{
 "date": "2026-11-06",
 "model": "Kimi K3.1 Pro: 2.8T MoE, 92 MoE layers, 896 experts per layer, 16 routed per token; Q8_0 spine, MXFP4 experts (82,432 files, 1.45 TB)",
 "engine": "argonautlabsai/argodrive-ds4-kimi-k3.1, tag k31-cuda-20261105 (port on top of antirez/ds4 c97964b)",
 "hardware": "one RTX PRO 6000 Blackwell (96 GB) in a 70 GiB VM; RAID0 of three NVMe (19 GB/s) with 79,365 experts; second NVMe (6.2 GB/s) with 41,078; Thunderbolt NVMe (3.7 GB/s) with 74,000",
 "settings": "defaults plus a third expert directory: full-precision attention cache, no drafter, no expert skipping, -c 2048, greedy, monitoring off",
 "prompts": {
  "france": "The capital of France is",
  "record": "The three main financial statements are"
 },
 "repeats": "token ids (fresh process) or text (chat endpoint) identical across the repeats of each arm",
 "not_established": [
  "comparison with Moonshot's own implementation",
  "identity across arithmetic paths: see the README's table"
 ],
 "runs": {
  "fresh_process": {
   "france_17_sequential": [
    {
     "generated": 17,
     "decode_seconds": 18.46,
     "steady_seconds": 7.74,
     "steady_tok_s": 2.07,
     "wall_seconds": 29.329
    },
    {
     "generated": 17,
     "decode_seconds": 18.45,
     "steady_seconds": 7.71,
     "steady_tok_s": 2.08,
     "wall_seconds": 29.231
    },
    {
     "generated": 17,
     "decode_seconds": 18.37,
     "steady_seconds": 7.71,
     "steady_tok_s": 2.08,
     "wall_seconds": 29.131
    }
   ],
   "france_17_batched": [
    {
     "generated": 17,
     "decode_seconds": 8.08,
     "steady_seconds": 8.07,
     "steady_tok_s": 1.98,
     "prompt_tokens": 5,
     "prompt_pass_seconds": 17.23,
     "prompt_pass_expert_gb": 100.07,
     "wall_seconds": 36.135
    },
    {
     "generated": 17,
     "decode_seconds": 8.13,
     "steady_seconds": 8.13,
     "steady_tok_s": 1.97,
     "prompt_tokens": 5,
     "prompt_pass_seconds": 17.22,
     "prompt_pass_expert_gb": 100.07,
     "wall_seconds": 36.032
    },
    {
     "generated": 17,
     "decode_seconds": 8.17,
     "steady_seconds": 8.17,
     "steady_tok_s": 1.96,
     "prompt_tokens": 5,
     "prompt_pass_seconds": 17.25,
     "prompt_pass_expert_gb": 100.07,
     "wall_seconds": 36.034
    }
   ],
   "record_128": [
    {
     "generated": 128,
     "decode_seconds": 62.23,
     "steady_seconds": 62.23,
     "steady_tok_s": 2.04,
     "prompt_tokens": 6,
     "prompt_pass_seconds": 17.25,
     "prompt_pass_expert_gb": 105.28,
     "wall_seconds": 90.275
    },
    {
     "generated": 128,
     "decode_seconds": 62.27,
     "steady_seconds": 62.27,
     "steady_tok_s": 2.04,
     "prompt_tokens": 6,
     "prompt_pass_seconds": 17.23,
     "prompt_pass_expert_gb": 105.28,
     "wall_seconds": 90.274
    }
   ],
   "record_512": [
    {
     "generated": 512,
     "decode_seconds": 263.22,
     "steady_seconds": 263.22,
     "steady_tok_s": 1.94,
     "prompt_tokens": 6,
     "prompt_pass_seconds": 17.34,
     "prompt_pass_expert_gb": 105.28,
     "wall_seconds": 291.319
    },
    {
     "generated": 512,
     "decode_seconds": 263.12,
     "steady_seconds": 263.12,
     "steady_tok_s": 1.94,
     "prompt_tokens": 6,
     "prompt_pass_seconds": 17.35,
     "prompt_pass_expert_gb": 105.28,
     "wall_seconds": 291.323
    }
   ]
  },
  "chat_endpoint": {
   "prompt_270_generated_200": [
    {
     "prompt_tokens": 270,
     "prompt_pass_seconds": 25.38,
     "prompt_pass_expert_gb": 924.53,
     "generated": 200,
     "decode_seconds": 101.99,
     "decode_tok_s": 1.96,
     "steady_tok_s": 1.95
    },
    {
     "prompt_tokens": 270,
     "prompt_pass_seconds": 25.34,
     "prompt_pass_expert_gb": 924.53,
     "generated": 200,
     "decode_seconds": 103.61,
     "decode_tok_s": 1.93,
     "steady_tok_s": 1.92
    }
   ],
   "prompt_618_generated_200": [
    {
     "prompt_tokens": 618,
     "prompt_pass_seconds": 31.71,
     "prompt_pass_expert_gb": 1122.31,
     "generated": 200,
     "decode_seconds": 105.15,
     "decode_tok_s": 1.9,
     "steady_tok_s": 1.89
    },
    {
     "prompt_tokens": 618,
     "prompt_pass_seconds": 31.51,
     "prompt_pass_expert_gb": 1122.31,
     "generated": 200,
     "decode_seconds": 105.11,
     "decode_tok_s": 1.9,
     "steady_tok_s": 1.89
    }
   ],
   "prompt_1533_generated_8": [
    {
     "prompt_tokens": 1533,
     "prompt_pass_seconds": 36.81,
     "prompt_pass_expert_gb": 1240.82,
     "generated": 8,
     "decode_seconds": 3.85,
     "decode_tok_s": 2.08,
     "steady_tok_s": 1.82
    },
    {
     "prompt_tokens": 1533,
     "prompt_pass_seconds": 37.12,
     "prompt_pass_expert_gb": 1240.82,
     "generated": 8,
     "decode_seconds": 3.84,
     "decode_tok_s": 2.08,
     "steady_tok_s": 1.82
    }
   ]
  }
 }
}
