{ "admission": { "ceiling_mib": 119325.0, "kv_gib": 0.5, "kv_how": "2 x 32 layers x 8 kv heads x 128 dims x 2 bytes x 4096 tokens (an upper bound: sliding/linear layers counted full)", "line_gib": 25.0, "predicted_growth_gib": 5.327610231935978, "rule": "store bytes + KV(num_ctx, f16) + 1.0 GiB <= 25.0 GiB beside (compute M-8); the guard's own ceiling on the bench's measured memory is 26 GiB", "store_gib": 3.827610231935978 }, "argv": [ "/workshop/bench-coi-2026-09-23/coilargecard.py", "--arm-id", "s-mistral-v0.1-spread-cd", "--model", "mistral:v0.1", "--num-ctx", "4096", "--think", "off", "--tasks", "c,d", "--mode", "full", "--ceiling-mib", "119325", "--out", "/workshop/bench-coi-2026-09-23/rows/s-mistral-v0.1-spread-cd", "--probes", "--limit", "5" ], "arm_id": "s-mistral-v0.1-spread-cd", "baseline": { "bench_memory": { "cgroup_pids": [ 1946485 ], "per_board_mib": { "": 0.0, "": 0.0 }, "procs": [], "stamp": "2026-09-24T06:36:13Z" }, "boards": [ { "memory_mib": 2.0, "sm_mhz": 180.0, "temp_c": 32.0, "uuid": "", "watts": 18.82 }, { "memory_mib": 1.0, "sm_mhz": 210.0, "temp_c": 35.0, "uuid": "", "watts": 27.15 } ] }, "bench_unit_env": "BindsTo=coi-guard-desktop.service\nRuntimeMaxUSec=12min\nExecStart={ path=/workshop/bench-coi-2026-09-23/runtime/ollama-0.32.13/bin/ollama ; argv[]=/workshop/bench-coi-2026-09-23/runtime/ollama-0.32.13/bin/ollama serve ; ignore_errors=no ; start_time=[Thu 2026-09-24 06:35:48 UTC] ; stop_time=[n/a] ; pid=1946485 ; code=(null) ; status=0/0 }\nEnvironment=TZ=UTC OLLAMA_HOST=127.0.0.1:withheld-2026-10-14 OLLAMA_KEEP_ALIVE=-1 OLLAMA_MAX_LOADED_MODELS=1 OLLAMA_=1 OLLAMA_FLASH_ATTENTION=1 OLLAMA_KV_CACHE_TYPE=f16 OLLAMA_NOPRUNE=1 OLLAMA_MODELS=/workshop/bench-coi-2026-09-23/store OLLAMA_GPU_OVERHEAD=536870912 OLLAMA_LOAD_TIMEOUT=10m CUDA_VISIBLE_DEVICES=, OLLAMA_VULKAN=0 GGML_VK_VISIBLE_DEVICES=-1 OLLAMA_SCHED_SPREAD=1 HIP_VISIBLE_DEVICES= ROCR_VISIBLE_DEVICES=", "board": "", "box": "the desktop", "box_unreadable_polls": 0, "card_class": "a 96 GB workstation card (P-6: energy only, no dollar field)", "disk_floor_gib": 400.0, "disk_free_gib": 425.68235778808594, "ended": "2026-09-24T06:37:43Z", "guard_state": "active", "host": "http://127.0.0.1:withheld-2026-10-14", "idle_preload": { "board": { "cards": [ "", "" ], "joules": 1706.4275320601464, "mean_watts": 121.58372287939795, "peak_watts": 64.95, "per_card": { "": { "joules": 795.5037913513183, "mean_watts": 56.68, "min_watts": 56.68, "peak_watts": 56.68, "samples": 15, "seconds": 14.03499984741211 }, "": { "joules": 910.9237407088281, "mean_watts": 64.95, "min_watts": 64.95, "peak_watts": 64.95, "samples": 15, "seconds": 14.024999856948853 } }, "samples": 30, "seconds": 14.03499984741211 }, "box": { "instrument": "box delta (CP2000 ups.realpower, integer W; includes the desktop's 3090 and the host; never wall)", "joules": 3672.9870026111603, "mean_watts": 282.49399909315855, "median_s_between_changes": null, "min_watts": 159.0, "peak_watts": 328.0, "samples": 14, "seconds": 13.002000093460083, "value_changes": 1 }, "what": "the desktop's larger card + the desktop's 3090, every seat on the desktop stopped and NO bench model", "window_epoch": [ 1790231758.653, 1790231772.688 ] }, "instrument": { "coilargecard_py": "ac9f40ebcfb62b21a9d3e916f358a92119e6ee0781e8630999995f93ae41773d", "guard_py": "3c7d72c55e927c698a41fb2dfad84a9cd5d781d4f597c90d76b3a12a3fdbc16e", "jev/kit/doorman_planted.json": "4daf059c26a47a84fa3b73630a64ba1a348da72c2e5df7f3c28c007b67da22f7", "jev/kit/task_c.json": "61a0b1d6e7cf66fd91992c0c3a4572f6fbae90e6471048fd7dc1eb41ff2d7f39", "jev/run.py": "withheld-2026-10-14", "jev/run_addenda.py": "a34aca61833ed0c5860c91c7e2e227429e8ea9e2e50fec68d7b7fec15d89acab" }, "limit": 5, "mode": "full", "model": "mistral:v0.1", "num_ctx": 4096, "ollama_version": "0.32.13", "pin": { "api_ps_model": { "context_length": 4096, "details": { "families": [ "llama" ], "family": "llama", "format": "gguf", "parameter_size": "7B", "parent_model": "", "quantization_level": "Q4_0" }, "digest": "b17615239298ea5bacfc1c46aa1842737b833779c805542b78f6be29c516d2f4", "expires_at": "2319-01-04T06:24:01.940395176Z", "model": "mistral:v0.1", "name": "mistral:v0.1", "size": 4898559097, "size_vram": 4898559097 }, "bench_growth_mib": 5590.0, "bench_growth_per_board_mib": { "": 4376.0, "": 1214.0 }, "bench_memory_after": { "cgroup_pids": [ 1946485, 1947811 ], "per_board_mib": { "": 4376.0, "": 1214.0 }, "procs": [ { "mib": 