{"row_id": "L-IQ3/speed/a1/run-001", "arm": "L-IQ3", "engine": "llama.cpp", "engine_commit": "3cf03257f219afbe7334045ff7c6a06ac68c627d", "quant": "IQ3_S", "file_sha256": {"1": "4c1eb2ceb4915e1192f4f386021897bde56a97f40a0bb78bb86465e0f7d2aca3", "2": "316b46f3a2dbd68c900f43136ab9449f9dcc3725dfd8c794847c204bc161e113"}, "command_line": ["/data/strata-bench/build/llama.cpp-3cf03257/bin/llama-server", "-m", "/data/strata-bench/models/ISTA-ed59f920/IQ3_S/Qwen3.8-Flash-Next-GSQ-RCO-IQ3_S-00001-of-00002.gguf", "--host", "127.0.0.1", "--port", "withheld-2026-10-13", "-ngl", "99", "--n-cpu-moe", "34", "-c", "131072", "-fa", "on", "-ctk", "q8_0", "-ctv", "q8_0", "-lm", "mmap", "--lazy-mode", "on", "-fit", "off", "-np", "1", "--cache-ram", "0", "--no-webui", "--jinja", "-lv", "4"], "ctx": "4K", "repeat": 0, "run": 1, "slice_sha256": "9c676dc1959897277fda62a2d77d01fc06c932a2e04641c753a24bde3e7d4a16", "prompt_sha256": "b42c6a181c3e02478e7e033585c5962f35eb046b680c0ae837ad8706a3c0674b", "nonce": "Ticket 85822155.", "requests_since_start": 1, "status": "ok", "fail_kind": null, "error": null, "decode_tok_s": 19.268, "decode_rule": "client clock: (completion_tokens - 1) / (t_finish - t_first_token) (the kit's one timing rule, seatlib.RULES['engine_timing'])", "decode_tok_s_server": 19.266071473498613, "prompt_tok_s_server": 136.77541294285615, "prompt_tok_s_client": 136.7, "ttft_ms": 29533.1, "prompt_tokens": 4038, "completion_tokens": 256, "cached_tokens": 0, "finish_reason": "length", "server_timings": {"cache_n": 0, "prompt_n": 4038, "prompt_ms": 29522.85, "prompt_per_token_ms": 7.311255572065378, "prompt_per_second": 136.77541294285615, "predicted_n": 256, "predicted_ms": 13235.703, "predicted_per_token_ms": 51.904717647058824, "predicted_per_second": 19.266071473498613}, "short_completion": false, "cache_flag": null, "completion_sha256": "af131881ad9069cf1dcebd272b9d9f1178e0af8992f7a64662ac100319bc92dd", "completion_head": "withheld-2026-10-13", "strata": null, "telemetry": {"window": {"t0": 1790974430.464, "t1": 1790974473.232, "seconds": 42.768, "host_samples": 43, "gpu_samples": 43}, "board": {"joules": 2010.98, "mean_w": 47.87, "max_w": 64.04, "rule": "trapezoid over nvidia-smi power.draw at 1 Hz"}, "rapl_package0": {"joules": 1437.1, "label": "RAPL package-0, the CPU package: NOT WALL", "source": "/run/rapl/energy", "span_s": 42.006}, "data_reads_per_s": {"mean": 2207.5, "max": 11075.3}, "data_read_mb_per_s": {"mean": 112.97}, "temps": {"gpu_max_c": 45.0, "ssd_max_c": 43.85}, "vram": {"max_used_mib": 22467.0, "baseline_mib": 15.0, "growth_mib": 22452.0, "rule": "max memory.used in the window minus the pre-load baseline (read before the unit started)"}, "pinned": {"vmlck_max_bytes": 0, "vmpin_max_bytes": 0, "vmrss_max_bytes": 38296444928, "rule": "sums over the engine unit's processes' /proc/<pid>/status, every 5 s"}, "load": {"load1_mean": 2.18, "claude_procs_max": 1}, "memory": {"mem_available_gib_min": 51.256, "psi_full_avg10_max": 1.6, "unit_max_events_delta": 0}, "power_limit": {"min_w": 150.0, "max_w": 150.0, "rule": "enforced.power.limit READ (Dynamic Boost floats it), never one number"}}}
