{
  "as_of": "2026-09-10",
  "sources": {
    "kimi_prices": "https://platform.kimi.ai/",
    "openai_prices": "https://developers.openai.com/api/docs/pricing",
    "gpt55_model": "https://developers.openai.com/api/docs/models/gpt-5.5",
    "sol_model": "https://developers.openai.com/api/docs/models/gpt-5.6-sol",
    "cache_write_rules": "https://developers.openai.com/api/docs/guides/prompt-caching",
    "astra_long_context": "https://developers.openai.com/api/docs/models/gpt-6-astra",
    "kimi_parameters": "https://huggingface.co/moonshotai/Kimi-K3",
    "deepseek_parameters": "https://huggingface.co/deepseek-ai/DeepSeek-V4-Pro",
    "nvl72_hardware": "https://www.nvidia.com/en-us/data-center/gb200-nvl72/",
    "hgx_hardware": "https://docs.nvidia.com/enterprise-reference-architectures/hgx-ai-factory/latest/components.html",
    "kimi_8k_benchmark": "https://zenn.dev/fixstars/articles/kimi-k3-benchmark"
  },
  "standard_prices_usd_per_million": {
    "kimi_k3": {
      "input": 3,
      "cache_read": 0.3,
      "cache_write": 3,
      "output": 15
    },
    "gpt55": {
      "input": 5,
      "cache_read": 0.5,
      "cache_write": null,
      "output": 30
    },
    "gpt56_sol": {
      "input": 4,
      "cache_read": 0.4,
      "cache_write": 5,
      "output": 20
    },
    "gpt6_astra": {
      "input": 10,
      "cache_read": 1,
      "cache_write": 12.5,
      "output": 50
    }
  },
  "openai_long_context_threshold_input_tokens": 272000,
  "assumptions": {
    "kappa_definition": "(Astra/Kimi service-resource cost shares) times (Kimi/Astra cost per active parameter under matched workload); local linear approximation",
    "kappa_is_measured": false,
    "hourly_costs_are_author_scenarios": true,
    "100k_workload_astra_throughput_is_measured": false,
    "same_numerical_token_counts_assumed": true,
    "effective_duty": 0.7,
    "target_resource_surplus_fraction": 0.3
  },
  "request_bills": [
    {
      "model": "kimi_k3",
      "input_tokens": 100000,
      "output_tokens": 1000,
      "cache_hit": 0.8,
      "miss_write_fraction": 0,
      "bill_usd": 0.09899999999999999,
      "revenue_usd_per_million_output": 98.99999999999999
    },
    {
      "model": "gpt55",
      "input_tokens": 100000,
      "output_tokens": 1000,
      "cache_hit": 0.8,
      "miss_write_fraction": 0,
      "bill_usd": 0.17,
      "revenue_usd_per_million_output": 170.0
    },
    {
      "model": "gpt56_sol",
      "input_tokens": 100000,
      "output_tokens": 1000,
      "cache_hit": 0.8,
      "miss_write_fraction": 0,
      "bill_usd": 0.132,
      "revenue_usd_per_million_output": 132.0
    },
    {
      "model": "gpt56_sol",
      "input_tokens": 100000,
      "output_tokens": 1000,
      "cache_hit": 0.8,
      "miss_write_fraction": 1,
      "bill_usd": 0.15199999999999997,
      "revenue_usd_per_million_output": 151.99999999999997
    },
    {
      "model": "gpt6_astra",
      "input_tokens": 100000,
      "output_tokens": 1000,
      "cache_hit": 0.8,
      "miss_write_fraction": 0,
      "bill_usd": 0.33,
      "revenue_usd_per_million_output": 330.0
    },
    {
      "model": "gpt6_astra",
      "input_tokens": 100000,
      "output_tokens": 1000,
      "cache_hit": 0.8,
      "miss_write_fraction": 1,
      "bill_usd": 0.38,
      "revenue_usd_per_million_output": 380.0
    },
    {
      "model": "kimi_k3",
      "input_tokens": 100000,
      "output_tokens": 1000,
      "cache_hit": 0.85,
      "miss_write_fraction": 0,
      "bill_usd": 0.0855,
      "revenue_usd_per_million_output": 85.5
    },
    {
      "model": "gpt55",
      "input_tokens": 100000,
      "output_tokens": 1000,
      "cache_hit": 0.85,
      "miss_write_fraction": 0,
      "bill_usd": 0.1475,
      "revenue_usd_per_million_output": 147.5
    },
    {
      "model": "gpt56_sol",
      "input_tokens": 100000,
      "output_tokens": 1000,
