
#17 · dba2a0bb-4ce4-4690-a98e-86e3595feff4
D02 SHA-256: b9432690b3648fba32878af9e4fe51cc1b8fdab7c0df15607f6ed0c145813e15
#30 · e44a30c4-5a37-4e5d-b140-837abc521150
D03 SHA-256: 0465ae37184793e6a496f9b1b388b18978ff4d835edc444fa06943291bab61c4
#31 · e5e14317-b51d-4e73-97a6-100c4941d3cc
D03 SHA-256: 0465ae37184793e6a496f9b1b388b18978ff4d835edc444fa06943291bab61c4

Apple SpeechTranscriber 6 runs

system Version 27.2 (Build 24B115)

Resource sample interval: 250 ms.

Run timings. Unknown values were not recorded. Native load and compile are parts of preparation.
Run	Mode	State	Outcome	Preparation	Native load	Native compile	Transcription	Batch speed	First partial	Final delay
#7 · D018.98 s audio	Batch	warmLoaded state unknown	OK	157.58 ms	Unknown	Unknown	152.14 ms	59×	Not applicable	Not applicable
#8 · D018.98 s audio	Live	warmLoaded state unknown	OK	0.02 ms	Unknown	Unknown	9.09 s	Not applicable	3.89 s	120.52 ms
#20 · D0210.28 s audio	Batch	warmLoaded state unknown	OK	139.8 ms	Unknown	Unknown	147.12 ms	69.9×	Not applicable	Not applicable
#21 · D0210.28 s audio	Live	warmLoaded state unknown	OK	0.02 ms	Unknown	Unknown	10.37 s	Not applicable	3.89 s	119.88 ms
#35 · D0328.436 s audio	Batch	warmLoaded state unknown	OK	118.01 ms	Unknown	Unknown	335.09 ms	84.9×	Not applicable	Not applicable
#36 · D0328.436 s audio	Live	warmLoaded state unknown	OK	0.03 ms	Unknown	Unknown	28.52 s	Not applicable	3.88 s	109.23 ms

App-process RAM and CPU. Peak RAM is sampled. CPU includes all app threads, but excludes system speech services.
Run	Start RAM	After preparation	After transcription	Sampled peak	Whole-run CPU	Transcription CPU	Thermal maximum	Dropped audio	Committed regressions
#7 · D01	146 MiB	149.1 MiB	149.3 MiB	149.3 MiB	8.9%0.028 s CPU in 0.313 s	8.3%0.013 s CPU in 0.152 s	nominal	Not applicable	Not applicable
#8 · D01	150.9 MiB	150.9 MiB	136 MiB	150.9 MiB	6.3%0.569 s CPU in 9.095 s	6.2%0.566 s CPU in 9.092 s	nominal	0 s	0
#20 · D02	195.8 MiB	194.3 MiB	194.4 MiB	195.8 MiB	6.2%0.018 s CPU in 0.296 s	7.4%0.011 s CPU in 0.147 s	fair	Not applicable	Not applicable
#21 · D02	199.2 MiB	197.8 MiB	171.3 MiB	199.2 MiB	6%0.624 s CPU in 10.374 s	6%0.622 s CPU in 10.371 s	fair	0 s	0
#35 · D03	200 MiB	198.8 MiB	197.6 MiB	200 MiB	13%0.059 s CPU in 0.457 s	9.5%0.032 s CPU in 0.336 s	fair	Not applicable	Not applicable
#36 · D03	198.7 MiB	198.7 MiB	176.6 MiB	198.7 MiB	5.8%1.641 s CPU in 28.526 s	5.7%1.638 s CPU in 28.523 s	fair	0 s	0

Run IDs and fixture hashes

#7 · abd72335-352f-46aa-b711-c40c0c04a000
D01 SHA-256: c09ae00386e8cb7f9450b2f535ac1b157cb72d1efbc35f8805810d72bc81ab7e
#8 · 4049b50a-2a8a-4d11-8ce1-8c62aa80ec88
D01 SHA-256: c09ae00386e8cb7f9450b2f535ac1b157cb72d1efbc35f8805810d72bc81ab7e
#20 · 6c96846b-8498-4af1-868d-49d085930947
D02 SHA-256: b9432690b3648fba32878af9e4fe51cc1b8fdab7c0df15607f6ed0c145813e15
#21 · 1fb525f0-2e5c-4b0d-9374-92a2750f2237
D02 SHA-256: b9432690b3648fba32878af9e4fe51cc1b8fdab7c0df15607f6ed0c145813e15
#35 · dadbe7c0-8fd9-481b-b976-5121d7d59a98
D03 SHA-256: 0465ae37184793e6a496f9b1b388b18978ff4d835edc444fa06943291bab61c4
#36 · 67ba1e05-357e-41a8-bc61-504079d170d9
D03 SHA-256: 0465ae37184793e6a496f9b1b388b18978ff4d835edc444fa06943291bab61c4

Apple shipped adapter 3 runs

@react-native-ai/apple 0.12.0 (patched)

Resource sample interval: 250 ms.