{"row_id": "L-IQ3/speed/a1/run-002", "arm": "L-IQ3", "engine": "llama.cpp", "engine_commit": "3cf03257f219afbe7334045ff7c6a06ac68c627d", "quant": "IQ3_S", "file_sha256": {"1": "4c1eb2ceb4915e1192f4f386021897bde56a97f40a0bb78bb86465e0f7d2aca3", "2": "316b46f3a2dbd68c900f43136ab9449f9dcc3725dfd8c794847c204bc161e113"}, "command_line": ["/data/strata-bench/build/llama.cpp-3cf03257/bin/llama-server", "-m", "/data/strata-bench/models/ISTA-ed59f920/IQ3_S/Qwen3.8-Flash-Next-GSQ-RCO-IQ3_S-00001-of-00002.gguf", "--host", "127.0.0.1", "--port", "withheld-2026-10-13", "-ngl", "99", "--n-cpu-moe", "34", "-c", "131072", "-fa", "on", "-ctk", "q8_0", "-ctv", "q8_0", "-lm", "mmap", "--lazy-mode", "on", "-fit", "off", "-np", "1", "--cache-ram", "0", "--no-webui", "--jinja", "-lv", "4"], "ctx": "4K", "repeat": 1, "run": 2, "slice_sha256": "9c676dc1959897277fda62a2d77d01fc06c932a2e04641c753a24bde3e7d4a16", "prompt_sha256": "9744fee0d5b25fee2aef36ad0bc7c5c4611f73d11f4e6bd9e5b84b17d93d319b", "nonce": "Ticket 42865306.", "requests_since_start": 2, "status": "ok", "fail_kind": null, "error": null, "decode_tok_s": 20.236, "decode_rule": "client clock: (completion_tokens - 1) / (t_finish - t_first_token) (the kit's one timing rule, seatlib.RULES['engine_timing'])", "decode_tok_s_server": 20.236105354402056, "prompt_tok_s_server": 265.7928720512739, "prompt_tok_s_client": 265.7, "ttft_ms": 15198.827, "prompt_tokens": 4038, "completion_tokens": 256, "cached_tokens": 0, "finish_reason": "length", "server_timings": {"cache_n": 0, "prompt_n": 4038, "prompt_ms": 15192.281, "prompt_per_token_ms": 3.76232813273898, "prompt_per_second": 265.7928720512739, "predicted_n": 256, "predicted_ms": 12601.239, "predicted_per_token_ms": 49.416623529411766, "predicted_per_second": 20.236105354402056}, "short_completion": false, "cache_flag": null, "completion_sha256": "496412a664bab2d70d56812f0b8e4d990e5c76186d61d271bda2c1071fe1938c", "completion_head": "withheld-2026-10-13", "strata": null, "telemetry": {"window": {"t0": 1790974473.232, "t1": 1790974501.032, "seconds": 27.8, "host_samples": 28, "gpu_samples": 28}, "board": {"joules": 1729.19, "mean_w": 64.03, "max_w": 72.56, "rule": "trapezoid over nvidia-smi power.draw at 1 Hz"}, "rapl_package0": {"joules": 1106.29, "label": "RAPL package-0, the CPU package: NOT WALL", "source": "/run/rapl/energy", "span_s": 27.004}, "data_reads_per_s": {"mean": 66.5, "max": 291.9}, "data_read_mb_per_s": {"mean": 2.3}, "temps": {"gpu_max_c": 47.0, "ssd_max_c": 43.85}, "vram": {"max_used_mib": 22467.0, "baseline_mib": 15.0, "growth_mib": 22452.0, "rule": "max memory.used in the window minus the pre-load baseline (read before the unit started)"}, "pinned": {"vmlck_max_bytes": 0, "vmpin_max_bytes": 0, "vmrss_max_bytes": 38192238592, "rule": "sums over the engine unit's processes' /proc/<pid>/status, every 5 s"}, "load": {"load1_mean": 2.44, "claude_procs_max": 1}, "memory": {"mem_available_gib_min": 51.182, "psi_full_avg10_max": 0.14, "unit_max_events_delta": 0}, "power_limit": {"min_w": 150.0, "max_w": 150.0, "rule": "enforced.power.limit READ (Dynamic Boost floats it), never one number"}}}
{"row_id": "L-IQ3/speed/a1/run-003", "arm": "L-IQ3", "engine": "llama.cpp", "engine_commit": "3cf03257f219afbe7334045ff7c6a06ac68c627d", "quant": "IQ3_S", "file_sha256": {"1": "4c1eb2ceb4915e1192f4f386021897bde56a97f40a0bb78bb86465e0f7d2aca3", "2": "316b46f3a2dbd68c900f43136ab9449f9dcc3725dfd8c794847c204bc161e113"}, "command_line": ["/data/strata-bench/build/llama.cpp-3cf03257/bin/llama-server", "-m", "/data/strata-bench/models/ISTA-ed59f920/IQ3_S/Qwen3.8-Flash-Next-GSQ-RCO-IQ3_S-00001-of-00002.gguf", "--host", "127.0.0.1", "--port", "withheld-2026-10-13", "-ngl", "99", "--n-cpu-moe", "34", "-c", "131072", "-fa", "on", "-ctk", "q8_0", "-ctv", "q8_0", "-lm", "mmap", "--lazy-mode", "on", "-fit", "off", "-np", "1", "--cache-ram", "0", "--no-webui", "--jinja", "-lv", "4"], "ctx": "4K", "repeat": 