      "cache_hit": 0.85,
      "miss_write_fraction": 0,
      "bill_usd": 0.11400000000000002,
      "revenue_usd_per_million_output": 114.00000000000001
    },
    {
      "model": "gpt56_sol",
      "input_tokens": 100000,
      "output_tokens": 1000,
      "cache_hit": 0.85,
      "miss_write_fraction": 1,
      "bill_usd": 0.129,
      "revenue_usd_per_million_output": 129.0
    },
    {
      "model": "gpt6_astra",
      "input_tokens": 100000,
      "output_tokens": 1000,
      "cache_hit": 0.85,
      "miss_write_fraction": 0,
      "bill_usd": 0.285,
      "revenue_usd_per_million_output": 285.0
    },
    {
      "model": "gpt6_astra",
      "input_tokens": 100000,
      "output_tokens": 1000,
      "cache_hit": 0.85,
      "miss_write_fraction": 1,
      "bill_usd": 0.3225,
      "revenue_usd_per_million_output": 322.5
    },
    {
      "model": "kimi_k3",
      "input_tokens": 100000,
      "output_tokens": 1000,
      "cache_hit": 0.9,
      "miss_write_fraction": 0,
      "bill_usd": 0.072,
      "revenue_usd_per_million_output": 72.0
    },
    {
      "model": "gpt55",
      "input_tokens": 100000,
      "output_tokens": 1000,
      "cache_hit": 0.9,
      "miss_write_fraction": 0,
      "bill_usd": 0.125,
      "revenue_usd_per_million_output": 125.0
    },
    {
      "model": "gpt56_sol",
      "input_tokens": 100000,
      "output_tokens": 1000,
      "cache_hit": 0.9,
      "miss_write_fraction": 0,
      "bill_usd": 0.096,
      "revenue_usd_per_million_output": 96.0
    },
    {
      "model": "gpt56_sol",
      "input_tokens": 100000,
      "output_tokens": 1000,
      "cache_hit": 0.9,
      "miss_write_fraction": 1,
      "bill_usd": 0.10599999999999998,
      "revenue_usd_per_million_output": 105.99999999999999
    },
    {
      "model": "gpt6_astra",
      "input_tokens": 100000,
      "output_tokens": 1000,
      "cache_hit": 0.9,
      "miss_write_fraction": 0,
      "bill_usd": 0.24,
      "revenue_usd_per_million_output": 240.0
    },
    {
      "model": "gpt6_astra",
      "input_tokens": 100000,
      "output_tokens": 1000,
      "cache_hit": 0.9,
      "miss_write_fraction": 1,
      "bill_usd": 0.265,
      "revenue_usd_per_million_output": 265.0
    },
    {
      "model": "kimi_k3",
      "input_tokens": 300000,
      "output_tokens": 1000,
      "cache_hit": 0.8,
      "miss_write_fraction": 0,
      "bill_usd": 0.267,
      "revenue_usd_per_million_output": 267.0
    },
    {
      "model": "gpt55",
      "input_tokens": 300000,
      "output_tokens": 1000,
      "cache_hit": 0.8,
      "miss_write_fraction": 0,
      "bill_usd": 0.885,
      "revenue_usd_per_million_output": 885.0
    },
    {
      "model": "gpt56_sol",
      "input_tokens": 300000,
      "output_tokens": 1000,
      "cache_hit": 0.8,
      "miss_write_fraction": 0,
      "bill_usd": 0.7019999999999998,
      "revenue_usd_per_million_output": 701.9999999999999
    },
    {
      "model": "gpt56_sol",
      "input_tokens": 300000,
      "output_tokens": 1000,
      "cache_hit": 0.8,
      "miss_write_fraction": 1,
      "bill_usd": 0.8219999999999998,
      "revenue_usd_per_million_output": 821.9999999999999
    },
    {
      "model": "gpt6_astra",
      "input_tokens": 300000,
      "output_tokens": 1000,
      "cache_hit": 0.8,
      "miss_write_fraction": 0,
      "bill_usd": 1.755,
      "revenue_usd_per_million_output": 1755.0
    },
    {
      "model": "gpt6_astra",
      "input_tokens": 300000,
      "output_tokens": 1000,
      "cache_hit": 0.8,
      "miss_write_fraction": 1,
      "bill_usd": 2.055,
      "revenue_usd_per_million_output": 2055.0
    },
    {
      "model": "kimi_k3",
      "input_tokens": 300000,
      "output_tokens": 1000,
      "cache_hit": 0.85,
      "miss_write_fraction": 0,
      "bill_usd": 0.22650000000000003,
      "revenue_usd_per_million_output": 226.50000000000003
    },