Run timings. Unknown values were not recorded. Native load and compile are parts of preparation.
Run	Mode	State	Outcome	Preparation	Native load	Native compile	Transcription	Batch speed	First partial	Final delay
#9 · D018.98 s audio	Batch	warmLoaded state unknown	OK	2 ms	Unknown	Unknown	126.82 ms	70.8×	Not applicable	Not applicable
#22 · D0210.28 s audio	Batch	warmLoaded state unknown	OK	0.52 ms	Unknown	Unknown	133.44 ms	77×	Not applicable	Not applicable
#27 · D0328.436 s audio	Batch	warmLoaded state unknown	OK	0.19 ms	Unknown	Unknown	323.98 ms	87.8×	Not applicable	Not applicable

App-process RAM and CPU. Peak RAM is sampled. CPU includes all app threads, but excludes system speech services.
Run	Start RAM	After preparation	After transcription	Sampled peak	Whole-run CPU	Transcription CPU	Thermal maximum	Dropped audio	Committed regressions
#9 · D01	140.5 MiB	140.5 MiB	139.6 MiB	140.5 MiB	9%0.012 s CPU in 0.129 s	7.9%0.01 s CPU in 0.127 s	nominal	Not applicable	Not applicable
#22 · D02	173.6 MiB	172.1 MiB	172.4 MiB	173.6 MiB	8.1%0.011 s CPU in 0.134 s	7.4%0.01 s CPU in 0.134 s	fair	Not applicable	Not applicable
#27 · D03	199.8 MiB	199.8 MiB	200.8 MiB	200.8 MiB	9.3%0.03 s CPU in 0.325 s	9.2%0.03 s CPU in 0.325 s	fair	Not applicable	Not applicable

Run IDs and fixture hashes

#9 · 080c0c29-4c7c-4012-af36-1e9b5d2f479a
D01 SHA-256: c09ae00386e8cb7f9450b2f535ac1b157cb72d1efbc35f8805810d72bc81ab7e
#22 · 440b04b4-00a7-4534-b993-4505e1e36fff
D02 SHA-256: b9432690b3648fba32878af9e4fe51cc1b8fdab7c0df15607f6ed0c145813e15
#27 · 91046866-231f-4016-ac79-5d06d5f60afa
D03 SHA-256: 0465ae37184793e6a496f9b1b388b18978ff4d835edc444fa06943291bab61c4

Underdog Bark 3.0 Clean · FluidAudio 3 runs

fluidaudio@0.17.5+ConwayResearch/Underdog-Bark-3.0-Clean@6d34ed8639b2b9a33e764acf06748519bbb950cf+bark-clean-fp16-fluidaudio-0.17.5-v1

Resource sample interval: 250 ms.

Run timings. Unknown values were not recorded. Native load and compile are parts of preparation.
Run	Mode	State	Outcome	Preparation	Native load	Native compile	Transcription	Batch speed	First partial	Final delay
#10 · D018.98 s audio	Batch	cold-loadNot loaded	OK	319.14 s	316.08 s	Unknown	609.66 ms	14.7×	Not applicable	Not applicable
#13 · D0210.28 s audio	Batch	cold-loadNot loaded	OK	14.94 s	14.94 s	Unknown	619.11 ms	16.6×	Not applicable	Not applicable
#32 · D0328.436 s audio	Batch	cold-loadNot loaded	OK	16.67 s	16.67 s	Unknown	1.52 s	18.7×	Not applicable	Not applicable

App-process RAM and CPU. Peak RAM is sampled. CPU includes all app threads, but excludes system speech services.
Run	Start RAM	After preparation	After transcription	Sampled peak	Whole-run CPU	Transcription CPU	Thermal maximum	Dropped audio	Committed regressions
#10 · D01	141.5 MiB	1,688.5 MiB	1,094.6 MiB	2,334.2 MiB	3.6%11.367 s CPU in 319.756 s	108.1%0.659 s CPU in 0.61 s	fair	Not applicable	Not applicable
#13 · D02	158.1 MiB	1,855.2 MiB	1,108.9 MiB	2,177.7 MiB	44.5%6.924 s CPU in 15.57 s	107.8%0.667 s CPU in 0.619 s	fair	Not applicable	Not applicable
#32 · D03	187 MiB	1,841 MiB	1,141.5 MiB	2,363.9 MiB	41.8%7.614 s CPU in 18.2 s	117%1.776 s CPU in 1.518 s	fair	Not applicable	Not applicable

Run IDs and fixture hashes

#10 · 0f5dc740-e12b-43ce-902e-b83cb2ffeaa3
D01 SHA-256: c09ae00386e8cb7f9450b2f535ac1b157cb72d1efbc35f8805810d72bc81ab7e
#13 · 57042c8d-8f47-44ab-b0ba-19617e1b7d19
D02 SHA-256: b9432690b3648fba32878af9e4fe51cc1b8fdab7c0df15607f6ed0c145813e15
#32 · 69ba678e-7c5f-43a6-babe-ea8ac0644d55
D03 SHA-256: 0465ae37184793e6a496f9b1b388b18978ff4d835edc444fa06943291bab61c4

Phonon-2 · Fermion 6 runs

phonon-coreml 1.1.1; FermionResearch/Phonon-2-CoreML@e931079

Resource sample interval: 250 ms.

