{ "model_id": "x4", "model_tag": "nemotron3:33b", "box_class": "gpu-3080", "arm": "cpu-box", "arm_label": "cpu-box (CPU-only on the same box, size_vram == 0)", "base_url": "http://127.0.0.1:11439", "started_utc": "2026-09-10T13:06:19Z", "nproc": 20, "num_predict": 256, "prompt_path": "withheld-2026-11-10", "prompt_sha256": "90eedd0c53f9554ae3837674504fcb7090013d9a432a653352183c0f25a7ce5c", "prompt_bytes": 2101, "scored_runs_requested": 3, "warmup_discarded_runs": 1, "num_gpu_option": null, "wall_decode_note": "decode_tok_s is ollama's own eval_count/eval_duration. decode_tok_s_wall is (eval_count - 1) / (wall_s - ttft_s): the tokens AFTER the first, over the time after the first, measured by this client's clock -- it carries the HTTP hop and the stream decode that ollama's counter does not. decode_tok_s_wall_gross is eval_count / wall_s and therefore includes prefill and any load; it is the slowest and least comparable of the three, printed so nobody has to derive it.", "loadavg_before_arm": "10.06 9.70 9.11 2/814 1049774", "free_m_before": "total used free shared buff/cache available\nMem: 47040 3727 18985 9 24861 43313\nSwap: 24575 969 23606", "runs": [ { "ttft_ms": 491.74, "first_token_was_thinking": false, "wall_s": 20.0075, "eval_count": 256, "eval_duration_ns": 19590706000, "decode_tok_s": 13.067, "prompt_eval_count": 562, "prompt_eval_duration_ns": 163739000, "prefill_tok_s": 3432.292, "load_duration_ns": 246121342, "total_duration_ns": 20004511521, "streamed_chunks": 249, "response_chars": 1047, "thinking_chars": 0, "done_reason": "length", "decode_tok_s_wall": 13.066, "decode_tok_s_wall_gross": 12.795, "run": 1 }, { "ttft_ms": 488.0, "first_token_was_thinking": false, "wall_s": 19.9815, "eval_count": 256, "eval_duration_ns": 19568226000, "decode_tok_s": 13.082, "prompt_eval_count": 562, "prompt_eval_duration_ns": 163364000, "prefill_tok_s": 3440.17, "load_duration_ns": 243100401, "total_duration_ns": 19978654547, "streamed_chunks": 249, "response_chars": 1047, "thinking_chars": 0, "done_reason": "length", "decode_tok_s_wall": 13.081, "decode_tok_s_wall_gross": 12.812, "run": 2 }, { "ttft_ms": 484.13, "first_token_was_thinking": false, "wall_s": 19.9863, "eval_count": 256, "eval_duration_ns": 19576890000, "decode_tok_s": 13.077, "prompt_eval_count": 562, "prompt_eval_duration_ns": 164780000, "prefill_tok_s": 3410.608, "load_duration_ns": 237527604, "total_duration_ns": 19983418870, "streamed_chunks": 249, "response_chars": 1047, "thinking_chars": 0, "done_reason": "length", "decode_tok_s_wall": 13.075, "decode_tok_s_wall_gross": 12.809, "run": 3 } ], "errors": [], "nvidia_before": "0, , NVIDIA GeForce RTX 3080, 256, 11.19, 380.00, 24, 210", "smi_before": [ { "index": "0", "uuid": "", "name": "NVIDIA GeForce RTX 3080", "memory.used": 256.0, "power.draw": 11.19, "power.limit": 380.0, "temperature.gpu": 24.0, "clocks.sm": 210.0 } ], "ollama_version": "0.32.13", "quantization": "Q4_K_M", "param_size": "33.0B", "family": "nemotron_h_omni", "context_length": 131072, "think_mode": false, "think_note": "think:false accepted", "warmup_discarded": { "ttft_ms": 482.32, "first_token_was_thinking": false, "wall_s": 19.8827, "eval_count": 256, "eval_duration_ns": 19475586000, "decode_tok_s": 13.145, "prompt_eval_count": 562, "prompt_eval_duration_ns": 161499000, "prefill_tok_s": 3479.898, "load_duration_ns": 239019007, "total_duration_ns": 19879866691, "streamed_chunks": 249, "response_chars": 1047, "thinking_chars": 0, "done_reason": "length", "decode_tok_s_wall": 13.144, "decode_tok_s_wall_gross": 12.875 }, "ps_after_load": [ { "name": "nemotron3:33b", "model": "nemotron3:33b", "size": 