    {
      "model": "gpt55",
      "input_tokens": 300000,
      "output_tokens": 1000,
      "cache_hit": 0.85,
      "miss_write_fraction": 0,
      "bill_usd": 0.75,
      "revenue_usd_per_million_output": 750.0
    },
    {
      "model": "gpt56_sol",
      "input_tokens": 300000,
      "output_tokens": 1000,
      "cache_hit": 0.85,
      "miss_write_fraction": 0,
      "bill_usd": 0.5940000000000001,
      "revenue_usd_per_million_output": 594.0000000000001
    },
    {
      "model": "gpt56_sol",
      "input_tokens": 300000,
      "output_tokens": 1000,
      "cache_hit": 0.85,
      "miss_write_fraction": 1,
      "bill_usd": 0.684,
      "revenue_usd_per_million_output": 684.0
    },
    {
      "model": "gpt6_astra",
      "input_tokens": 300000,
      "output_tokens": 1000,
      "cache_hit": 0.85,
      "miss_write_fraction": 0,
      "bill_usd": 1.485,
      "revenue_usd_per_million_output": 1485.0
    },
    {
      "model": "gpt6_astra",
      "input_tokens": 300000,
      "output_tokens": 1000,
      "cache_hit": 0.85,
      "miss_write_fraction": 1,
      "bill_usd": 1.71,
      "revenue_usd_per_million_output": 1710.0
    },
    {
      "model": "kimi_k3",
      "input_tokens": 300000,
      "output_tokens": 1000,
      "cache_hit": 0.9,
      "miss_write_fraction": 0,
      "bill_usd": 0.18599999999999997,
      "revenue_usd_per_million_output": 185.99999999999997
    },
    {
      "model": "gpt55",
      "input_tokens": 300000,
      "output_tokens": 1000,
      "cache_hit": 0.9,
      "miss_write_fraction": 0,
      "bill_usd": 0.615,
      "revenue_usd_per_million_output": 615.0
    },
    {
      "model": "gpt56_sol",
      "input_tokens": 300000,
      "output_tokens": 1000,
      "cache_hit": 0.9,
      "miss_write_fraction": 0,
      "bill_usd": 0.486,
      "revenue_usd_per_million_output": 486.0
    },
    {
      "model": "gpt56_sol",
      "input_tokens": 300000,
      "output_tokens": 1000,
      "cache_hit": 0.9,
      "miss_write_fraction": 1,
      "bill_usd": 0.546,
      "revenue_usd_per_million_output": 546.0
    },
    {
      "model": "gpt6_astra",
      "input_tokens": 300000,
      "output_tokens": 1000,
      "cache_hit": 0.9,
      "miss_write_fraction": 0,
      "bill_usd": 1.215,
      "revenue_usd_per_million_output": 1215.0
    },
    {
      "model": "gpt6_astra",
      "input_tokens": 300000,
      "output_tokens": 1000,
      "cache_hit": 0.9,
      "miss_write_fraction": 1,
      "bill_usd": 1.365,
      "revenue_usd_per_million_output": 1365.0
    }
  ],
  "throughput_thresholds": [
    {
      "model": "kimi_k3",
      "input_tokens": 100000,
      "output_tokens": 1000,
      "cache_hit": 0.8,
      "miss_write_fraction": 0,
      "bill_usd": 0.09899999999999999,
      "revenue_usd_per_million_output": 98.99999999999999,
      "assumed_hourly_cost_usd": 80,
      "assumed_duty": 0.7,
      "target_resource_surplus_fraction": 0,
      "required_active_output_tokens_per_second": 320.66698733365405
    },
    {
      "model": "kimi_k3",
      "input_tokens": 100000,
      "output_tokens": 1000,
      "cache_hit": 0.8,
      "miss_write_fraction": 0,
      "bill_usd": 0.09899999999999999,
      "revenue_usd_per_million_output": 98.99999999999999,
      "assumed_hourly_cost_usd": 80,
      "assumed_duty": 0.7,
      "target_resource_surplus_fraction": 0.3,
      "required_active_output_tokens_per_second": 458.09569619093435
    },
    {
      "model": "gpt6_astra",
      "input_tokens": 100000,
      "output_tokens": 1000,
      "cache_hit": 0.8,
      "miss_write_fraction": 0,
      "bill_usd": 0.33,
      "revenue_usd_per_million_output": 330.0,
      "assumed_hourly_cost_usd": 360,
      "assumed_duty": 0.7,
      "target_resource_surplus_fraction": 0,
      "required_active_output_tokens_per_second": 432.9004329004329
    },
    {
      "model": "gpt6_astra",