4376.0, "name": "/workshop/bench-coi-2026-09-23/runtime/ollama-0.32.13/lib/ollama/llama-server", "pid": 1947811, "uuid": "" }, { "mib": 1214.0, "name": "/workshop/bench-coi-2026-09-23/runtime/ollama-0.32.13/lib/ollama/llama-server", "pid": 1947811, "uuid": "" } ], "stamp": "2026-09-24T06:36:48Z" }, "bench_on_othercard_mib": 1214.0, "board_total_before_mib": { "": 2.0, "": 1.0 }, "board_total_growth_mib": { "": 4386.0, "": 1224.0 }, "ok": true, "ps_size": 4898559097, "ps_size_vram": 4898559097, "ps_whole": true, "rule": "PER-PROCESS: WHOLE when the bench cgroup's own memory on the desktop's larger card grew by >= 90 % of the tag's store bytes AND /api/ps reports no CPU share; none of it on the desktop's 3090; the board totals are corroboration only (four live seats share the board)", "split": true, "store_bytes": 4109865192, "whole": true }, "postflight": { "after_unload": { "api_ps": { "models": [] }, "bench_memory": { "cgroup_pids": [ 1946485 ], "per_board_mib": { "": 0.0, "": 0.0 }, "procs": [], "stamp": "2026-09-24T06:37:43Z" }, "rows": [ { "memory_mib": 2.0, "sm_mhz": 405.0, "temp_c": 38.0, "uuid": "", "watts": 70.13 }, { "memory_mib": 1.0, "sm_mhz": 210.0, "temp_c": 46.0, "uuid": "", "watts": 55.24 } ], "stamp": "2026-09-24T06:37:43Z" }, "posture_after": { "boards_on_bus": [ "0, , NVIDIA RTX PRO 6000 Blackwell Workstation Edition, withheld-2026-10-14", "1, , NVIDIA GeForce RTX 3090, withheld-2026-10-14" ], "card": { "clocks.max.sm": "3090", "clocks.mem": "13365", "clocks.sm": "2640", "clocks_throttle_reasons.active": "0x0000000000000000", "driver_version": "595.84", "enforced.power.limit": "420.00", "memory.free": "92847", "memory.total": "97887", "memory.used": "4404", "name": "NVIDIA RTX PRO 6000 Blackwell Workstation Edition", "pci.bus_id": "withheld-2026-10-14", "pcie.link.gen.current": "4", "pcie.link.gen.max": "4", "pcie.link.width.current": "16", "pcie.link.width.max": "16", "persistence_mode": "Disabled", "power.default_limit": "600.00", "power.draw": "139.53", "power.limit": "420.00", "power.max_limit": "600.00", "pstate": "P1", "temperature.gpu": "42", "utilization.gpu": "19", "uuid": "", "vbios_version": "98.02.81.00.07" }, "compute_apps": [ ", 1947811, /workshop/bench-coi-2026-09-23/runtime/ollama-0.32.13/lib/ollama/llama-server, 4394 MiB", ", 1947811, /workshop/bench-coi-2026-09-23/runtime/ollama-0.32.13/lib/ollama/llama-server, 1226 MiB" ], "how": "nvidia-smi -i ; enforced.power.limit READ, never set (the seats' posture)", "stamp": "2026-09-24T06:37:37Z" } }, "posture_before": { "boards_on_bus": [ "0, , NVIDIA RTX PRO 6000 Blackwell Workstation Edition, withheld-2026-10-14", "1, , NVIDIA GeForce RTX 3090, withheld-2026-10-14" ], "card": { "clocks.max.sm": "3090", "clocks.mem": "14001", "clocks.sm": "2610", "clocks_throttle_reasons.active": "0x0000000000000000", "driver_version": "595.84", "enforced.power.limit": "420.00", "memory.free": "97249", "memory.total": "97887", "memory.used": "2", "name": "NVIDIA RTX PRO 6000 Blackwell Workstation Edition", "pci.bus_id": "withheld-2026-10-14", "pcie.link.gen.current": "4", "pcie.link.gen.max": "4", "pcie.link.width.current": "16", "pcie.link.width.max": "16", "persistence_mode": "Disabled", "power.default_limit": "600.00", "power.draw": "91.53", "power.limit": "420.00", "power.max_limit": "600.00", "pstate": "P0", "temperature.gpu": "34", "utilization.gpu": "0", "uuid": "", "vbios_version": "98.02.81.00.07" }, "compute_apps": [], "how": "nvidia-smi -i ; enforced.power.limit READ, never set (the seats' posture)", "stamp": "2026-09-24T06:35:55Z" }, "posture_othercard_before": { "clocks.max.sm": "2130", "clocks.mem": "9751", "clocks.sm": "1725", "clocks_throttle_reasons.active": "0x0000000000000000", "driver_version": "595.84", "enforced.power.limit": "300.00", "memory.free": "24124", "memory.total": "24576", "memory.used": "1", "name": "NVIDIA GeForce RTX 3090", "pci.bus_id": "withheld-2026-10-14", "pcie.link.gen.current": "4", "pcie.link.gen.max": "4", "pcie.link.width.current": "16", "pcie.link.width.max": "16", "persistence_mode": "Disabled", "power.default_limit": "350.00", "power.draw": "112.11", "power.limit": "300.00", "power.max_limit": "366.00", "pstate": "P0", "temperature.gpu": "37", "utilization.gpu": "0", "uuid": "", "vbios_version": "94.02.42.C0.10" }, "posture_verdict": "420.00 W enforced as run; the desktop's larger card + the desktop's 3090 on the bus", "probes": { "decode": { "board": { "cards": [ "", "" ], "joules": 1551.0695893085003, "mean_watts": 386.5112347085049, "peak_watts": 