2, "run": 3, "slice_sha256": "9c676dc1959897277fda62a2d77d01fc06c932a2e04641c753a24bde3e7d4a16", "prompt_sha256": "214ec62ff2b1550ea59fc2fa7cc264e79ceed8841f5fd91e1709a8b73be2fa8f", "nonce": "Ticket 90569497.", "requests_since_start": 3, "status": "ok", "fail_kind": null, "error": null, "decode_tok_s": 20.525, "decode_rule": "client clock: (completion_tokens - 1) / (t_finish - t_first_token) (the kit's one timing rule, seatlib.RULES['engine_timing'])", "decode_tok_s_server": 20.5246800423726, "prompt_tok_s_server": 272.75085243084635, "prompt_tok_s_client": 272.6, "ttft_ms": 14810.306, "prompt_tokens": 4038, "completion_tokens": 256, "cached_tokens": 0, "finish_reason": "length", "server_timings": {"cache_n": 0, "prompt_n": 4038, "prompt_ms": 14804.72, "prompt_per_token_ms": 3.666349678058445, "prompt_per_second": 272.75085243084635, "predicted_n": 256, "predicted_ms": 12424.067, "predicted_per_token_ms": 48.72183137254902, "predicted_per_second": 20.5246800423726}, "short_completion": false, "cache_flag": null, "completion_sha256": "2f42cb414309477fc9d218ba3a726ae2762280c21580e79f45f37104fed104e0", "completion_head": "withheld-2026-10-13", "strata": null, "telemetry": {"window": {"t0": 1790974501.033, "t1": 1790974528.267, "seconds": 27.234, "host_samples": 27, "gpu_samples": 27}, "board": {"joules": 1704.13, "mean_w": 65.53, "max_w": 71.87, "rule": "trapezoid over nvidia-smi power.draw at 1 Hz"}, "rapl_package0": {"joules": 1080.66, "label": "RAPL package-0, the CPU package: NOT WALL", "source": "/run/rapl/energy", "span_s": 26.004}, "data_reads_per_s": {"mean": 59.8, "max": 288.9}, "data_read_mb_per_s": {"mean": 1.42}, "temps": {"gpu_max_c": 50.0, "ssd_max_c": 43.85}, "vram": {"max_used_mib": 22467.0, "baseline_mib": 15.0, "growth_mib": 22452.0, "rule": "max memory.used in the window minus the pre-load baseline (read before the unit started)"}, "pinned": {"vmlck_max_bytes": 0, "vmpin_max_bytes": 0, "vmrss_max_bytes": 38216355840, "rule": "sums over the engine unit's processes' /proc/<pid>/status, every 5 s"}, "load": {"load1_mean": 2.22, "claude_procs_max": 1}, "memory": {"mem_available_gib_min": 51.155, "psi_full_avg10_max": 0.0, "unit_max_events_delta": 0}, "power_limit": {"min_w": 150.0, "max_w": 150.0, "rule": "enforced.power.limit READ (Dynamic Boost floats it), never one number"}}}
{"row_id": "L-IQ3/speed/a1/run-004", "arm": "L-IQ3", "engine": "llama.cpp", "engine_commit": "3cf03257f219afbe7334045ff7c6a06ac68c627d", "quant": "IQ3_S", "file_sha256": {"1": "4c1eb2ceb4915e1192f4f386021897bde56a97f40a0bb78bb86465e0f7d2aca3", "2": "316b46f3a2dbd68c900f43136ab9449f9dcc3725dfd8c794847c204bc161e113"}, "command_line": ["/data/strata-bench/build/llama.cpp-3cf03257/bin/llama-server", "-m", "/data/strata-bench/models/ISTA-ed59f920/IQ3_S/Qwen3.8-Flash-Next-GSQ-RCO-IQ3_S-00001-of-00002.gguf", "--host", "127.0.0.1", "--port", "withheld-2026-10-13", "-ngl", "99", "--n-cpu-moe", "34", "-c", "131072", "-fa", "on", "-ctk", "q8_0", "-ctv", "q8_0", "-lm", "mmap", "--lazy-mode", "on", "-fit", "off", "-np", "1", "--cache-ram", "0", "--no-webui", "--jinja", "-lv", "4"], "ctx": "32K", "repeat": 0, "run": 4, "slice_sha256": "b72b07858a845e76331ec38d46abae3c09d3e1cb6e4d968e887fa8b132c64977", "prompt_sha256": "507b7bc9961d968eaf07467339f314a3c76a036b2542852f07b9a638adcd5ad5", "nonce": "Ticket 10426037.", "requests_since_start": 4, "status": "ok", "fail_kind": null, "error": null, "decode_tok_s": 17.693, "decode_rule": "client clock: (completion_tokens - 1) / (t_finish - t_first_token) (the kit's