      "input_tokens": 100000,
      "output_tokens": 1000,
      "cache_hit": 0.8,
      "miss_write_fraction": 0,
      "bill_usd": 0.33,
      "revenue_usd_per_million_output": 330.0,
      "assumed_hourly_cost_usd": 360,
      "assumed_duty": 0.7,
      "target_resource_surplus_fraction": 0.3,
      "required_active_output_tokens_per_second": 618.4291898577612
    },
    {
      "model": "gpt6_astra",
      "input_tokens": 100000,
      "output_tokens": 1000,
      "cache_hit": 0.8,
      "miss_write_fraction": 0,
      "bill_usd": 0.33,
      "revenue_usd_per_million_output": 330.0,
      "assumed_hourly_cost_usd": 720,
      "assumed_duty": 0.7,
      "target_resource_surplus_fraction": 0,
      "required_active_output_tokens_per_second": 865.8008658008658
    },
    {
      "model": "gpt6_astra",
      "input_tokens": 100000,
      "output_tokens": 1000,
      "cache_hit": 0.8,
      "miss_write_fraction": 0,
      "bill_usd": 0.33,
      "revenue_usd_per_million_output": 330.0,
      "assumed_hourly_cost_usd": 720,
      "assumed_duty": 0.7,
      "target_resource_surplus_fraction": 0.3,
      "required_active_output_tokens_per_second": 1236.8583797155225
    },
    {
      "model": "gpt6_astra",
      "input_tokens": 100000,
      "output_tokens": 1000,
      "cache_hit": 0.8,
      "miss_write_fraction": 1,
      "bill_usd": 0.38,
      "revenue_usd_per_million_output": 380.0,
      "assumed_hourly_cost_usd": 360,
      "assumed_duty": 0.7,
      "target_resource_surplus_fraction": 0,
      "required_active_output_tokens_per_second": 375.9398496240602
    },
    {
      "model": "gpt6_astra",
      "input_tokens": 100000,
      "output_tokens": 1000,
      "cache_hit": 0.8,
      "miss_write_fraction": 1,
      "bill_usd": 0.38,
      "revenue_usd_per_million_output": 380.0,
      "assumed_hourly_cost_usd": 360,
      "assumed_duty": 0.7,
      "target_resource_surplus_fraction": 0.3,
      "required_active_output_tokens_per_second": 537.0569280343716
    },
    {
      "model": "gpt6_astra",
      "input_tokens": 100000,
      "output_tokens": 1000,
      "cache_hit": 0.8,
      "miss_write_fraction": 1,
      "bill_usd": 0.38,
      "revenue_usd_per_million_output": 380.0,
      "assumed_hourly_cost_usd": 720,
      "assumed_duty": 0.7,
      "target_resource_surplus_fraction": 0,
      "required_active_output_tokens_per_second": 751.8796992481203
    },
    {
      "model": "gpt6_astra",
      "input_tokens": 100000,
      "output_tokens": 1000,
      "cache_hit": 0.8,
      "miss_write_fraction": 1,
      "bill_usd": 0.38,
      "revenue_usd_per_million_output": 380.0,
      "assumed_hourly_cost_usd": 720,
      "assumed_duty": 0.7,
      "target_resource_surplus_fraction": 0.3,
      "required_active_output_tokens_per_second": 1074.1138560687432
    },
    {
      "model": "kimi_k3",
      "input_tokens": 100000,
      "output_tokens": 1000,
      "cache_hit": 0.85,
      "miss_write_fraction": 0,
      "bill_usd": 0.0855,
      "revenue_usd_per_million_output": 85.5,
      "assumed_hourly_cost_usd": 80,
      "assumed_duty": 0.7,
      "target_resource_surplus_fraction": 0,
      "required_active_output_tokens_per_second": 371.298616912652
    },
    {
      "model": "kimi_k3",
      "input_tokens": 100000,
      "output_tokens": 1000,
      "cache_hit": 0.85,
      "miss_write_fraction": 0,
      "bill_usd": 0.0855,
      "revenue_usd_per_million_output": 85.5,
      "assumed_hourly_cost_usd": 80,
      "assumed_duty": 0.7,
      "target_resource_surplus_fraction": 0.3,
      "required_active_output_tokens_per_second": 530.4265955895029
    },
    {
      "model": "gpt6_astra",
      "input_tokens": 100000,
      "output_tokens": 1000,
      "cache_hit": 0.85,
      "miss_write_fraction": 0,
      "bill_usd": 0.285,
      "revenue_usd_per_million_output": 285.0,
      "assumed_hourly_cost_usd": 360,