322.14, "per_card": { "": { "joules": 959.2013452410697, "mean_watts": 239.02350922144598, "min_watts": 107.0, "peak_watts": 322.14, "samples": 5, "seconds": 4.013000011444092 }, "": { "joules": 591.8682440674305, "mean_watts": 147.70856335379892, "min_watts": 115.9, "peak_watts": 167.1, "samples": 5, "seconds": 4.007000207901001 } }, "samples": 10, "seconds": 4.013000011444092 }, "box": { "instrument": "box delta (CP2000 ups.realpower, integer W; includes the desktop's 3090 and the host; never wall)", "joules": 804.4710030555725, "mean_watts": 267.0, "median_s_between_changes": null, "min_watts": 267.0, "peak_watts": 267.0, "samples": 4, "seconds": 3.013000011444092, "value_changes": 0 }, "num_predict": 256, "prompt": "Describe, in about three hundred words, a lighthouse on a rocky coast during a winter storm.", "runs": [ { "decode_tok_s": 239.7854669400471, "done_reason": "length", "end_epoch": 1790231836.1496778, "eval_count": 256, "eval_s": 1.067621, "load_s": 0.147295125, "prefill_tok_s": 3750.586029067042, "prompt_eval_count": 32, "prompt_eval_s": 0.008532, "start_epoch": 1790231834.8391304, "text_head": "A lighthouse stands tall and proud on a rocky coast, overlooking the churning ocean. The waves crash against the cliffs,", "thinking_chars": 0, "total_s": 1.279116275 }, { "decode_tok_s": 240.23505498661348, "done_reason": "length", "end_epoch": 1790231837.469826, "eval_count": 256, "eval_s": 1.065623, "load_s": 0.187151729, "prefill_tok_s": 5207.485760781123, "prompt_eval_count": 32, "prompt_eval_s": 0.006145, "start_epoch": 1790231836.1497357, "text_head": "A lighthouse stands tall and proud on a rocky coast, overlooking the churning ocean. The waves crash against the cliffs,", "thinking_chars": 0, "total_s": 1.291049378 }, { "decode_tok_s": 240.62841614804663, "done_reason": "length", "end_epoch": 1790231838.7672296, "eval_count": 256, "eval_s": 1.063881, "load_s": 0.16902059, "prefill_tok_s": 6334.125098970705, "prompt_eval_count": 32, "prompt_eval_s": 0.005052, "start_epoch": 1790231837.4698772, "text_head": "A lighthouse stands tall and proud on a rocky coast, overlooking the churning ocean. The waves crash against the cliffs,", "thinking_chars": 0, "total_s": 1.270701988 } ] }, "prefill": { "board": { "cards": [ "", "" ], "joules": 4362.446548572779, "mean_watts": 335.18606172300775, "peak_watts": 227.08, "per_card": { "": { "joules": 2601.0369558727743, "mean_watts": 199.84917268962872, "min_watts": 104.26, "peak_watts": 227.08, "samples": 14, "seconds": 13.014999866485596 }, "": { "joules": 1761.4095927000046, "mean_watts": 135.3993094364935, "min_watts": 121.85, "peak_watts": 140.92, "samples": 14, "seconds": 13.008999824523926 } }, "samples": 28, "seconds": 13.014999866485596 }, "box": { "instrument": "box delta (CP2000 ups.realpower, integer W; includes the desktop's 3090 and the host; never wall)", "joules": 7465.243909835815, "mean_watts": 622.0, "median_s_between_changes": null, "min_watts": 622.0, "peak_watts": 622.0, "samples": 13, "seconds": 12.001999855041504, "value_changes": 0 }, "runs": [ { "decode_tok_s": 1000000.0, "done_reason": "length", "end_epoch": 1790231840.2967658, "eval_count": 1, "eval_s": 1e-06, "load_s": 0.186136673, "prefill_tok_s": 16309.725404256733, "prompt_eval_count": 2456, "prompt_eval_s": 0.150585, "start_epoch": 1790231839.9020472, "text_head": "\n", "thinking_chars": 0, "total_s": 0.391155003 }, { "decode_tok_s": 1000000.0, "done_reason": "length", "end_epoch": 1790231840.9066696, "eval_count": 1, "eval_s": 1e-06, "load_s": 0.18686048, "prefill_tok_s": 16620.04835661991, "prompt_eval_count": 2454, "prompt_eval_s": 0.147653, "start_epoch": 1790231840.2968397, "text_head": "\n", "thinking_chars": 0, "total_s": 0.608222765 }, { "decode_tok_s": 1000000.0, "done_reason": "length", "end_epoch": 1790231841.518549, "eval_count": 1, "eval_s": 1e-06, "load_s": 0.188917452, "prefill_tok_s": 16603.43160130122, "prompt_eval_count": 2455, "prompt_eval_s": 0.147861, "start_epoch": 1790231840.9067354, "text_head": "\n", "thinking_chars": 0, "total_s": 0.610305387 }, { "decode_tok_s": 1000000.0, "done_reason": "length", "end_epoch": 1790231842.1414568, "eval_count": 1, "eval_s": 1e-06, "load_s": 0.197224, "prefill_tok_s": 16581.11576387951, "prompt_eval_count": 2455, "prompt_eval_s": 0.14806, "start_epoch": 1790231841.5186114, "text_head": "\n", "thinking_chars": 0, "total_s": 0.621469754 }, { "decode_tok_s": 1000000.0, "done_reason": "length", "end_epoch": 