one timing rule, seatlib.RULES['engine_timing'])", "decode_tok_s_server": 17.693614457806248, "prompt_tok_s_server": 239.66088769845229, "prompt_tok_s_client": 239.6, "ttft_ms": 133709.454, "prompt_tokens": 32038, "completion_tokens": 256, "cached_tokens": 0, "finish_reason": "length", "server_timings": {"cache_n": 0, "prompt_n": 32038, "prompt_ms": 133680.553, "prompt_per_token_ms": 4.1725623634434115, "prompt_per_second": 239.66088769845229, "predicted_n": 256, "predicted_ms": 14411.979, "predicted_per_token_ms": 56.51756470588235, "predicted_per_second": 17.693614457806248}, "short_completion": false, "cache_flag": null, "completion_sha256": "cb128e39f75fa87c982f733d315111a4f0e7d23a6bd9fccf814d49bf09ad3f4a", "completion_head": "withheld-2026-10-13", "strata": null, "telemetry": {"window": {"t0": 1790974528.268, "t1": 1790974676.39, "seconds": 148.122, "host_samples": 148, "gpu_samples": 148}, "board": {"joules": 9661.29, "mean_w": 65.64, "max_w": 79.48, "rule": "trapezoid over nvidia-smi power.draw at 1 Hz"}, "rapl_package0": {"joules": 5125.78, "label": "RAPL package-0, the CPU package: NOT WALL", "source": "/run/rapl/energy", "span_s": 147.024}, "data_reads_per_s": {"mean": 1226.1, "max": 6965.5}, "data_read_mb_per_s": {"mean": 10.66}, "temps": {"gpu_max_c": 54.0, "ssd_max_c": 45.85}, "vram": {"max_used_mib": 22469.0, "baseline_mib": 15.0, "growth_mib": 22454.0, "rule": "max memory.used in the window minus the pre-load baseline (read before the unit started)"}, "pinned": {"vmlck_max_bytes": 0, "vmpin_max_bytes": 0, "vmrss_max_bytes": 39899971584, "rule": "sums over the engine unit's processes' /proc/<pid>/status, every 5 s"}, "load": {"load1_mean": 1.69, "claude_procs_max": 1}, "memory": {"mem_available_gib_min": 51.046, "psi_full_avg10_max": 0.0, "unit_max_events_delta": 0}, "power_limit": {"min_w": 150.0, "max_w": 150.0, "rule": "enforced.power.limit READ (Dynamic Boost floats it), never one number"}}}
{"row_id": "L-IQ3/speed/a1/run-005", "arm": "L-IQ3", "engine": "llama.cpp", "engine_commit": "3cf03257f219afbe7334045ff7c6a06ac68c627d", "quant": "IQ3_S", "file_sha256": {"1": "4c1eb2ceb4915e1192f4f386021897bde56a97f40a0bb78bb86465e0f7d2aca3", "2": "316b46f3a2dbd68c900f43136ab9449f9dcc3725dfd8c794847c204bc161e113"}, "command_line": ["/data/strata-bench/build/llama.cpp-3cf03257/bin/llama-server", "-m", "/data/strata-bench/models/ISTA-ed59f920/IQ3_S/Qwen3.8-Flash-Next-GSQ-RCO-IQ3_S-00001-of-00002.gguf", "--host", "127.0.0.1", "--port", "withheld-2026-10-13", "-ngl", "99", "--n-cpu-moe", "34", "-c", "131072", "-fa", "on", "-ctk", "q8_0", "-ctv", "q8_0", "-lm", "mmap", "--lazy-mode", "on", "-fit", "off", "-np", "1", "--cache-ram", "0", "--no-webui", "--jinja", "-lv", "4"], "ctx": "32K", "repeat": 1, "run": 5, "slice_sha256": "b72b07858a845e76331ec38d46abae3c09d3e1cb6e4d968e887fa8b132c64977", "prompt_sha256": "c8a1ccb2149bdf9595fe8d97b7183fd9efd61c32259e907606ddbac7710955b2", "nonce": "Ticket 97542637.", "requests_since_start": 5, "status": "ok", "fail_kind": null, "error": null, "decode_tok_s": 18.368, "decode_rule": "client clock: (completion_tokens - 1) / (t_finish - t_first_token) (the kit's one timing rule, seatlib.RULES['engine_timing'])", "decode_tok_s_server": 18.368962878270974, "prompt_tok_s_server": 263.35459884591694, "prompt_tok_s_client": 263.3, "ttft_ms": 121675.901, "prompt_tokens": 32038, "completion_tokens": 256, "cached_tokens": 0, "finish_reason": "length", "server_timings": {"cache_n": 0, "prompt_n": 32038, "prompt_ms": 121653.467, "prompt_per_token_ms": 