      "assumed_duty": 0.7,
      "target_resource_surplus_fraction": 0,
      "required_active_output_tokens_per_second": 501.2531328320802
    },
    {
      "model": "gpt6_astra",
      "input_tokens": 100000,
      "output_tokens": 1000,
      "cache_hit": 0.85,
      "miss_write_fraction": 0,
      "bill_usd": 0.285,
      "revenue_usd_per_million_output": 285.0,
      "assumed_hourly_cost_usd": 360,
      "assumed_duty": 0.7,
      "target_resource_surplus_fraction": 0.3,
      "required_active_output_tokens_per_second": 716.0759040458289
    },
    {
      "model": "gpt6_astra",
      "input_tokens": 100000,
      "output_tokens": 1000,
      "cache_hit": 0.85,
      "miss_write_fraction": 0,
      "bill_usd": 0.285,
      "revenue_usd_per_million_output": 285.0,
      "assumed_hourly_cost_usd": 720,
      "assumed_duty": 0.7,
      "target_resource_surplus_fraction": 0,
      "required_active_output_tokens_per_second": 1002.5062656641604
    },
    {
      "model": "gpt6_astra",
      "input_tokens": 100000,
      "output_tokens": 1000,
      "cache_hit": 0.85,
      "miss_write_fraction": 0,
      "bill_usd": 0.285,
      "revenue_usd_per_million_output": 285.0,
      "assumed_hourly_cost_usd": 720,
      "assumed_duty": 0.7,
      "target_resource_surplus_fraction": 0.3,
      "required_active_output_tokens_per_second": 1432.1518080916578
    },
    {
      "model": "gpt6_astra",
      "input_tokens": 100000,
      "output_tokens": 1000,
      "cache_hit": 0.85,
      "miss_write_fraction": 1,
      "bill_usd": 0.3225,
      "revenue_usd_per_million_output": 322.5,
      "assumed_hourly_cost_usd": 360,
      "assumed_duty": 0.7,
      "target_resource_surplus_fraction": 0,
      "required_active_output_tokens_per_second": 442.96788482834995
    },
    {
      "model": "gpt6_astra",
      "input_tokens": 100000,
      "output_tokens": 1000,
      "cache_hit": 0.85,
      "miss_write_fraction": 1,
      "bill_usd": 0.3225,
      "revenue_usd_per_million_output": 322.5,
      "assumed_hourly_cost_usd": 360,
      "assumed_duty": 0.7,
      "target_resource_surplus_fraction": 0.3,
      "required_active_output_tokens_per_second": 632.8112640405
    },
    {
      "model": "gpt6_astra",
      "input_tokens": 100000,
      "output_tokens": 1000,
      "cache_hit": 0.85,
      "miss_write_fraction": 1,
      "bill_usd": 0.3225,
      "revenue_usd_per_million_output": 322.5,
      "assumed_hourly_cost_usd": 720,
      "assumed_duty": 0.7,
      "target_resource_surplus_fraction": 0,
      "required_active_output_tokens_per_second": 885.9357696566999
    },
    {
      "model": "gpt6_astra",
      "input_tokens": 100000,
      "output_tokens": 1000,
      "cache_hit": 0.85,
      "miss_write_fraction": 1,
      "bill_usd": 0.3225,
      "revenue_usd_per_million_output": 322.5,
      "assumed_hourly_cost_usd": 720,
      "assumed_duty": 0.7,
      "target_resource_surplus_fraction": 0.3,
      "required_active_output_tokens_per_second": 1265.622528081
    },
    {
      "model": "kimi_k3",
      "input_tokens": 100000,
      "output_tokens": 1000,
      "cache_hit": 0.9,
      "miss_write_fraction": 0,
      "bill_usd": 0.072,
      "revenue_usd_per_million_output": 72.0,
      "assumed_hourly_cost_usd": 80,
      "assumed_duty": 0.7,
      "target_resource_surplus_fraction": 0,
      "required_active_output_tokens_per_second": 440.9171075837742
    },
    {
      "model": "kimi_k3",
      "input_tokens": 100000,
      "output_tokens": 1000,
      "cache_hit": 0.9,
      "miss_write_fraction": 0,
      "bill_usd": 0.072,
      "revenue_usd_per_million_output": 72.0,
      "assumed_hourly_cost_usd": 80,
      "assumed_duty": 0.7,
      "target_resource_surplus_fraction": 0.3,
      "required_active_output_tokens_per_second": 629.8815822625346
    },
    {
      "model": "gpt6_astra",
      "input_tokens": 100000,
      "output_tokens": 1000,