24953225214, "digest": "f6d8b7ff496ccc53429cc480ad53971d522b443ee4a5aa58a6da49e57acf42cf", "details": { "parent_model": "", "format": "gguf", "family": "nemotron_h_omni", "families": [ "nemotron_h_omni" ], "parameter_size": "33.0B", "quantization_level": "Q4_K_M" }, "expires_at": "2026-09-10T13:16:57.348078259Z", "size_vram": 0, "context_length": 4096 } ], "size_total": 24953225214, "size_vram": 0, "gpu": { "arm": "cpu-box", "requested_uuid": "", "requested_index": null, "fits_on_card": false, "card_untouched": true, "size_bytes": 24953225214, "size_vram_bytes": 0, "offload_fraction": 0.0, "offload_split": "0.0% VRAM / 100.0% RAM", "verdict": "cpu-only: size_vram == 0, nothing on the card", "smi_before": [ { "index": "0", "uuid": "", "name": "NVIDIA GeForce RTX 3080", "memory.used": 256.0, "power.draw": 11.19, "power.limit": 380.0, "temperature.gpu": 24.0, "clocks.sm": 210.0 } ], "during": { "sample_hz": 1.0, "samples": 59, "errors": [], "per_gpu": { "": { "index": "0", "name": "NVIDIA GeForce RTX 3080", "power_w": { "min": 10.88, "median": 11.09, "max": 11.34, "mean": 11.1, "n": 59 }, "power_limit_w": 380.0, "memory_used_mib": { "min": 256.0, "median": 256.0, "max": 256.0, "mean": 256.0, "n": 59 }, "temperature_c": { "min": 24.0, "median": 25.0, "max": 25.0, "mean": 24.9, "n": 59 }, "clocks_sm_mhz": { "min": 210.0, "median": 210.0, "max": 210.0, "mean": 210.0, "n": 59 }, "n": 59 } }, "target_uuid": "", "target": { "index": "0", "name": "NVIDIA GeForce RTX 3080", "power_w": { "min": 10.88, "median": 11.09, "max": 11.34, "mean": 11.1, "n": 59 }, "power_limit_w": 380.0, "memory_used_mib": { "min": 256.0, "median": 256.0, "max": 256.0, "mean": 256.0, "n": 59 }, "temperature_c": { "min": 24.0, "median": 25.0, "max": 25.0, "mean": 24.9, "n": 59 }, "clocks_sm_mhz": { "min": 210.0, "median": 210.0, "max": 210.0, "mean": 210.0, "n": 59 }, "n": 59 } }, "smi_after": [ { "index": "0", "uuid": "", "name": "NVIDIA GeForce RTX 3080", "memory.used": 256.0, "power.draw": 10.95, "power.limit": 380.0, "temperature.gpu": 25.0, "clocks.sm": 210.0 } ] }, "scored": true, "peak_runner_rss_kb": 26083200, "peak_runner_rss_gib": 24.87, "peak_runner_threads": 32, "rss_samples": 117, "loadavg_during_runs": { "min": 8.64, "median": 9.17, "max": 9.45, "mean": 9.12, "samples": 117, "nproc": 20 }, "contention_suspected": false, "nvidia_after": "0, , NVIDIA GeForce RTX 3080, 256, 10.95, 380.00, 25, 210", "smi_after": [ { "index": "0", "uuid": "", "name": "NVIDIA GeForce RTX 3080", "memory.used": 256.0, "power.draw": 10.95, "power.limit": 380.0, "temperature.gpu": 25.0, "clocks.sm": 210.0 } ], "free_m_after": "total used free shared buff/cache available\nMem: 47040 29183 356 9 18034 17857\nSwap: 24575 969 23606", "derived": { "decode_tok_s": { "min": 13.067, "median": 13.077, "max": 13.082, "mean": 13.075, "n": 3 }, "decode_tok_s_wall": { "min": 13.066, "median": 13.075, "max": 13.081, "mean": 13.074, "n": 3 }, "ttft_ms": { "min": 484.13, "median": 488.0, "max": 491.74, "mean": 487.96, "n": 3 }, "wall_s": { "min": 19.9815, "median": 19.9863, "max": 20.0075, "mean": 19.9918, "n": 3 }, "prefill_tok_s": { "min": 3410.608, "median": 3432.292, "max": 3440.17, "mean": 3427.69, "n": 3 }, "eval_count": { "min": 256, "median": 256, "max": 256, "mean": 256.0, "n": 3 }, "thinking_chars_total": 0 }, "ttft_note": "Every scored run sends the SAME frozen prompt, so from the second request the KV prefix is cached and derived.ttft_ms is a CACHED-PREFIX first-token time, not a fresh-prompt one. LEDGER.md, 'THE HEADLINE TRAP'.", "ps_final": [], "finished_utc": "2026-09-10T13:07:58Z" }