3.7971617142143708, "prompt_per_second": 263.35459884591694, "predicted_n": 256, "predicted_ms": 13882.112, "predicted_per_token_ms": 54.43965490196078, "predicted_per_second": 18.368962878270974}, "short_completion": false, "cache_flag": null, "completion_sha256": "527b05689e576cbd8f447cc7c9e940ace9a3d29398f01daffdcfaec25b968db0", "completion_head": "withheld-2026-10-13", "strata": null, "telemetry": {"window": {"t0": 1790974676.391, "t1": 1790974811.949, "seconds": 135.559, "host_samples": 136, "gpu_samples": 135}, "board": {"joules": 9340.07, "mean_w": 69.63, "max_w": 81.9, "rule": "trapezoid over nvidia-smi power.draw at 1 Hz"}, "rapl_package0": {"joules": 4710.02, "label": "RAPL package-0, the CPU package: NOT WALL", "source": "/run/rapl/energy", "span_s": 135.02}, "data_reads_per_s": {"mean": 827.2, "max": 14716.8}, "data_read_mb_per_s": {"mean": 7.73}, "temps": {"gpu_max_c": 56.0, "ssd_max_c": 46.85}, "vram": {"max_used_mib": 22469.0, "baseline_mib": 15.0, "growth_mib": 22454.0, "rule": "max memory.used in the window minus the pre-load baseline (read before the unit started)"}, "pinned": {"vmlck_max_bytes": 0, "vmpin_max_bytes": 0, "vmrss_max_bytes": 39924826112, "rule": "sums over the engine unit's processes' /proc/<pid>/status, every 5 s"}, "load": {"load1_mean": 1.8, "claude_procs_max": 1}, "memory": {"mem_available_gib_min": 51.06, "psi_full_avg10_max": 0.0, "unit_max_events_delta": 0}, "power_limit": {"min_w": 148.36, "max_w": 150.0, "rule": "enforced.power.limit READ (Dynamic Boost floats it), never one number"}}}
{"row_id": "L-IQ3/speed/a1/run-006", "arm": "L-IQ3", "engine": "llama.cpp", "engine_commit": "3cf03257f219afbe7334045ff7c6a06ac68c627d", "quant": "IQ3_S", "file_sha256": {"1": "4c1eb2ceb4915e1192f4f386021897bde56a97f40a0bb78bb86465e0f7d2aca3", "2": "316b46f3a2dbd68c900f43136ab9449f9dcc3725dfd8c794847c204bc161e113"}, "command_line": ["/data/strata-bench/build/llama.cpp-3cf03257/bin/llama-server", "-m", "/data/strata-bench/models/ISTA-ed59f920/IQ3_S/Qwen3.8-Flash-Next-GSQ-RCO-IQ3_S-00001-of-00002.gguf", "--host", "127.0.0.1", "--port", "withheld-2026-10-13", "-ngl", "99", "--n-cpu-moe", "34", "-c", "131072", "-fa", "on", "-ctk", "q8_0", "-ctv", "q8_0", "-lm", "mmap", "--lazy-mode", "on", "-fit", "off", "-np", "1", "--cache-ram", "0", "--no-webui", "--jinja", "-lv", "4"], "ctx": "32K", "repeat": 2, "run": 6, "slice_sha256": "b72b07858a845e76331ec38d46abae3c09d3e1cb6e4d968e887fa8b132c64977", "prompt_sha256": "d6c964051c51baa9ef5cb009be0a921422c2b28350d2dbea610a6038552bdc18", "nonce": "Ticket 24502846.", "requests_since_start": 6, "status": "ok", "fail_kind": null, "error": null, "decode_tok_s": 18.494, "decode_rule": "client clock: (completion_tokens - 1) / (t_finish - t_first_token) (the kit's one timing rule, seatlib.RULES['engine_timing'])", "decode_tok_s_server": 18.494870067010176, "prompt_tok_s_server": 252.88145185431338, "prompt_tok_s_client": 252.8, "ttft_ms": 126714.694, "prompt_tokens": 32038, "completion_tokens": 256, "cached_tokens": 0, "finish_reason": "length", "server_timings": {"cache_n": 0, "prompt_n": 32038, "prompt_ms": 126691.775, "prompt_per_token_ms": 3.954422092515138, "prompt_per_second": 252.88145185431338, "predicted_n": 256, "predicted_ms": 13787.607, "predicted_per_token_ms": 54.06904705882353, "predicted_per_second": 18.494870067010176}, "short_completion": false, "cache_flag": null, "completion_sha256": "7d7031a1e2a44bcffbe84fc6f34a4e8138201596094b3252fa74034b288578f7", "completion_head": "withheld-2026-10-13", "strata": null, "telemetry": {"window": {"t0": 