File size: 9,442 Bytes
9cce565 7f5f1cf 9cce565 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 | {
"schema_version": "1.0",
"package_id": "ralph-sub2-iphone-matrix-v1",
"generated_at_utc": "2026-09-15T14:12:31Z",
"publication_status": "published",
"model": {
"repository": "RalphLabsAI/ralph-crowns",
"filename": "ralph-qwen3-8b-sub2.gguf",
"size_bytes": 2937263168,
"display_size_decimal_gb": 2.94,
"expected_sha256": "9c45bf0ac486e793bd761a99ce4b9d965ec12e70f2d420e02af16797c71f6b50",
"sha256_verification": {
"status": "expected_and_local_mirror_verified",
"local_mirror_match": true,
"computed_on_device": false
},
"profile": "sub2",
"quantization": "Q2_K-led mixed quantization",
"embedded_chat_template": true,
"round": 7
},
"app": {
"name": "PocketPal",
"version": "1.17.3",
"build": 147,
"source_commit": "6683a28876d031484cc72dde251b6139a804ad64",
"test_only_device_lab_build": true
},
"test_configuration": {
"provider": "BrowserStack",
"physical_devices": true,
"network_state": "connected",
"airplane_mode_verified": false,
"context_tokens": 2048,
"requested_backend": "gpu",
"effective_backend": "metal",
"runtime_device": "MTL0",
"benchmark": {
"prompt_tokens": 128,
"generated_tokens": 32,
"parallel": 1,
"repeats": 3,
"mmap": true
},
"chat_prompt": "In two sentences, explain how lower-bit weights reduce memory use without changing a model's architecture."
},
"summary": {
"devices_tested": 4,
"successful_benchmarks": 4,
"successful_normal_chat_takes": 4,
"devices_showing_memory_tight_warning": 4,
"normal_chat_tokens_per_second_range": {
"min": 10.58,
"max": 18.54
},
"controlled_generation_tokens_per_second_range": {
"min": 12.358422759719948,
"max": 22.539080509295804
}
},
"devices": [
{
"id": "iphone-15-ios-26.6",
"device": "iPhone 15",
"os": "iOS 26.6",
"physical_device": true,
"session_started_at_utc": "2026-09-15T14:02:27Z",
"session_duration_seconds": 458,
"benchmark_captured_at_utc": "2026-09-15T14:08:56.887Z",
"status": "passed",
"model_loaded": true,
"chat_completed": true,
"memory_tight_warning_visible": true,
"normal_pocketpal_chat": {
"tokens_per_second": 10.58,
"milliseconds_per_token": 95,
"ttft_ms": 838,
"response": "Lower-bit weights reduce memory use by quantizing the model's parameters to fewer bits, which decreases the storage requirements per parameter without altering the model's structure. This allows the model to run efficiently on devices with limited memory while maintaining its performance."
},
"controlled_benchmark": {
"prompt_tokens_per_second": 81.67847742889315,
"generation_tokens_per_second": 12.358422759719948,
"wall_ms": 159967,
"observed_peak_memory_mb": 1359.25,
"runtime_gpu": true,
"runtime_devices": [
"MTL0"
]
},
"artifact_sha256_on_device_verified": false,
"evidence": {
"receipt": "devices/iphone-15-ios-26.6/RECEIPT.md",
"benchmark_screenshot": "devices/iphone-15-ios-26.6/benchmark.png",
"model_screenshot": "devices/iphone-15-ios-26.6/model-ready.png",
"chat_screenshot": "devices/iphone-15-ios-26.6/chat-response.png",
"raw_screen_recording": "devices/iphone-15-ios-26.6/receipt-raw.mp4"
}
},
{
"id": "iphone-16-ios-18.6",
"device": "iPhone 16",
"os": "iOS 18.6",
"physical_device": true,
"session_started_at_utc": "2026-09-15T14:02:27Z",
"session_duration_seconds": 450,
"benchmark_captured_at_utc": "2026-09-15T14:08:40.550Z",
"status": "passed",
"model_loaded": true,
"chat_completed": true,
"memory_tight_warning_visible": true,
"normal_pocketpal_chat": {
"tokens_per_second": 12.36,
"milliseconds_per_token": 81,
"ttft_ms": 439,
"response": "Lower-bit weights reduce memory use by quantizing the model's parameters to fewer bits, which decreases the storage requirements without altering the model's structure. This allows for more efficient memory utilization while maintaining the same number of layers and operations."