1790231842.7472363, "eval_count": 1, "eval_s": 1e-06, "load_s": 0.182640367, "prefill_tok_s": 16616.283552111017, "prompt_eval_count": 2457, "prompt_eval_s": 0.147867, "start_epoch": 1790231842.1415105, "text_head": "\n", "thinking_chars": 0, "total_s": 0.604232711 }, { "decode_tok_s": 1000000.0, "done_reason": "length", "end_epoch": 1790231843.3539302, "eval_count": 1, "eval_s": 1e-06, "load_s": 0.188144625, "prefill_tok_s": 16555.374757779744, "prompt_eval_count": 2452, "prompt_eval_s": 0.148109, "start_epoch": 1790231842.7473, "text_head": "\n", "thinking_chars": 0, "total_s": 0.605317886 }, { "decode_tok_s": 1000000.0, "done_reason": "length", "end_epoch": 1790231843.95571, "eval_count": 1, "eval_s": 1e-06, "load_s": 0.188033836, "prefill_tok_s": 16600.427177116286, "prompt_eval_count": 2456, "prompt_eval_s": 0.147948, "start_epoch": 1790231843.3539784, "text_head": "\n", "thinking_chars": 0, "total_s": 0.600277796 }, { "decode_tok_s": 1000000.0, "done_reason": "length", "end_epoch": 1790231844.5590906, "eval_count": 1, "eval_s": 1e-06, "load_s": 0.182111608, "prefill_tok_s": 16601.6946858334, "prompt_eval_count": 2453, "prompt_eval_s": 0.147756, "start_epoch": 1790231843.9557729, "text_head": "\n", "thinking_chars": 0, "total_s": 0.60165734 }, { "decode_tok_s": 1000000.0, "done_reason": "length", "end_epoch": 1790231845.1613023, "eval_count": 1, "eval_s": 1e-06, "load_s": 0.180856053, "prefill_tok_s": 16597.84687540126, "prompt_eval_count": 2456, "prompt_eval_s": 0.147971, "start_epoch": 1790231844.5591455, "text_head": "\n", "thinking_chars": 0, "total_s": 0.600596684 }, { "decode_tok_s": 1000000.0, "done_reason": "length", "end_epoch": 1790231845.7387602, "eval_count": 1, "eval_s": 1e-06, "load_s": 0.166305221, "prefill_tok_s": 16597.398209156952, "prompt_eval_count": 2456, "prompt_eval_s": 0.147975, "start_epoch": 1790231845.1613688, "text_head": "\n", "thinking_chars": 0, "total_s": 0.575984871 }, { "decode_tok_s": 1000000.0, "done_reason": "length", "end_epoch": 1790231846.3302834, "eval_count": 1, "eval_s": 1e-06, "load_s": 0.171304201, "prefill_tok_s": 16574.026604786704, "prompt_eval_count": 2457, "prompt_eval_s": 0.148244, "start_epoch": 1790231845.7388153, "text_head": "\n", "thinking_chars": 0, "total_s": 0.590028936 }, { "decode_tok_s": 1000000.0, "done_reason": "length", "end_epoch": 1790231846.9378438, "eval_count": 1, "eval_s": 1e-06, "load_s": 0.190676986, "prefill_tok_s": 16506.245333656196, "prompt_eval_count": 2454, "prompt_eval_s": 0.148671, "start_epoch": 1790231846.3303418, "text_head": "\n", "thinking_chars": 0, "total_s": 0.606199083 }, { "decode_tok_s": 1000000.0, "done_reason": "length", "end_epoch": 1790231847.5206838, "eval_count": 1, "eval_s": 1e-06, "load_s": 0.16885855, "prefill_tok_s": 16615.138989426116, "prompt_eval_count": 2456, "prompt_eval_s": 0.147817, "start_epoch": 1790231846.9378853, "text_head": "\n", "thinking_chars": 0, "total_s": 0.581455539 }, { "decode_tok_s": 1000000.0, "done_reason": "length", "end_epoch": 1790231848.1222818, "eval_count": 1, "eval_s": 1e-06, "load_s": 0.189969969, "prefill_tok_s": 16618.247567194197, "prompt_eval_count": 2454, "prompt_eval_s": 0.147669, "start_epoch": 1790231847.5207274, "text_head": "\n", "thinking_chars": 0, "total_s": 0.600146597 }, { "decode_tok_s": 1000000.0, "done_reason": "length", "end_epoch": 1790231848.7263348, "eval_count": 1, "eval_s": 1e-06, "load_s": 0.187972536, "prefill_tok_s": 16640.89836439092, "prompt_eval_count": 2454, "prompt_eval_s": 0.147468, "start_epoch": 1790231848.122324, "text_head": "\n", "thinking_chars": 0, "total_s": 0.602583628 }, { "decode_tok_s": 1000000.0, "done_reason": "length", "end_epoch": 1790231849.3234909, "eval_count": 1, "eval_s": 1e-06, "load_s": 0.182185999, "prefill_tok_s": 16614.23981058684, "prompt_eval_count": 2456, "prompt_eval_s": 0.147825, "start_epoch": 1790231848.7264035, "text_head": "\n", "thinking_chars": 0, "total_s": 0.595667224 }, { "decode_tok_s": 1000000.0, "done_reason": "length", "end_epoch": 1790231849.9218693, "eval_count": 1, "eval_s": 1e-06, "load_s": 0.185809394, "prefill_tok_s": 16614.44339922714, "prompt_eval_count": 2455, "prompt_eval_s": 0.147763, "start_epoch": 1790231849.323534, "text_head": "\n", "thinking_chars": 0, "total_s": 0.596893949 }, { "decode_tok_s": 1000000.0, "done_reason": "length", "end_epoch": 1790231850.51822, "eval_count": 1, "eval_s": 1e-06, "load_s": 0.180004277, "prefill_tok_s": 16636.84909598547, "prompt_eval_count": 2455, "prompt_eval_s": 