      "cache_hit": 0.9,
      "miss_write_fraction": 0,
      "bill_usd": 0.24,
      "revenue_usd_per_million_output": 240.0,
      "assumed_hourly_cost_usd": 360,
      "assumed_duty": 0.7,
      "target_resource_surplus_fraction": 0,
      "required_active_output_tokens_per_second": 595.2380952380952
    },
    {
      "model": "gpt6_astra",
      "input_tokens": 100000,
      "output_tokens": 1000,
      "cache_hit": 0.9,
      "miss_write_fraction": 0,
      "bill_usd": 0.24,
      "revenue_usd_per_million_output": 240.0,
      "assumed_hourly_cost_usd": 360,
      "assumed_duty": 0.7,
      "target_resource_surplus_fraction": 0.3,
      "required_active_output_tokens_per_second": 850.3401360544218
    },
    {
      "model": "gpt6_astra",
      "input_tokens": 100000,
      "output_tokens": 1000,
      "cache_hit": 0.9,
      "miss_write_fraction": 0,
      "bill_usd": 0.24,
      "revenue_usd_per_million_output": 240.0,
      "assumed_hourly_cost_usd": 720,
      "assumed_duty": 0.7,
      "target_resource_surplus_fraction": 0,
      "required_active_output_tokens_per_second": 1190.4761904761904
    },
    {
      "model": "gpt6_astra",
      "input_tokens": 100000,
      "output_tokens": 1000,
      "cache_hit": 0.9,
      "miss_write_fraction": 0,
      "bill_usd": 0.24,
      "revenue_usd_per_million_output": 240.0,
      "assumed_hourly_cost_usd": 720,
      "assumed_duty": 0.7,
      "target_resource_surplus_fraction": 0.3,
      "required_active_output_tokens_per_second": 1700.6802721088436
    },
    {
      "model": "gpt6_astra",
      "input_tokens": 100000,
      "output_tokens": 1000,
      "cache_hit": 0.9,
      "miss_write_fraction": 1,
      "bill_usd": 0.265,
      "revenue_usd_per_million_output": 265.0,
      "assumed_hourly_cost_usd": 360,
      "assumed_duty": 0.7,
      "target_resource_surplus_fraction": 0,
      "required_active_output_tokens_per_second": 539.0835579514825
    },
    {
      "model": "gpt6_astra",
      "input_tokens": 100000,
      "output_tokens": 1000,
      "cache_hit": 0.9,
      "miss_write_fraction": 1,
      "bill_usd": 0.265,
      "revenue_usd_per_million_output": 265.0,
      "assumed_hourly_cost_usd": 360,
      "assumed_duty": 0.7,
      "target_resource_surplus_fraction": 0.3,
      "required_active_output_tokens_per_second": 770.1193685021178
    },
    {
      "model": "gpt6_astra",
      "input_tokens": 100000,
      "output_tokens": 1000,
      "cache_hit": 0.9,
      "miss_write_fraction": 1,
      "bill_usd": 0.265,
      "revenue_usd_per_million_output": 265.0,
      "assumed_hourly_cost_usd": 720,
      "assumed_duty": 0.7,
      "target_resource_surplus_fraction": 0,
      "required_active_output_tokens_per_second": 1078.167115902965
    },
    {
      "model": "gpt6_astra",
      "input_tokens": 100000,
      "output_tokens": 1000,
      "cache_hit": 0.9,
      "miss_write_fraction": 1,
      "bill_usd": 0.265,
      "revenue_usd_per_million_output": 265.0,
      "assumed_hourly_cost_usd": 720,
      "assumed_duty": 0.7,
      "target_resource_surplus_fraction": 0.3,
      "required_active_output_tokens_per_second": 1540.2387370042356
    }
  ],
  "active_parameter_sensitivity": [
    {
      "astra_miss_write_fraction": 0,
      "cache_hit": 0.85,
      "input_tokens": 100000,
      "output_tokens": 1000,
      "reference_active_parameters_b": 104,
      "revenue_ratio": 3.3333333333333326,
      "assumed_kappa": 0.75,
      "implied_astra_active_parameters_b": 259.99999999999994
    },
    {
      "astra_miss_write_fraction": 0,
      "cache_hit": 0.85,
      "input_tokens": 100000,
      "output_tokens": 1000,
      "reference_active_parameters_b": 104,
      "revenue_ratio": 3.3333333333333326,
      "assumed_kappa": 1,
      "implied_astra_active_parameters_b": 346.6666666666666
    },
    {
      "astra_miss_write_fraction": 0,
