{
 "features_resolved": {
  "snapshot_any": false,
  "snapshot_mem": false,
  "fork": false,
  "pause_resume": false,
  "persistent_disk": false,
  "whole_disk": false,
  "persist_scope": null,
  "volumes": true,
  "auto_stop_idle": true,
  "long_sessions": true,
  "custom_image": true,
  "image_snapshot": false,
  "image_oci": true,
  "vm_isolation": false,
  "docker_inside": false,
  "nested_virt": false,
  "root": null,
  "systemd": null,
  "browser": false,
  "desktop": false,
  "computer_use": false,
  "code_interpreter": null,
  "browser_control": null,
  "desktop_control": null,
  "browser_and_desktop": false,
  "anti_bot": false,
  "captcha_solving": false,
  "residential_ip": false,
  "agent_harnesses": false,
  "ssh": true,
  "public_ipv4": false,
  "inbound_https": true,
  "custom_domain": null,
  "raw_tcp_inbound": null,
  "egress_allowlist": true,
  "open_internet": true,
  "static_egress_ip": false,
  "private_network": null,
  "self_host": true,
  "byoc": true,
  "open_source": false,
  "soc2": true,
  "hipaa": true,
  "sso": true,
  "gpu": true,
  "arm64": false,
  "windows": false,
  "macos": false,
  "lang_python": true,
  "lang_node": false,
  "lang_go": null,
  "lang_rust": null,
  "lang_java": null,
  "gpu_desktop": null,
  "credential_injection": null,
  "egress_http_rules": false,
  "wake_on_request": true,
  "live_resize": false,
  "fork_running": false,
  "memory_snapshot_fork": false,
  "webhooks": null,
  "scheduled_wakeups": null,
  "audit_logs": null,
  "eu_data_residency": true,
  "zero_data_retention": null,
  "scoped_api_keys": null,
  "spend_limits": false,
  "mcp_server": true,
  "usage_api": true,
  "snapshot_auto": false,
  "snapshot_on_demand": false,
  "harness_api": false,
  "own_agent_api": false,
  "fast_boot": null,
  "strong_isolation": false,
  "ingress_rules": true,
  "guest_firewall": false,
  "secret_proxy": false,
  "secret_proxy_any": false,
  "volume_attach": true,
  "volume_shared": true,
  "sidecar": null,
  "fuse": null,
  "kernel_features": null,
  "burst": null
 },
 "company": {
  "id": "baseten",
  "checked": "2026-09-30",
  "company": "Baseten Labs Inc.",
  "ownership": {
   "value": "independent",
   "parent": null,
   "source": "https://www.baseten.co/about/",
   "quote": "We're an interdisciplinary team of researchers, engineers, and operators building the Inference Cloud our AI future demands."
  },
  "stage": {
   "value": "series-d+",
   "source": "https://www.baseten.co/blog/announcing-our-series-f/",
   "quote": "Today, we are thrilled to announce Baseten's $1.5B Series F, led by Altimeter Capital, Conviction Partners, and Spark Capital, co-led by Sands Capital and Wellington Management",
   "confidence": "high",
   "note": "Last priced round is Series F ($1.5B, June 2026, $13B valuation); mapped to the 'series-d+' bucket which covers Series D and later."
  },
  "accelerators": [
   {
    "name": "South Park Commons",
    "batch": null,
    "source": "https://www.southparkcommons.com/companies",
    "quote": "listed in the South Park Commons portfolio as \"Baseten\"",
    "confidence": "high",
    "via": "portfolio-domain"
   }
  ],
  "total_funding_usd": {
   "value": 1915000000,
   "status": "estimated",
   "basis": "Sum of publicly reported priced/strategic rounds found: Series B $40M (Mar 2024) + Series C $75M (Feb 2025) + Series D $150M (Sep 2025) + Nvidia strategic investment $150M (Jan 2026) + Series F $1.5B (Jun 2026). Pre-Series-B funding (seed/Series A amounts) not found, so this total is a floor, not a confirmed cumulative figure.",
   "source": "https://www.baseten.co/blog/announcing-our-series-f/",
   "quote": "Baseten raised a $1.5B Series F and achieved a $13B valuation"
  },
  "last_round": {
   "type": "series-f",
   "amount_usd": 1500000000,
   "date": "2026-06",
   "leads": [
    "Altimeter Capital",
    "Conviction Partners",
    "Spark Capital"
   ],
   "status": "confirmed",
   "source": "https://www.baseten.co/blog/announcing-our-series-f/",
   "quote": "we are thrilled to announce Baseten's $1.5B Series F, led by Altimeter Capital, Conviction Partners, and Spark Capital, co-led by Sands Capital and Wellington Management, with participation from Battery Ventures, Blackbird, D.E. Shaw Ventures, Durable Capital Partners, Greylock, IVP, Verified Capital, and 01A."