0.147564, "start_epoch": 1790231849.9219325, "text_head": "\n", "thinking_chars": 0, "total_s": 0.594864317 }, { "decode_tok_s": 1000000.0, "done_reason": "length", "end_epoch": 1790231851.1170263, "eval_count": 1, "eval_s": 1e-06, "load_s": 0.182339578, "prefill_tok_s": 16628.72011709221, "prompt_eval_count": 2454, "prompt_eval_s": 0.147576, "start_epoch": 1790231850.5182652, "text_head": "\n", "thinking_chars": 0, "total_s": 0.597348977 }, { "decode_tok_s": 1000000.0, "done_reason": "length", "end_epoch": 1790231851.7164695, "eval_count": 1, "eval_s": 1e-06, "load_s": 0.187124829, "prefill_tok_s": 16606.503084935626, "prompt_eval_count": 2452, "prompt_eval_s": 0.147653, "start_epoch": 1790231851.117094, "text_head": "\n", "thinking_chars": 0, "total_s": 0.597934546 }, { "decode_tok_s": 1000000.0, "done_reason": "length", "end_epoch": 1790231852.316577, "eval_count": 1, "eval_s": 1e-06, "load_s": 0.183837902, "prefill_tok_s": 16604.10537350783, "prompt_eval_count": 2455, "prompt_eval_s": 0.147855, "start_epoch": 1790231851.716531, "text_head": "\n", "thinking_chars": 0, "total_s": 0.598467553 } ], "target_tokens_approx": 3000.0 }, "ttft": { "how": "the first 12 task_c prompts, run.py's body with stream true, in task order; TTFT = client clock on localhost from the request to the first chunk carrying content or thinking", "median_s": 0.25082387041766196, "p95_s": 0.26371916104108095, "runs": [ { "end_epoch": 1790231854.4432738, "eval_count": 7, "prompt_eval_count": 266, "server_load_ms": 180.170917, "server_prompt_eval_ms": 5.237, "start_epoch": 1790231854.002795, "total_s": 0.44035082100890577, "ttft_s": 0.41913272405508906 }, { "end_epoch": 1790231854.693157, "eval_count": 8, "prompt_eval_count": 367, "server_load_ms": 168.440872, "server_prompt_eval_ms": 4.2, "start_epoch": 1790231854.4433455, "total_s": 0.24971186404582113, "ttft_s": 0.22038872900884598 }, { "end_epoch": 1790231854.9533722, "eval_count": 6, "prompt_eval_count": 280, "server_load_ms": 182.812576, "server_prompt_eval_ms": 4.184999, "start_epoch": 1790231854.693247, "total_s": 0.26003233494702727, "ttft_s": 0.2430449309758842 }, { "end_epoch": 1790231855.2238863, "eval_count": 8, "prompt_eval_count": 389, "server_load_ms": 187.646718, "server_prompt_eval_ms": 4.213, "start_epoch": 1790231854.9534373, "total_s": 0.2703537350753322, "ttft_s": 0.24090996000450104 }, { "end_epoch": 1790231855.4947762, "eval_count": 7, "prompt_eval_count": 392, "server_load_ms": 182.607087, "server_prompt_eval_ms": 4.208, "start_epoch": 1790231855.2239606, "total_s": 0.27071093290578574, "ttft_s": 0.2495738249272108 }, { "end_epoch": 1790231855.7888172, "eval_count": 8, "prompt_eval_count": 393, "server_load_ms": 187.391208, "server_prompt_eval_ms": 28.53, "start_epoch": 1790231855.4948392, "total_s": 0.2938821429852396, "ttft_s": 0.26254716503899544 }, { "end_epoch": 1790231856.0687082, "eval_count": 7, "prompt_eval_count": 307, "server_load_ms": 182.671476, "server_prompt_eval_ms": 23.408, "start_epoch": 1790231855.7889144, "total_s": 0.27966883790213615, "ttft_s": 0.2520739159081131 }, { "end_epoch": 1790231856.3390603, "eval_count": 7, "prompt_eval_count": 278, "server_load_ms": 184.203381, "server_prompt_eval_ms": 21.693, "start_epoch": 1790231856.06877, "total_s": 0.2701993950176984, "ttft_s": 0.24780204298440367 }, { "end_epoch": 1790231856.6185005, "eval_count": 7, "prompt_eval_count": 359, "server_load_ms": 191.369362, "server_prompt_eval_ms": 25.394, "start_epoch": 1790231856.3391259, "total_s": 0.27926027006469667, "ttft_s": 0.2567435579840094 }, { "end_epoch": 1790231856.9040504, "eval_count": 8, "prompt_eval_count": 383, "server_load_ms": 182.183489, "server_prompt_eval_ms": 28.048, "start_epoch": 1790231856.6185641, "total_s": 0.2853941459907219, "ttft_s": 0.25369384000077844 }, { "end_epoch": 1790231857.1780365, "eval_count": 7, "prompt_eval_count": 387, "server_load_ms": 172.772155, "server_prompt_eval_ms": 28.133, "start_epoch": 1790231856.9041107, "total_s": 0.2738384410040453, "ttft_s": 0.24640977790113539 }, { "end_epoch": 1790231857.4734294, "eval_count": 8, "prompt_eval_count": 384, "server_load_ms": 190.773066, "server_prompt_eval_ms": 28.081, "start_epoch": 1790231857.1781032, "total_s": 0.2952313970308751, "ttft_s": 0.26371916104108095 } ] } }, "show": { "capabilities": [ "completion" ], "details": { "families": [ "llama" ], "family": "llama", "format": "gguf", "parameter_size": "7B", "parent_model": "", "quantization_level": "Q4_0" }, "model_info": { "general.architecture": "llama", "general.file_type": 