      "cache_hit": 0.85,
      "input_tokens": 100000,
      "output_tokens": 1000,
      "reference_active_parameters_b": 104,
      "revenue_ratio": 3.3333333333333326,
      "assumed_kappa": 1.25,
      "implied_astra_active_parameters_b": 433.3333333333332
    },
    {
      "astra_miss_write_fraction": 1,
      "cache_hit": 0.85,
      "input_tokens": 100000,
      "output_tokens": 1000,
      "reference_active_parameters_b": 104,
      "revenue_ratio": 3.7719298245614032,
      "assumed_kappa": 0.75,
      "implied_astra_active_parameters_b": 294.2105263157894
    },
    {
      "astra_miss_write_fraction": 1,
      "cache_hit": 0.85,
      "input_tokens": 100000,
      "output_tokens": 1000,
      "reference_active_parameters_b": 104,
      "revenue_ratio": 3.7719298245614032,
      "assumed_kappa": 1,
      "implied_astra_active_parameters_b": 392.2807017543859
    },
    {
      "astra_miss_write_fraction": 1,
      "cache_hit": 0.85,
      "input_tokens": 100000,
      "output_tokens": 1000,
      "reference_active_parameters_b": 104,
      "revenue_ratio": 3.7719298245614032,
      "assumed_kappa": 1.25,
      "implied_astra_active_parameters_b": 490.3508771929824
    }
  ],
  "model_scenarios": [
    {
      "active_parameters_b": 300,
      "total_active_ratio": 26.923076923076923,
      "total_parameters_t": 8.076923076923077,
      "raw_4bit_weight_tb": 4.038461538461538,
      "nvl72_hbm_after_raw_4bit_weights_tb": 9.361538461538462,
      "project_flops": 1.0368e+27,
      "pretraining_share": 0.6,
      "cumulative_training_tokens_t": 345.59999999999997
    },
    {
      "active_parameters_b": 300,
      "total_active_ratio": 30,
      "total_parameters_t": 9.0,
      "raw_4bit_weight_tb": 4.5,
      "nvl72_hbm_after_raw_4bit_weights_tb": 8.9,
      "project_flops": 1.0368e+27,
      "pretraining_share": 0.6,
      "cumulative_training_tokens_t": 345.59999999999997
    },
    {
      "active_parameters_b": 300,
      "total_active_ratio": 32.6530612244898,
      "total_parameters_t": 9.795918367346939,
      "raw_4bit_weight_tb": 4.8979591836734695,
      "nvl72_hbm_after_raw_4bit_weights_tb": 8.50204081632653,
      "project_flops": 1.0368e+27,
      "pretraining_share": 0.6,
      "cumulative_training_tokens_t": 345.59999999999997
    },
    {
      "active_parameters_b": 350,
      "total_active_ratio": 26.923076923076923,
      "total_parameters_t": 9.423076923076923,
      "raw_4bit_weight_tb": 4.711538461538462,
      "nvl72_hbm_after_raw_4bit_weights_tb": 8.688461538461539,
      "project_flops": 1.0368e+27,
      "pretraining_share": 0.6,
      "cumulative_training_tokens_t": 296.2285714285714
    },
    {
      "active_parameters_b": 350,
      "total_active_ratio": 30,
      "total_parameters_t": 10.5,
      "raw_4bit_weight_tb": 5.25,
      "nvl72_hbm_after_raw_4bit_weights_tb": 8.15,
      "project_flops": 1.0368e+27,
      "pretraining_share": 0.6,
      "cumulative_training_tokens_t": 296.2285714285714
    },
    {
      "active_parameters_b": 350,
      "total_active_ratio": 32.6530612244898,
      "total_parameters_t": 11.428571428571429,
      "raw_4bit_weight_tb": 5.714285714285714,
      "nvl72_hbm_after_raw_4bit_weights_tb": 7.685714285714286,
      "project_flops": 1.0368e+27,
      "pretraining_share": 0.6,
      "cumulative_training_tokens_t": 296.2285714285714
    },
    {
      "active_parameters_b": 400,
      "total_active_ratio": 26.923076923076923,
      "total_parameters_t": 10.76923076923077,
      "raw_4bit_weight_tb": 5.384615384615385,
      "nvl72_hbm_after_raw_4bit_weights_tb": 8.015384615384615,
      "project_flops": 1.0368e+27,
      "pretraining_share": 0.6,
      "cumulative_training_tokens_t": 259.2
    },
    {
      "active_parameters_b": 400,
      "total_active_ratio": 30,
      "total_parameters_t": 12.0,
      "raw_4bit_weight_tb": 6.0,