},
"controlled_benchmark": {
"prompt_tokens_per_second": 88.19278523444183,
"generation_tokens_per_second": 14.401330754649772,
"wall_ms": 301949,
"observed_peak_memory_mb": 3151.52,
"runtime_gpu": true,
"runtime_devices": [
"MTL0"
]
},
"artifact_sha256_on_device_verified": false,
"evidence": {
"receipt": "devices/iphone-16-ios-18.6/RECEIPT.md",
"benchmark_screenshot": "devices/iphone-16-ios-18.6/benchmark.png",
"model_screenshot": "devices/iphone-16-ios-18.6/model-ready.png",
"chat_screenshot": "devices/iphone-16-ios-18.6/chat-response.png",
"raw_screen_recording": "devices/iphone-16-ios-18.6/receipt-raw.mp4"
}
},
{
"id": "iphone-17-ios-26.6",
"device": "iPhone 17",
"os": "iOS 26.6",
"physical_device": true,
"session_started_at_utc": "2026-09-15T14:02:28Z",
"session_duration_seconds": 455,
"benchmark_captured_at_utc": "2026-09-15T14:08:51.361Z",
"status": "passed",
"model_loaded": true,
"chat_completed": true,
"memory_tight_warning_visible": true,
"normal_pocketpal_chat": {
"tokens_per_second": 16.77,
"milliseconds_per_token": 60,
"ttft_ms": 427,
"response": "Lower-bit weights reduce memory use by representing model parameters with fewer bits, which decreases the amount of memory required to store each weight without altering the model's structure. This allows for more efficient memory utilization while maintaining the same number of layers and neurons in the model."
},
"controlled_benchmark": {
"prompt_tokens_per_second": 288.4833486243861,
"generation_tokens_per_second": 20.356770485468502,
"wall_ms": 311703,
"observed_peak_memory_mb": 1897.22,
"runtime_gpu": true,
"runtime_devices": [
"MTL0"
]
},
"artifact_sha256_on_device_verified": false,
"evidence": {
"receipt": "devices/iphone-17-ios-26.6/RECEIPT.md",
"benchmark_screenshot": "devices/iphone-17-ios-26.6/benchmark.png",
"model_screenshot": "devices/iphone-17-ios-26.6/model-ready.png",
"chat_screenshot": "devices/iphone-17-ios-26.6/chat-response.png",
"raw_screen_recording": "devices/iphone-17-ios-26.6/receipt-raw.mp4"
}
},
{
"id": "iphone-17-pro-max-ios-26.6",
"device": "iPhone 17 Pro Max",
"os": "iOS 26.6",
"physical_device": true,
"session_started_at_utc": "2026-09-15T13:28:50Z",
"session_duration_seconds": 316,
"benchmark_captured_at_utc": "2026-09-15T13:33:04.160Z",
"status": "passed",
"model_loaded": true,
"chat_completed": true,
"memory_tight_warning_visible": true,
"normal_pocketpal_chat": {
"tokens_per_second": 18.54,
"milliseconds_per_token": 54,
"ttft_ms": 400,
"response": "Lower-bit weights reduce memory use by quantizing the model's weights to fewer bits, which decreases the storage requirements per weight without altering the model's structure. This allows for more efficient memory utilization while maintaining the same computational graph and parameter distribution."
},
"controlled_benchmark": {
"prompt_tokens_per_second": 328.47583335972814,
"generation_tokens_per_second": 22.539080509295804,
"wall_ms": 171784,
"observed_peak_memory_mb": 3224.59,
"runtime_gpu": true,
"runtime_devices": [
"MTL0"
]
},
"artifact_sha256_on_device_verified": false,
"evidence": {
"receipt": "devices/iphone-17-pro-max-ios-26.6/RECEIPT.md",
"benchmark_screenshot": "devices/iphone-17-pro-max-ios-26.6/benchmark.png",
"model_screenshot": "devices/iphone-17-pro-max-ios-26.6/model-ready.png",
"chat_screenshot": "devices/iphone-17-pro-max-ios-26.6/chat-response.png",
"raw_screen_recording": "devices/iphone-17-pro-max-ios-26.6/receipt-raw.mp4"
}
}
],
"limitations": [
"BrowserStack re-signing strips PocketPal increased-memory-limit and extended-virtual-addressing entitlements.",
"These are physical-cloud-device throughput and backend observations, not App Store memory-fit certification.",
"Devices remained network-connected; these recordings are not airplane-mode or offline proof.",
"The expected model SHA-256 was verified against a local mirror, not computed on-device.",
"The Memory tight warning was visible on every tested device even though all four loaded and generated successfully.",
"Controlled benchmark generation speed and normal PocketPal chat speed are separate measurements and must not be substituted for one another.",
"Observed peak memory is a harness observation, not required device RAM.",
"Each row is one session without thermal normalization and does not establish universal performance or model quality."
],
"publication_repository": "https://huggingface.co/datasets/RalphLabsAI/ralph-device-lab"
}
|