1790974811.95, "t1": 1790974952.453, "seconds": 140.503, "host_samples": 140, "gpu_samples": 141}, "board": {"joules": 9531.06, "mean_w": 67.99, "max_w": 79.82, "rule": "trapezoid over nvidia-smi power.draw at 1 Hz"}, "rapl_package0": {"joules": 5243.42, "label": "RAPL package-0, the CPU package: NOT WALL", "source": "/run/rapl/energy", "span_s": 139.021}, "data_reads_per_s": {"mean": 617.9, "max": 25544.9}, "data_read_mb_per_s": {"mean": 16.43}, "temps": {"gpu_max_c": 56.0, "ssd_max_c": 48.85}, "vram": {"max_used_mib": 22469.0, "baseline_mib": 15.0, "growth_mib": 22454.0, "rule": "max memory.used in the window minus the pre-load baseline (read before the unit started)"}, "pinned": {"vmlck_max_bytes": 0, "vmpin_max_bytes": 0, "vmrss_max_bytes": 39924826112, "rule": "sums over the engine unit's processes' /proc/<pid>/status, every 5 s"}, "load": {"load1_mean": 2.05, "claude_procs_max": 1}, "memory": {"mem_available_gib_min": 50.504, "psi_full_avg10_max": 1.99, "unit_max_events_delta": 0}, "power_limit": {"min_w": 150.0, "max_w": 150.0, "rule": "enforced.power.limit READ (Dynamic Boost floats it), never one number"}}}
{"row_id": "L-IQ3/speed/a1/run-007", "arm": "L-IQ3", "engine": "llama.cpp", "engine_commit": "3cf03257f219afbe7334045ff7c6a06ac68c627d", "quant": "IQ3_S", "file_sha256": {"1": "4c1eb2ceb4915e1192f4f386021897bde56a97f40a0bb78bb86465e0f7d2aca3", "2": "316b46f3a2dbd68c900f43136ab9449f9dcc3725dfd8c794847c204bc161e113"}, "command_line": ["/data/strata-bench/build/llama.cpp-3cf03257/bin/llama-server", "-m", "/data/strata-bench/models/ISTA-ed59f920/IQ3_S/Qwen3.8-Flash-Next-GSQ-RCO-IQ3_S-00001-of-00002.gguf", "--host", "127.0.0.1", "--port", "withheld-2026-10-13", "-ngl", "99", "--n-cpu-moe", "34", "-c", "131072", "-fa", "on", "-ctk", "q8_0", "-ctv", "q8_0", "-lm", "mmap", "--lazy-mode", "on", "-fit", "off", "-np", "1", "--cache-ram", "0", "--no-webui", "--jinja", "-lv", "4"], "ctx": "128K", "repeat": 0, "run": 7, "slice_sha256": "286827657f5df11459f4ee550246f7e0e06ce7af3238fe74a53f0aa4c12676ac", "prompt_sha256": "2491e963527b8a39f7fff1aa2d04dab5eb4dd3ca12c7417c451f6216c665972c", "nonce": "Ticket 14239051.", "requests_since_start": 7, "status": "ok", "fail_kind": null, "error": null, "decode_tok_s": 13.642, "decode_rule": "client clock: (completion_tokens - 1) / (t_finish - t_first_token) (the kit's one timing rule, seatlib.RULES['engine_timing'])", "decode_tok_s_server": 13.643601311497862, "prompt_tok_s_server": 229.63944066370064, "prompt_tok_s_client": 229.6, "ttft_ms": 557645.092, "prompt_tokens": 128037, "completion_tokens": 256, "cached_tokens": 0, "finish_reason": "length", "server_timings": {"cache_n": 0, "prompt_n": 128037, "prompt_ms": 557556.662, "prompt_per_token_ms": 4.354652655091887, "prompt_per_second": 229.63944066370064, "predicted_n": 256, "predicted_ms": 18690.08, "predicted_per_token_ms": 73.29443137254903, "predicted_per_second": 13.643601311497862}, "short_completion": false, "cache_flag": null, "completion_sha256": "89d971db96555afeff7b6ad7708de40edacc94cadc572442d7ca246d598fe72f", "completion_head": "withheld-2026-10-13", "strata": null, "telemetry": {"window": {"t0": 1790974952.454, "t1": 1790975528.791, "seconds": 576.337, "host_samples": 577, "gpu_samples": 575}, "board": {"joules": 38632.66, "mean_w": 67.23, "max_w": 126.48, "rule": "trapezoid over nvidia-smi power.draw at 1 Hz"}, "rapl_package0": {"joules": 24008.97, "label": "RAPL package-0, the CPU package: NOT WALL", "source": "/run/rapl/energy", "span_s": 