{
  "updated": "2026-10-16",
  "currency": "USD",
  "model": "GLM-5.3",
  "workload": {
    "date": "2026-10-14",
    "start_utc": "02:03",
    "end_utc": "07:30",
    "elapsed_seconds": 19620,
    "configured_agents": 2048,
    "ready_environments_at_start": 2048,
    "completed_model_responses_estimate": 201852,
    "input_tokens_estimate": 3284084176,
    "cached_input_tokens_estimate": 3073472181,
    "output_tokens_estimate": 154863956,
    "observations": 1309,
    "outcome": "Run ended in a crash at approximately 07:30:55 UTC after degraded service.",
    "method": "Rounded Prometheus increase estimates over a fixed 19,620-second window. Includes cache rebuilding and degraded service. The selected prompt counters had no resets.",
    "limits": [
      "Configured agents and ready environments are not simultaneous inference requests.",
      "No task-quality improvement or advantage over another harness is established.",
      "These are aggregate internal measurements, not a public per-request trace.",
      "The recovery run is excluded. A full-day extrapolation is not an observed 24-hour run."
    ]
  },
  "providers": [
    {
      "id": "prime",
      "name": "Prime Intellect via OpenRouter",
      "input_per_million": 1.4,
      "cached_input_per_million": 0.26,
      "output_per_million": 4.4
    },
    {
      "id": "deepinfra",
      "name": "DeepInfra via OpenRouter (promotional)",
      "input_per_million": 0.5625,
      "cached_input_per_million": 0.125,
      "output_per_million": 2.5
    },
    {
      "id": "baseten",
      "name": "Baseten FP4 via OpenRouter",
      "input_per_million": 1.4,
      "cached_input_per_million": 0.14,
      "output_per_million": 4.4
    }
  ],
  "environments": {
    "provider": "Daytona",
    "vcpu_hour": 0.0504,
    "gib_ram_hour": 0.0162,
    "gib_storage_hour": 0.000108,
    "free_storage_gib": 5,
    "disk_gib": 16,
    "sizes": [
      {
        "id": "small",
        "vcpu": 1,
        "ram_gib": 2
      },
      {
        "id": "medium",
        "vcpu": 2,
        "ram_gib": 4
      }
    ],
    "scope": "Rented Daytona environments only. These sizes do not specify the Multiplet standard package."
  },
  "method": "Scale recorded token counts by requested agent-hours / (2048 * 19620 / 3600). Price input at the chosen cache rate and output at the selected provider rate. Charge each Daytona environment for the full duration.",
  "limits": [
    "Scenario estimate, not a remote benchmark or provider quote.",
    "Recorded workload included degraded service and ended in a crash.",
    "Configured agent-hours do not measure simultaneous inference or useful work.",
    "Same model label does not establish equal quantization, speed, quality or capacity.",
    "Provider availability, quota and achievable cache reuse require a real test.",
    "Excludes router fees, engineering, orchestration, operations, network, plan charges, discounts and taxes.",
    "Does not model paid cache retention or cache-write premiums."
  ],
  "sources": [
    {
      "label": "GLM-5.3 provider prices",
      "url": "https://openrouter.ai/z-ai/glm-5.3"
    },
    {
      "label": "Endpoint pricing API",
      "url": "https://openrouter.ai/api/v1/models/z-ai/glm-5.3/endpoints"
    },
    {
      "label": "Daytona pricing",
      "url": "https://www.daytona.io/pricing"
    },
    {
      "label": "DeepInfra automatic caching",
      "url": "https://docs.deepinfra.com/chat/prompt-caching"
    },
    {
      "label": "OpenRouter sticky routing",
      "url": "https://openrouter.ai/docs/guides/best-practices/prompt-caching"
    }
  ]
}