{ "model_id": "m2", "model_tag": "gemma4:12b", "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": "2026-09-10T12:32:28Z", "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": "9.05 6.04 4.76 2/812 944044", "free_m_before": "total used free shared buff/cache available\nMem: 47040 6509 7197 9 33875 40531\nSwap: 24575 978 23597", "runs": [ { "ttft_ms": 1046.12, "first_token_was_thinking": false, "wall_s": 53.8635, "eval_count": 256, "eval_duration_ns": 52814504000, "decode_tok_s": 4.847, "prompt_eval_count": 536, "prompt_eval_duration_ns": 575831000, "prefill_tok_s": 930.829, "load_duration_ns": 465823375, "total_duration_ns": 53860283239, "streamed_chunks": 256, "response_chars": 1056, "thinking_chars": 0, "done_reason": "length", "decode_tok_s_wall": 4.828, "decode_tok_s_wall_gross": 4.753, "run": 1 }, { "ttft_ms": 1063.5, "first_token_was_thinking": false, "wall_s": 53.9049, "eval_count": 256, "eval_duration_ns": 52838763000, "decode_tok_s": 4.845, "prompt_eval_count": 536, "prompt_eval_duration_ns": 578764000, "prefill_tok_s": 926.112, "load_duration_ns": 472086646, "total_duration_ns": 53901839806, "streamed_chunks": 256, "response_chars": 1056, "thinking_chars": 0, "done_reason": "length", "decode_tok_s_wall": 4.826, "decode_tok_s_wall_gross": 4.749, "run": 2 }, { "ttft_ms": 1044.5, "first_token_was_thinking": false, "wall_s": 53.8565, "eval_count": 256, "eval_duration_ns": 52809142000, "decode_tok_s": 4.848, "prompt_eval_count": 536, "prompt_eval_duration_ns": 574737000, "prefill_tok_s": 932.6, "load_duration_ns": 465532175, "total_duration_ns": 53853225546, "streamed_chunks": 256, "response_chars": 1056, "thinking_chars": 0, "done_reason": "length", "decode_tok_s_wall": 4.828, "decode_tok_s_wall_gross": 4.753, "run": 3 } ], "errors": [], "nvidia_before": "0, , NVIDIA GeForce RTX 3080, 256, 11.54, 380.00, 36, 210", "smi_before": [ { "index": "0", "uuid": "", "name": "NVIDIA GeForce RTX 3080", "memory.used": 256.0, "power.draw": 11.54, "power.limit": 380.0, "temperature.gpu": 36.0, "clocks.sm": 210.0 } ], "ollama_version": "0.32.13", "quantization": "Q4_K_M", "param_size": "11.9B", "family": "gemma4", "context_length": 262144, "think_mode": false, "think_note": "think:false accepted", "warmup_discarded": { "ttft_ms": 1039.75, "first_token_was_thinking": false, "wall_s": 53.5837, "eval_count": 256, "eval_duration_ns": 52535664000, "decode_tok_s": 4.873, "prompt_eval_count": 536, "prompt_eval_duration_ns": 567954000, "prefill_tok_s": 943.738, "load_duration_ns": 467293478, "total_duration_ns": 53575027384, "streamed_chunks": 256, "response_chars": 1056, "thinking_chars": 0, "done_reason": "length", "decode_tok_s_wall": 4.853, "decode_tok_s_wall_gross": 4.778 }, "ps_after_load": [ { "name": "gemma4:12b", "model": "gemma4:12b", "size": 8897649703, "digest": "4eb23ef187e2c5462566d6a1d3bbbc2f1346d0b4327cbb66d58fffbcc9b2b05c", "details": { "parent_model": "", "format": "gguf", "family": "gemma4", "families": [ "gemma4" ], "parameter_size": "11.9B", "quantization_level": "Q4_K_M" }, "expires_at": "2026-09-10T12:43:44.154289131Z", "size_vram": 0, "context_length": 4096 } ], "size_total": 8897649703, "size_vram": 0, "gpu": { "arm": "cpu-box", "requested_uuid": "", "requested_index": null, "fits_on_card": false, "card_untouched": true, "size_bytes": 8897649703, "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": 11.54, "power.limit": 380.0, "temperature.gpu": 36.0, "clocks.sm": 210.0 } ], "during": { "sample_hz": 1.0, "samples": 157, "errors": [], "per_gpu": { "": { "index": "0", "name": "NVIDIA GeForce RTX 3080", "power_w": { "min": 10.94, "median": 11.41, "max": 11.8, "mean": 11.4, "n": 157 }, "power_limit_w": 380.0, "memory_used_mib": { "min": 256.0, "median": 256.0, "max": 256.0, "mean": 256.0, "n": 157 }, "temperature_c": { "min": 30.0, "median": 31.0, "max": 34.0, "mean": 31.6, "n": 157 }, "clocks_sm_mhz": { "min": 210.0, "median": 210.0, "max": 210.0, "mean": 210.0, "n": 157 }, "n": 157 } }, "target_uuid": "", "target": { "index": "0", "name": "NVIDIA GeForce RTX 3080", "power_w": { "min": 10.94, "median": 11.41, "max": 11.8, "mean": 11.4, "n": 157 }, "power_limit_w": 380.0, "memory_used_mib": { "min": 256.0, "median": 256.0, "max": 256.0, "mean": 256.0, "n": 157 }, "temperature_c": { "min": 30.0, "median": 31.0, "max": 34.0, "mean": 31.6, "n": 157 }, "clocks_sm_mhz": { "min": 210.0, "median": 210.0, "max": 210.0, "mean": 210.0, "n": 157 }, "n": 157 } }, "smi_after": [ { "index": "0", "uuid": "", "name": "NVIDIA GeForce RTX 3080", "memory.used": 256.0, "power.draw": 11.4, "power.limit": 380.0, "temperature.gpu": 30.0, "clocks.sm": 210.0 } ] }, "scored": true, "peak_runner_rss_kb": 9502860, "peak_runner_rss_gib": 9.06, "peak_runner_threads": 32, "rss_samples": 314, "loadavg_during_runs": { "min": 8.95, "median": 9.73, "max": 9.99, "mean": 9.65, "samples": 314, "nproc": 20 }, "contention_suspected": false, "nvidia_after": "0, , NVIDIA GeForce RTX 3080, 256, 11.40, 380.00, 30, 210", "smi_after": [ { "index": "0", "uuid": "", "name": "NVIDIA GeForce RTX 3080", "memory.used": 256.0, "power.draw": 11.4, "power.limit": 380.0, "temperature.gpu": 30.0, "clocks.sm": 210.0 } ], "free_m_after": "total used free shared buff/cache available\nMem: 47040 15848 391 9 31334 31192\nSwap: 24575 971 23604", "derived": { "decode_tok_s": { "min": 4.845, "median": 4.847, "max": 4.848, "mean": 4.847, "n": 3 }, "decode_tok_s_wall": { "min": 4.826, "median": 4.828, "max": 4.828, "mean": 4.827, "n": 3 }, "ttft_ms": { "min": 1044.5, "median": 1046.12, "max": 1063.5, "mean": 1051.37, "n": 3 }, "wall_s": { "min": 53.8565, "median": 53.8635, "max": 53.9049, "mean": 53.875, "n": 3 }, "prefill_tok_s": { "min": 926.112, "median": 930.829, "max": 932.6, "mean": 929.847, "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": "2026-09-10T12:36:26Z" }