576.086}, "data_reads_per_s": {"mean": 1246.6, "max": 7182.1}, "data_read_mb_per_s": {"mean": 5.48}, "temps": {"gpu_max_c": 59.0, "ssd_max_c": 47.85}, "vram": {"max_used_mib": 22469.0, "baseline_mib": 15.0, "growth_mib": 22454.0, "rule": "max memory.used in the window minus the pre-load baseline (read before the unit started)"}, "pinned": {"vmlck_max_bytes": 0, "vmpin_max_bytes": 0, "vmrss_max_bytes": 40383188992, "rule": "sums over the engine unit's processes' /proc/<pid>/status, every 5 s"}, "load": {"load1_mean": 3.33, "claude_procs_max": 1}, "memory": {"mem_available_gib_min": 48.38, "psi_full_avg10_max": 5.35, "unit_max_events_delta": 0}, "power_limit": {"min_w": 135.0, "max_w": 150.0, "rule": "enforced.power.limit READ (Dynamic Boost floats it), never one number"}}}
{"row_id": "L-IQ3/speed/a1/run-008", "arm": "L-IQ3", "engine": "llama.cpp", "engine_commit": "3cf03257f219afbe7334045ff7c6a06ac68c627d", "quant": "IQ3_S", "file_sha256": {"1": "4c1eb2ceb4915e1192f4f386021897bde56a97f40a0bb78bb86465e0f7d2aca3", "2": "316b46f3a2dbd68c900f43136ab9449f9dcc3725dfd8c794847c204bc161e113"}, "command_line": ["/data/strata-bench/build/llama.cpp-3cf03257/bin/llama-server", "-m", "/data/strata-bench/models/ISTA-ed59f920/IQ3_S/Qwen3.8-Flash-Next-GSQ-RCO-IQ3_S-00001-of-00002.gguf", "--host", "127.0.0.1", "--port", "withheld-2026-10-13", "-ngl", "99", "--n-cpu-moe", "34", "-c", "131072", "-fa", "on", "-ctk", "q8_0", "-ctv", "q8_0", "-lm", "mmap", "--lazy-mode", "on", "-fit", "off", "-np", "1", "--cache-ram", "0", "--no-webui", "--jinja", "-lv", "4"], "ctx": "128K", "repeat": 1, "run": 8, "slice_sha256": "286827657f5df11459f4ee550246f7e0e06ce7af3238fe74a53f0aa4c12676ac", "prompt_sha256": "94613594c1c300728a39aa8436592de0bdf88c05edeceb88cd9c6794de1fa097", "nonce": "Ticket 81078402.", "requests_since_start": 8, "status": "ok", "fail_kind": null, "error": null, "decode_tok_s": 13.648, "decode_rule": "client clock: (completion_tokens - 1) / (t_finish - t_first_token) (the kit's one timing rule, seatlib.RULES['engine_timing'])", "decode_tok_s_server": 13.649502194625843, "prompt_tok_s_server": 216.4157428414468, "prompt_tok_s_client": 216.4, "ttft_ms": 591716.107, "prompt_tokens": 128037, "completion_tokens": 256, "cached_tokens": 0, "finish_reason": "length", "server_timings": {"cache_n": 0, "prompt_n": 128037, "prompt_ms": 591625.167, "prompt_per_token_ms": 4.620735935706085, "prompt_per_second": 216.4157428414468, "predicted_n": 256, "predicted_ms": 18682.0, "predicted_per_token_ms": 73.26274509803922, "predicted_per_second": 13.649502194625843}, "short_completion": false, "cache_flag": null, "completion_sha256": "01e9814828fa3a94d4c5551c8b4426ac4601987232da3e9d2950e4fa06db2288", "completion_head": "withheld-2026-10-13", "strata": null, "telemetry": {"window": {"t0": 1790975528.793, "t1": 1790976139.193, "seconds": 610.4, "host_samples": 602, "gpu_samples": 610}, "board": {"joules": 39862.11, "mean_w": 65.38, "max_w": 172.53, "rule": "trapezoid over nvidia-smi power.draw at 1 Hz"}, "rapl_package0": {"joules": 32468.23, "label": "RAPL package-0, the CPU package: NOT WALL", "source": "/run/rapl/energy", "span_s": 609.263}, "data_reads_per_s": {"mean": 1153.7, "max": 23677.1}, "data_read_mb_per_s": {"mean": 6.2}, "temps": {"gpu_max_c": 61.0, "ssd_max_c": 49.85}, "vram": {"max_used_mib": 22469.0, "baseline_mib": 15.0, "growth_mib": 22454.0, "rule": "max memory.used in the window minus the pre-load baseline (read before the unit started)"}, "pinned": {"vmlck_max_bytes": 0, "vmpin_max_bytes": 0, "vmrss_max_bytes": 40884989952, "rule": "sums over the engine unit's processes' /proc/<pid>/status, every 5 s"}, "load": {"load1_mean": 4.57, "claude_procs_max": 1}, "memory": {"mem_available_gib_min": 49.068, "psi_full_avg10_max": 1.26, "unit_max_events_delta": 0}, "power_limit": {"min_w": 135.0, "max_w": 150.0, "rule": "enforced.power.limit READ (Dynamic Boost floats it), never one number"}}}
