Buildkite

• $.reliability_note none-observed

{
 "wave2": {
  "class": "ci-runner-capacity",
  "free_tier_kind": "monthly-vcpu-minute",
  "shape": {
   "vcpu": null,
   "ram_gib": null,
   "gpu_class": null,
   "gpu_count": 0
  }
 },
 "shape": {
  "vcpu": 2,
  "ram_gib": 4,
  "disk_gib": 47,
  "gpu_class": null,
  "gpu_count": null,
  "note": "`LINUX_AMD64_2X4`, the only shape the Free plan offers, quoted as 'Small | AMD64 | 2 | 4 GB | 47 GB'. The vendor writes decimal GB; kept unconverted, so 4 GB is about 3.7 GiB and 47 GB about 43.8 GiB. Larger shapes exist (4/16/95, 8/32/158 and up to 64 vCPU) but are paid-plan or Enterprise only."
 },
 "billing_model": "free",
 "term": "none",
 "m10h_usd": null,
 "m24h_usd": null,
 "price_arithmetic": "2,000 vCPU-minutes / 2 vCPU = 1,000 wall-clock minutes = 16 h 40 min a month at USD 0.00. 10 h a day x 30 = 300 h = 18,000 wall-clock minutes = 36,000 vCPU-minutes, eighteen times the pool. 24/7 x 30 = 730 h needs 87,600 vCPU-minutes, forty-four times the pool. Past the pool the rate is USD 0.004 per vCPU-minute on a paid platform plan, so 36,000 vCPU-minutes would be USD 144.00 and 87,600 USD 350.40.",
 "free_tier_kind": "monthly-vcpu-minute",
 "session_cap": {
  "wall_clock_min": 240,
  "idle_reap_s": null,
  "restart_required": true,
  "note": "4 hours per job on the Free plan; the 2,000 vCPU-minute monthly pool runs out far sooner. The environment is destroyed at job end."
 },
 "concurrency": {
  "jobs_at_once": 10,
  "note": "20 vCPU of Linux concurrency on the Free plan, and the only shape is 2 vCPU, so 10 machines at once. Quoted: 'the Free plan's 20-vCPU Linux limit supports up to 10 Small Linux machines at 2 vCPU each'."
 },
 "keep_rate": {
  "value": 0.0,
  "basis": "the environment is destroyed at job end; cache volumes persist at no stated charge",
  "source": "https://buildkite.com/docs/pipelines/hosted-agents"
 },
 "persistent_storage": {
  "included_gib": 0,
  "price_gib_month": "not published; cache volumes are 'included at no additional cost' with a 50 GB container cache and 5 GB git mirror default quota",
  "durability": "the job environment is destroyed; cache volumes persist across jobs",
  "survives_stop": false,
  "survives_destroy": false,
  "max_retention": "the job, except cache volumes"
 },
 "network_out": {
  "allow_outbound": true,
  "allow_inbound_listen": false,
  "ports": "none inbound: no page read offers a public address and the environment is destroyed at job end",
  "static_ip": false,
  "ipv6": null,
  "per_session_bytes": null
 },
 "units": "RAM and disk are the vendor's decimal GB, kept unconverted. 4 GB reads as 4, 47 GB as 47.",
 "granularity": {
  "min_billed_s": 1,
  "rounding_s": 1,
  "note": "quoted for the paid rate: 'billed per vCPU minute and metered to the second, with no minimums and no rounding'"
 },
 "burst_x": 10.0,
 "burst_note": "10 concurrent 2 vCPU small machines on the Free plan's 20 vCPU of Linux concurrency. burst_x is that concurrency multiplier over a single runner.",
 "census_rows": [
  "buildkite"
 ],
 "topic_note": "belongs to ci-cd-runners; recorded here to stop the double count",
 "wave3": {
  "verified": [
   "2,000 Linux vCPU-minutes a month on the Free plan",
   "20 vCPU Linux concurrency = 10 small machines",
   "the small shape is 2 vCPU / 4 GB / 47 GB",
   "4-hour job timeout on the Free plan",
   "no credit card required"
  ],
  "corrected": "It does and they are not. The earlier arithmetic also priced the pool against a 4 vCPU agent that a Free namespace cannot select.",
  "new": "the shape, the concurrency and the 4-hour Free-plan job timeout, none of which the corpus had"
 }
}

• $.egress.metered egress-unknown
• $.interface interface-unknown
• $.per_object_cap.notes process
• $.per_object_cap.stated not-applicable
• $.reliability_note none-observed
• $.tooling not-stated

{
 "talaria_m10h_usd": 174.0,
 "talaria_m24h_usd": 375.6,
 "nemo_m10h_usd": 136.0,
 "keep_rate": 1.0,
 "keep_rate_basis": null,
 "billing_model": "unclassified",
 "shape": null,
 "session_cap": {
  "wall_clock_min": null,
  "idle_reap_s": null,
  "restart_required": false,
  "note": "a job has its own timeout; the monthly vCPU-minute pool is what runs out"
 },
 "persistent_storage": null,
 "network_out": null,
 "price_note": null,
 "free_tier_kind": "allowance",
 "granularity": {
  "min_billed_s": 0,
  "rounding_s": null
 },
 "burst_x": 1.0,
 "ps": null,
 "net": null
}

• $.chaining not-stated
• $.egress.metered egress-unknown
• $.egress.terms not-stated
• $.evidence[0].proved process
• $.extras.keep_rate_basis process
• $.extras.price_note process
• $.interface interface-unknown
• $.interface_detail withheld
• $.limits_note not-stated
• $.per_object_cap.notes process
• $.per_object_cap.stated not-stated
• $.reliability_note none-observed
• $.retention.notes not-extracted
• $.retention.stated not-stated
• $.tooling not-stated

{
 "dead_end": {
  "failure_class": "retention",
  "why": "Metered both ways: storage billed per GB-month and transfer billed per GB beyond plan inclusion, and retention maxes at 7/30/90 days depending on plan. Full custom retention is an Enterprise",
  "retry_when": null,
  "first_recorded": "2026-10-21"
 }
}

• $.chaining not-recorded
• $.confidence_note withheld
• $.egress.terms not-recorded
• $.extras.dead_end.retry_when process
• $.free_tier.notes process
• $.identity_cost.detail defaulted
• $.identity_cost.kind identity-defaulted
• $.interface interface-unknown
• $.interface_detail withheld
• $.limits_note process
• $.per_object_cap.notes not-recorded
• $.per_object_cap.stated not-recorded
• $.retention.notes process
• $.retention.stated not-recorded
• $.tooling not-recorded