2, "general.parameter_count": 7241732096, "general.quantization_version": 2, "llama.attention.head_count": 32, "llama.attention.head_count_kv": 8, "llama.attention.layer_norm_rms_epsilon": 1e-05, "llama.block_count": 32, "llama.context_length": 32768, "llama.embedding_length": 4096, "llama.feed_forward_length": 14336, "llama.rope.dimension_count": 128, "llama.rope.freq_base": 10000, "tokenizer.ggml.add_bos_token": true, "tokenizer.ggml.add_eos_token": false, "tokenizer.ggml.bos_token_id": 1, "tokenizer.ggml.eos_token_id": 2, "tokenizer.ggml.merges": null, "tokenizer.ggml.model": "llama", "tokenizer.ggml.scores": null, "tokenizer.ggml.token_type": null, "tokenizer.ggml.tokens": null, "tokenizer.ggml.unknown_token_id": 0 }, "parameters": "stop \"[INST]\"\nstop \"[/INST]\"", "system_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", "template_head": "[INST] {{ .System }} {{ .Prompt }} [/INST]\n", "template_sha256": "22e1b2e8dc2fbc3ac38b50f59e49f594034462c1cd02764353a8a076d97c3a59" }, "started": "2026-09-24T06:35:55Z", "tag": { "details": { "context_length": 32768, "embedding_length": 4096, "families": [ "llama" ], "family": "llama", "format": "gguf", "parameter_size": "7B", "parent_model": "", "quantization_level": "Q4_0" }, "digest": "b17615239298ea5bacfc1c46aa1842737b833779c805542b78f6be29c516d2f4", "modified_at": "2026-09-23T19:10:40.495213565Z", "name": "mistral:v0.1", "size": 4109865192 }, "tasks": [ "c", "d" ], "tasks_run": { "c": { "board_idle_window_epoch": [ 1790231808.406, 1790231817.417 ], "board_task_window_epoch": [ 1790231818.468, 1790231819.476 ], "box_idle_resident": { "instrument": "box delta (CP2000 ups.realpower, integer W; includes the desktop's 3090 and the host; never wall)", "joules": 2421.0470366477966, "mean_watts": 302.5552372370606, "median_s_between_changes": null, "min_watts": 227.0, "peak_watts": 320.0, "samples": 9, "seconds": 8.002000093460083, "value_changes": 1 }, "box_task": { "instrument": "box delta (CP2000 ups.realpower, integer W; includes the desktop's 3090 and the host; never wall)", "joules": null, "samples": 1, "seconds": null }, "energy": { "cards": [ "", "" ], "decisions": 5, "joules": 214.9506715595722, "joules_per_decision": 42.990134311914446, "mean_watts": 213.24468520000312, "net_joules_per_decision": 3.2840757548666604, "peak_watts": 116.27, "per_card": { "": { "joules": 101.72233366608619, "mean_watts": 100.91499999999999, "min_watts": 85.56, "peak_watts": 116.27, "samples": 2, "seconds": 1.008000135421753 }, "": { "joules": 113.22833789348603, "mean_watts": 112.665, "min_watts": 110.79, "peak_watts": 114.54, "samples": 2, "seconds": 1.005000114440918 } }, "samples": 4, "seconds": 1.008000135421753 }, "idle": { "cards": [ "", "" ], "joules": 1774.758193449974, "mean_watts": 196.95462908065258, "peak_watts": 111.06, "per_card": { "": { "joules": 776.3359484398364, "mean_watts": 86.1542486921639, "min_watts": 85.62, "peak_watts": 87.03, "samples": 10, "seconds": 9.011000156402588 }, "": { "joules": 998.4222450101375, "mean_watts": 110.83728130554603, "min_watts": 110.7, "peak_watts": 111.06, "samples": 10, "seconds": 9.008000135421753 } }, "samples": 20, "seconds": 9.011000156402588, "what": "the draw of these cards immediately before the arm, with the model resident and no work in flight" }, "posture_end": { "boards_on_bus": [ "0, , NVIDIA RTX PRO 6000 Blackwell Workstation Edition, withheld-2026-10-14", "1, , NVIDIA GeForce RTX 3090, withheld-2026-10-14" ], "card": { "clocks.max.sm": "3090", "clocks.mem": "13365", "clocks.sm": "2610", "clocks_throttle_reasons.active": "0x0000000000000000", "driver_version": "595.84", "enforced.power.limit": "420.00", "memory.free": "92851", "memory.total": "97887", "memory.used": "4400", "name": "NVIDIA RTX PRO 6000 Blackwell Workstation Edition", "pci.bus_id": "withheld-2026-10-14", "pcie.link.gen.current": "4", "pcie.link.gen.max": "4", "pcie.link.width.current": "16", "pcie.link.width.max": "16", "persistence_mode": "Disabled", "power.default_limit": "600.00", "power.draw": "100.32", "power.limit": "420.00", "power.max_limit": "600.00", "pstate": "P1", "temperature.gpu": "35", "utilization.gpu": "0", "uuid": "", "vbios_version": "98.02.81.00.07" }, "compute_apps": [ ", 1947811, /workshop/bench-coi-2026-09-23/runtime/ollama-0.32.13/lib/ollama/llama-server, 4390 MiB", ", 1947811, /workshop/bench-coi-2026-09-23/runtime/ollama-0.32.13/lib/ollama/llama-server, 