{"row_id": "L-IQ3/speed/a1/run-009", "arm": "L-IQ3", "engine": "llama.cpp", "engine_commit": "3cf03257f219afbe7334045ff7c6a06ac68c627d", "quant": "IQ3_S", "file_sha256": {"1": "4c1eb2ceb4915e1192f4f386021897bde56a97f40a0bb78bb86465e0f7d2aca3", "2": "316b46f3a2dbd68c900f43136ab9449f9dcc3725dfd8c794847c204bc161e113"}, "command_line": ["/data/strata-bench/build/llama.cpp-3cf03257/bin/llama-server", "-m", "/data/strata-bench/models/ISTA-ed59f920/IQ3_S/Qwen3.8-Flash-Next-GSQ-RCO-IQ3_S-00001-of-00002.gguf", "--host", "127.0.0.1", "--port", "withheld-2026-10-13", "-ngl", "99", "--n-cpu-moe", "34", "-c", "131072", "-fa", "on", "-ctk", "q8_0", "-ctv", "q8_0", "-lm", "mmap", "--lazy-mode", "on", "-fit", "off", "-np", "1", "--cache-ram", "0", "--no-webui", "--jinja", "-lv", "4"], "ctx": "128K", "repeat": 2, "run": 9, "slice_sha256": "286827657f5df11459f4ee550246f7e0e06ce7af3238fe74a53f0aa4c12676ac", "prompt_sha256": "38d3c9a3177694d5b9006f1c813fddc9e6c5745048a59c545c41d199f780b364", "nonce": "Ticket 81461625.", "requests_since_start": 9, "status": "ok", "fail_kind": null, "error": null, "decode_tok_s": 13.936, "decode_rule": "client clock: (completion_tokens - 1) / (t_finish - t_first_token) (the kit's one timing rule, seatlib.RULES['engine_timing'])", "decode_tok_s_server": 13.93714968564349, "prompt_tok_s_server": 244.54154403237152, "prompt_tok_s_client": 244.5, "ttft_ms": 523674.355, "prompt_tokens": 128037, "completion_tokens": 256, "cached_tokens": 0, "finish_reason": "length", "server_timings": {"cache_n": 0, "prompt_n": 128037, "prompt_ms": 523579.748, "prompt_per_token_ms": 4.089284722384936, "prompt_per_second": 244.54154403237152, "predicted_n": 256, "predicted_ms": 18296.424, "predicted_per_token_ms": 71.75068235294117, "predicted_per_second": 13.93714968564349}, "short_completion": false, "cache_flag": null, "completion_sha256": "a193e766197392c352f902cef964014d90634551b35bce4cf9e3cb96c9523238", "completion_head": "withheld-2026-10-13", "strata": null, "telemetry": {"window": {"t0": 1790976139.195, "t1": 1790976681.167, "seconds": 541.973, "host_samples": 542, "gpu_samples": 542}, "board": {"joules": 37760.75, "mean_w": 69.72, "max_w": 104.29, "rule": "trapezoid over nvidia-smi power.draw at 1 Hz"}, "rapl_package0": {"joules": 21691.72, "label": "RAPL package-0, the CPU package: NOT WALL", "source": "/run/rapl/energy", "span_s": 541.076}, "data_reads_per_s": {"mean": 31.7, "max": 11964.5}, "data_read_mb_per_s": {"mean": 0.4}, "temps": {"gpu_max_c": 61.0, "ssd_max_c": 49.85}, "vram": {"max_used_mib": 22469.0, "baseline_mib": 15.0, "growth_mib": 22454.0, "rule": "max memory.used in the window minus the pre-load baseline (read before the unit started)"}, "pinned": {"vmlck_max_bytes": 0, "vmpin_max_bytes": 0, "vmrss_max_bytes": 40909615104, "rule": "sums over the engine unit's processes' /proc/<pid>/status, every 5 s"}, "load": {"load1_mean": 2.52, "claude_procs_max": 1}, "memory": {"mem_available_gib_min": 49.662, "psi_full_avg10_max": 0.0, "unit_max_events_delta": 0}, "power_limit": {"min_w": 135.0, "max_w": 150.0, "rule": "enforced.power.limit READ (Dynamic Boost floats it), never one number"}}}