Run timings. Unknown values were not recorded. Native load and compile are parts of preparation.
Run	Mode	State	Outcome	Preparation	Native load	Native compile	Transcription	Batch speed	First partial	Final delay
#11 · D018.98 s audio	Batch	cold-loadNot loaded	Failedpreparation-failed	13.51 s	Unknown	Unknown	Unknown	Not measured	Not applicable	Not applicable
#12 · D018.98 s audio	Live	cold-loadNot loaded	OK	197.17 s	187.93 s	669.47 ms	9.03 s	Not applicable	1.01 s	65.64 ms
#23 · D0210.28 s audio	Batch	cold-loadNot loaded	OK	1.31 s	1.31 s	Unknown	36.42 ms	282.2×	Not applicable	Not applicable
#24 · D0210.28 s audio	Live	warmAlready loaded	OK	0.02 ms	Unknown	Unknown	10.31 s	Not applicable	1.01 s	61.48 ms
#28 · D0328.436 s audio	Batch	cold-loadNot loaded	OK	737.64 ms	735.57 ms	Unknown	70 ms	406.2×	Not applicable	Not applicable
#29 · D0328.436 s audio	Live	warmAlready loaded	OK	0.02 ms	Unknown	Unknown	28.47 s	Not applicable	985.75 ms	68.46 ms

App-process RAM and CPU. Peak RAM is sampled. CPU includes all app threads, but excludes system speech services.
Run	Start RAM	After preparation	After transcription	Sampled peak	Whole-run CPU	Transcription CPU	Thermal maximum	Dropped audio	Committed regressions
#11 · D01	650.8 MiB	204.5 MiB	Unknown	654.5 MiB	40.6%5.488 s CPU in 13.513 s	UnknownUnknown CPU in Unknown	fair	Not applicable	Not applicable
#12 · D01	212.1 MiB	473.7 MiB	1,103.6 MiB	1,108.3 MiB	5.2%10.798 s CPU in 206.204 s	12.4%1.118 s CPU in 9.031 s	fair	0 s	0
#23 · D02	174.9 MiB	1,140.5 MiB	1,181.4 MiB	1,181.5 MiB	57.8%0.782 s CPU in 1.354 s	118.7%0.043 s CPU in 0.036 s	fair	Not applicable	Not applicable
#24 · D02	1,179.8 MiB	1,179.8 MiB	819.3 MiB	1,190.5 MiB	11.5%1.19 s CPU in 10.315 s	11.5%1.189 s CPU in 10.314 s	fair	0 s	0
#28 · D03	202.2 MiB	1,123.9 MiB	1,165.4 MiB	1,165.4 MiB	96%0.789 s CPU in 0.822 s	70.1%0.049 s CPU in 0.07 s	fair	Not applicable	Not applicable
#29 · D03	1,173.2 MiB	1,173.2 MiB	502.1 MiB	1,173.2 MiB	13.8%3.937 s CPU in 28.47 s	13.8%3.936 s CPU in 28.469 s	fair	0 s	0

Run IDs and fixture hashes

#11 · 2d72047e-9a11-4f56-b032-e19ba941de10
D01 SHA-256: c09ae00386e8cb7f9450b2f535ac1b157cb72d1efbc35f8805810d72bc81ab7e
#12 · d72e25d7-f82d-4b2d-80fb-ce7c293008b7
D01 SHA-256: c09ae00386e8cb7f9450b2f535ac1b157cb72d1efbc35f8805810d72bc81ab7e
#23 · 7736cf86-fa64-41ce-8396-2e9c919bc636
D02 SHA-256: b9432690b3648fba32878af9e4fe51cc1b8fdab7c0df15607f6ed0c145813e15
#24 · f51d780e-24cf-4431-b5b1-1c5ed09ab238
D02 SHA-256: b9432690b3648fba32878af9e4fe51cc1b8fdab7c0df15607f6ed0c145813e15
#28 · f7e3d053-6663-498b-a1d7-dcd66ca445d3
D03 SHA-256: 0465ae37184793e6a496f9b1b388b18978ff4d835edc444fa06943291bab61c4
#29 · f9120cea-421d-454c-ad80-d98307322cd4
D03 SHA-256: 0465ae37184793e6a496f9b1b388b18978ff4d835edc444fa06943291bab61c4

Session provenance

Raw export SHA-256: 2c45e1fe51a324ccb49e4fad7a8210d4925d5970aebde22a704bc53cc5b63a72

Session ID: 4e00f424-067b-45cc-ad1f-d0bf48aab901

The local dataset retains measurements and identifiers. It excludes transcripts, reference text, audio paths, and personal device names.