      "nvl72_hbm_after_raw_4bit_weights_tb": 7.4,
      "project_flops": 1.0368e+27,
      "pretraining_share": 0.6,
      "cumulative_training_tokens_t": 259.2
    },
    {
      "active_parameters_b": 400,
      "total_active_ratio": 32.6530612244898,
      "total_parameters_t": 13.061224489795919,
      "raw_4bit_weight_tb": 6.530612244897959,
      "nvl72_hbm_after_raw_4bit_weights_tb": 6.869387755102041,
      "project_flops": 1.0368e+27,
      "pretraining_share": 0.6,
      "cumulative_training_tokens_t": 259.2
    },
    {
      "active_parameters_b": 350,
      "total_active_ratio": 28.21224489795918,
      "total_parameters_t": 9.874285714285714,
      "raw_4bit_weight_tb": 4.937142857142857,
      "nvl72_hbm_after_raw_4bit_weights_tb": 8.462857142857143,
      "project_flops": 1.0368e+27,
      "pretraining_share": 0.6,
      "cumulative_training_tokens_t": 296.2285714285714
    }
  ],
  "generation_price_calibration": [
    {
      "model": "kimi_k3",
      "input_tokens": 100000,
      "output_tokens": 1000,
      "cache_hit": 0.85,
      "miss_write_fraction": 0,
      "bill_usd": 0.0855,
      "revenue_ratio_vs_kimi": 1.0,
      "assumed_kappa": 1,
      "price_normalized_active_parameters_b": 104.0,
      "parameter_status": "reported"
    },
    {
      "model": "gpt55",
      "input_tokens": 100000,
      "output_tokens": 1000,
      "cache_hit": 0.85,
      "miss_write_fraction": 0,
      "bill_usd": 0.1475,
      "revenue_ratio_vs_kimi": 1.725146198830409,
      "assumed_kappa": 1,
      "price_normalized_active_parameters_b": 179.41520467836256,
      "parameter_status": "conditional_price_normalization"
    },
    {
      "model": "gpt56_sol",
      "input_tokens": 100000,
      "output_tokens": 1000,
      "cache_hit": 0.85,
      "miss_write_fraction": 0,
      "bill_usd": 0.11400000000000002,
      "revenue_ratio_vs_kimi": 1.3333333333333335,
      "assumed_kappa": 1,
      "price_normalized_active_parameters_b": 138.66666666666669,
      "parameter_status": "conditional_price_normalization"
    },
    {
      "model": "gpt6_astra",
      "input_tokens": 100000,
      "output_tokens": 1000,
      "cache_hit": 0.85,
      "miss_write_fraction": 0,
      "bill_usd": 0.285,
      "revenue_ratio_vs_kimi": 3.3333333333333326,
      "assumed_kappa": 1,
      "price_normalized_active_parameters_b": 346.6666666666666,
      "parameter_status": "conditional_price_normalization"
    }
  ],
  "agent_vs_8_to_1_at_90pct_cache": {
    "old_revenue_per_million_output": 19.560000000000002,
    "new_revenue_per_million_output": 72.0,
    "revenue_ratio": 3.680981595092024,
    "old_fresh_forward_tokens_per_output": 1.8,
    "new_fresh_forward_tokens_per_output": 11.0,
    "fresh_forward_work_ratio": 6.111111111111111
  },
  "central_model_scenario": {
    "active_parameters_b": 350,
    "total_active_ratio": 28.21224489795918,
    "total_parameters_t": 9.874285714285714,
    "raw_4bit_weight_tb": 4.937142857142857,
    "nvl72_hbm_after_raw_4bit_weights_tb": 8.462857142857143,
    "project_flops": 1.0368e+27,
    "pretraining_share": 0.6,
    "cumulative_training_tokens_t": 296.2285714285714
  },
  "preferred_author_scenario": {
    "active_parameters_b": [
      300,
      400
    ],
    "total_active_ratio": [
      26.923076923076923,
      32.6530612244898
    ],
    "architecture_only_total_parameters_t": [
      8.076923076923077,
      13.061224489795919
    ],
    "joint_training_constraint_file": "training-constraints.json"
  },
  "memory_examples": [
    {
      "total_parameters_t": 8,
      "raw_4bit_weight_tb": 4.0,
      "nvl72_hbm_after_raw_4bit_weights_tb": 9.4
    },
    {
      "total_parameters_t": 10,
      "raw_4bit_weight_tb": 5.0,
      "nvl72_hbm_after_raw_4bit_weights_tb": 8.4
    },
    {
      "total_parameters_t": 12,
      "raw_4bit_weight_tb": 6.0,
      "nvl72_hbm_after_raw_4bit_weights_tb": 7.4
    }
  ]
}