1222 MiB" ], "how": "nvidia-smi -i ; enforced.power.limit READ, never set (the seats' posture)", "stamp": "2026-09-24T06:37:01Z" }, "posture_start": { "boards_on_bus": [ "0, , NVIDIA RTX PRO 6000 Blackwell Workstation Edition, withheld-2026-10-14", "1, , NVIDIA GeForce RTX 3090, withheld-2026-10-14" ], "card": { "clocks.max.sm": "3090", "clocks.mem": "13365", "clocks.sm": "2610", "clocks_throttle_reasons.active": "0x0000000000000000", "driver_version": "595.84", "enforced.power.limit": "420.00", "memory.free": "92865", "memory.total": "97887", "memory.used": "4386", "name": "NVIDIA RTX PRO 6000 Blackwell Workstation Edition", "pci.bus_id": "withheld-2026-10-14", "pcie.link.gen.current": "4", "pcie.link.gen.max": "4", "pcie.link.width.current": "16", "pcie.link.width.max": "16", "persistence_mode": "Disabled", "power.default_limit": "600.00", "power.draw": "86.57", "power.limit": "420.00", "power.max_limit": "600.00", "pstate": "P1", "temperature.gpu": "35", "utilization.gpu": "0", "uuid": "", "vbios_version": "98.02.81.00.07" }, "compute_apps": [ ", 1947811, /workshop/bench-coi-2026-09-23/runtime/ollama-0.32.13/lib/ollama/llama-server, 4376 MiB", ", 1947811, /workshop/bench-coi-2026-09-23/runtime/ollama-0.32.13/lib/ollama/llama-server, 1214 MiB" ], "how": "nvidia-smi -i ; enforced.power.limit READ, never set (the seats' posture)", "stamp": "2026-09-24T06:36:48Z" }, "summary": { "accuracy": 0.2, "brier": 1.6, "correct": 1, "errors": 0, "eval_tokens_total": 36, "floored_rows": 0, "letters_floored_total": 0, "median_seconds": 0.27879087103065103, "n": 5, "n_labelled": 5, "n_reached": 5, "p95_seconds": 0.2896995289484039, "prompt_tokens_total": 1694, "refused": 0, "refused_by_reason": {}, "reliability": [ { "accuracy": null, "correct": 0, "gap": null, "hi": 0.1, "lo": 0.0, "mean_p": null, "n": 0, "sum_p": 0.0 }, { "accuracy": null, "correct": 0, "gap": null, "hi": 0.2, "lo": 0.1, "mean_p": null, "n": 0, "sum_p": 0.0 }, { "accuracy": null, "correct": 0, "gap": null, "hi": 0.3, "lo": 0.2, "mean_p": null, "n": 0, "sum_p": 0.0 }, { "accuracy": null, "correct": 0, "gap": null, "hi": 0.4, "lo": 0.3, "mean_p": null, "n": 0, "sum_p": 0.0 }, { "accuracy": null, "correct": 0, "gap": null, "hi": 0.5, "lo": 0.4, "mean_p": null, "n": 0, "sum_p": 0.0 }, { "accuracy": null, "correct": 0, "gap": null, "hi": 0.6, "lo": 0.5, "mean_p": null, "n": 0, "sum_p": 0.0 }, { "accuracy": null, "correct": 0, "gap": null, "hi": 0.7, "lo": 0.6, "mean_p": null, "n": 0, "sum_p": 0.0 }, { "accuracy": null, "correct": 0, "gap": null, "hi": 0.8, "lo": 0.7, "mean_p": null, "n": 0, "sum_p": 0.0 }, { "accuracy": null, "correct": 0, "gap": null, "hi": 0.9, "lo": 0.8, "mean_p": null, "n": 0, "sum_p": 0.0 }, { "accuracy": 0.2, "correct": 1, "gap": 0.8, "hi": 1.0, "lo": 0.9, "mean_p": 1.0, "n": 5, "sum_p": 5.0 } ], "unparsed": 0 }, "wall_seconds": 1.4032654899638146 }, "d": { "board_idle_window_epoch": [ 1790231821.734, 1790231830.745 ], "board_task_window_epoch": [ 1790231831.803, 1790231832.811 ], "box_idle_resident": { "instrument": "box delta (CP2000 ups.realpower, integer W; includes the desktop's 3090 and the host; never wall)", "joules": 2269.0869574546814, "mean_watts": 283.56498357408583, "median_s_between_changes": null, "min_watts": 267.0, "peak_watts": 320.0, "samples": 9, "seconds": 8.001999855041504, "value_changes": 1 }, "box_task": { "instrument": "box delta (CP2000 ups.realpower, integer W; includes the desktop's 3090 and the host; never wall)", "joules": null, "samples": 1, "seconds": null }, "energy": { "cards": [ "", "" ], "decisions": 5, "joules": 216.08402670741083, "joules_per_decision": 43.216805341482164, "mean_watts": 214.36904531466163, "net_joules_per_decision": 3.326216506979108, "peak_watts": 117.76, "per_card": { "": { "joules": 102.63961378932, "mean_watts": 101.825, "min_watts": 85.89, "peak_watts": 117.76, "samples": 2, "seconds": 1.008000135421753 }, "": { "joules": 113.44441291809082, "mean_watts": 112.88, "min_watts": 111.25, "peak_watts": 114.51, "samples": 2, "seconds": 1.005000114440918 } }, "samples": 4, "seconds": 1.008000135421753 }, "idle": { "cards": [ "", "" ], "joules": 1783.0061727404593, "mean_watts": 197.86995771489956, "peak_watts": 114.26, "per_card": { "": { "joules": 781.3246339821815, "mean_watts": 86.70787272151965, "min_watts": 85.69, "peak_watts": 100.32, "samples": 10, "seconds": 9.010999917984009 }, "": { "joules": 1001.6815387582778, "mean_watts": 111.19910637338286, "min_watts": 110.94, "peak_watts": 114.26, "samples": 10, "seconds": 9.007999897003174 } }, "samples": 20, "seconds": 9.010999917984009, "what": "the draw of these cards immediately before the arm, with the model resident and no work in flight" }, "posture_end": { "boards_on_bus": [ "0, , NVIDIA RTX PRO 6000 Blackwell Workstation Edition, withheld-2026-10-14", "1, , NVIDIA GeForce RTX 3090, withheld-2026-10-14" ], "card": { "clocks.max.sm": "3090", "clocks.mem": "13365", "clocks.sm": "2610", "clocks_throttle_reasons.active": "0x0000000000000000", "driver_version": "595.84", "enforced.power.limit": "420.00", "memory.free": "92847", "memory.total": "97887", "memory.used": "4404", "name": "NVIDIA RTX PRO 6000 Blackwell Workstation Edition", "pci.bus_id": "withheld-2026-10-14", "pcie.link.gen.current": "4", "pcie.link.gen.max": "4", "pcie.link.width.current": "16", "pcie.link.width.max": "16", "persistence_mode": "Disabled", "power.default_limit": "600.00", "power.draw": "107.00", "power.limit": "420.00", "power.max_limit": "600.00", "pstate": "P1", "temperature.gpu": "36", "utilization.gpu": "0", "uuid": "", "vbios_version": "98.02.81.00.07" }, "compute_apps": [ ", 1947811, /workshop/bench-coi-2026-09-23/runtime/ollama-0.32.13/lib/ollama/llama-server, 4394 MiB", ", 1947811, /workshop/bench-coi-2026-09-23/runtime/ollama-0.32.13/lib/ollama/llama-server, 1226 MiB" ], "how": "nvidia-smi -i ; enforced.power.limit READ, never set (the seats' posture)", "stamp": "2026-09-24T06:37:14Z" }, "posture_start": { "boards_on_bus": [ "0, , NVIDIA RTX PRO 6000 Blackwell Workstation Edition, withheld-2026-10-14", "1, , NVIDIA GeForce RTX 3090, withheld-2026-10-14" ], "card": { "clocks.max.sm": "3090", "clocks.mem": "13365", "clocks.sm": "2610", "clocks_throttle_reasons.active": "0x0000000000000000", "driver_version": "595.84", "enforced.power.limit": "420.00", "memory.free": "92851", "memory.total": "97887", "memory.used": "4400", "name": "NVIDIA RTX PRO 6000 Blackwell Workstation Edition", "pci.bus_id": "withheld-2026-10-14", "pcie.link.gen.current": "4", "pcie.link.gen.max": "4", "pcie.link.width.current": "16", "pcie.link.width.max": "16", "persistence_mode": "Disabled", "power.default_limit": "600.00", "power.draw": "100.32", "power.limit": "420.00", "power.max_limit": "600.00", "pstate": "P1", "temperature.gpu": "35", "utilization.gpu": "0", "uuid": "", "vbios_version": "98.02.81.00.07" }, "compute_apps": [ ", 1947811, /workshop/bench-coi-2026-09-23/runtime/ollama-0.32.13/lib/ollama/llama-server, 4390 MiB", ", 1947811, /workshop/bench-coi-2026-09-23/runtime/ollama-0.32.13/lib/ollama/llama-server, 1222 MiB" ], "how": "nvidia-smi -i ; enforced.power.limit READ, never set (the seats' posture)", "stamp": "2026-09-24T06:37:01Z" }, "summary": { "accuracy": 0.0, "brier": 0.5, "correct": 0, "errors": 0, "eval_tokens_total": 40, "floored_rows": 0, "letters_floored_total": 0, "median_seconds": 0.32310157804749906, "n": 5, "n_labelled": 5, "n_reached": 5, "p95_seconds": 0.3373433630913496, "prompt_tokens_total": 2682, "refused": 0, "refused_by_reason": {}, "reliability": [ { "accuracy": null, "correct": 0, "gap": null, "hi": 0.1, "lo": 0.0, "mean_p": null, "n": 0, "sum_p": 0.0 }, { "accuracy": null, "correct": 0, "gap": null, "hi": 0.2, "lo": 0.1, "mean_p": null, "n": 0, "sum_p": 0.0 }, { "accuracy": null, "correct": 0, "gap": null, "hi": 0.3, "lo": 0.2, "mean_p": null, "n": 0, "sum_p": 0.0 }, { "accuracy": null, "correct": 0, "gap": null, "hi": 0.4, "lo": 0.3, "mean_p": null, "n": 0, "sum_p": 0.0 }, { "accuracy": null, "correct": 0, "gap": null, "hi": 0.5, "lo": 0.4, "mean_p": null, "n": 0, "sum_p": 0.0 }, { "accuracy": 0.0, "correct": 0, "gap": 0.5, "hi": 0.6, "lo": 0.5, "mean_p": 0.5, "n": 5, "sum_p": 2.5 }, { "accuracy": null, "correct": 0, "gap": null, "hi": 0.7, "lo": 0.6, "mean_p": null, "n": 0, "sum_p": 0.0 }, { "accuracy": null, "correct": 0, "gap": null, "hi": 0.8, "lo": 0.7, "mean_p": null, "n": 0, "sum_p": 0.0 }, { "accuracy": null, "correct": 0, "gap": null, "hi": 0.9, "lo": 0.8, "mean_p": null, "n": 0, "sum_p": 0.0 }, { "accuracy": null, "correct": 0, "gap": null, "hi": 1.0, "lo": 0.9, "mean_p": null, "n": 0, "sum_p": 0.0 } ], "unparsed": 5 }, "wall_seconds": 1.6031520100077614 } }, "think": "off", "think_budget": 8, "ups_before": { "available": true, "load_pct": 27, "read_utc": "2026-09-24T06:35:55Z", "ups.realpower": 328, "ups.status": "OL" }, "warm": { "done_reason": "length", "load_ms": 24934.582913, "seconds": 31.233268456999213 } }