  },
  "investors": [
   "Altimeter Capital",
   "Conviction Partners",
   "Spark Capital",
   "Sands Capital",
   "Wellington Management",
   "Nvidia",
   "IVP",
   "Greylock",
   "Battery Ventures",
   "Blackbird",
   "D.E. Shaw Ventures",
   "Durable Capital Partners",
   "Verified Capital",
   "01A",
   "South Park Commons",
   "Lachy Groom",
   "Base Case",
   "Conviction (Sarah Guo)",
   "Sapphire Ventures"
  ],
  "portfolio_listings": [
   {
    "source": "Conviction",
    "company": "Baseten",
    "batch": null,
    "how": "domain"
   },
   {
    "source": "Greylock",
    "company": "Baseten",
    "batch": null,
    "how": "domain"
   },
   {
    "source": "IVP (Institutional Venture Partners)",
    "company": "Baseten",
    "batch": null,
    "how": "exact-name"
   },
   {
    "source": "Sapphire Ventures",
    "company": "Baseten",
    "batch": null,
    "how": "exact-name"
   },
   {
    "source": "South Park Commons",
    "company": "Baseten",
    "batch": null,
    "how": "domain"
   },
   {
    "source": "Spark Capital",
    "company": "Baseten",
    "batch": null,
    "how": "domain"
   }
  ],
  "revenue": {
   "value": null,
   "status": "unknown"
  },
  "employees": {
   "value": null,
   "source": ""
  },
  "note": "Baseten (founded 2019, CEO Tuhin Srivastava) is an independent AI inference-cloud startup, not a YC company (ycombinator.com/companies/baseten returns 404 — hint rejected). Rapid raise cadence: Series B $40M (Mar 2024, IVP+Spark led, Greylock/South Park Commons/Lachy Groom/Base Case participated) -> Series C $75M at $825M valuation (Feb 2025, per company Twitter) -> Series D $150M at $2.15B valuation (Sep 2025, Fortune) -> Nvidia $150M strategic investment (Jan 2026, WSJ) -> Series F $1.5B at $13B valuation (Jun 2026, company blog, 'fourth fundraise in 18 months'). Also acquihired Inferless (Feb 2026). Separately, a security researcher (Strix) reported gaining admin access to a Baseten production GitHub org in Sep 2026 — a notable security incident buyers should be aware of."
 },
 "id": "baseten",
 "name": "Baseten",
 "url": "https://docs.baseten.co/deployment/resources",
 "category": "gpu-cloud",
 "isolation": "container",
 "modes": [
  {
   "key": "serverless-t4-0",
   "label": "T4, serverless (×1)",
   "flags": [
    "alt"
   ],
   "os": [
    "linux"
   ],
   "regions": null,
   "gpu_only": true,
   "features": {},
   "pricing": "resource",
   "vcpu_h": 0,
   "ram_gib_h": 0,
   "cpu_basis": "alloc",
   "ram_basis": "alloc",
   "active_floor": 0,
   "ram_per_vcpu": [
    null,
    null
   ],
   "ram_fixed_per_vcpu": null,
   "vcpu_options": null,
   "min_vcpu": null,
   "max_vcpu": 4,
   "max_ram_gib": 16,
   "sizes": null,
   "pool_tiers": null,
   "requires_always_on": false,
   "always_on_month_per_instance": null,
   "commit_note": null,
   "min_billed_seconds": 60,
   "granularity_s": 60,
   "start_fee": null,
   "boot_overhead_s": null,
   "multiplier": 1,
   "windows_vcpu_h": null,
   "gpu": {
    "T4": 0.6312
   },
   "note": " Source: https://docs.baseten.co/deployment/resources",
   "cpu_class": "shared"
  },
  {
   "key": "serverless-l4-1",
   "label": "L4, serverless (×1)",
   "flags": [
    "alt"
   ],
   "os": [
    "linux"
   ],
   "regions": null,
   "gpu_only": true,
   "features": {},
   "pricing": "resource",
   "vcpu_h": 0,
   "ram_gib_h": 0,
   "cpu_basis": "alloc",
   "ram_basis": "alloc",
   "active_floor": 0,
   "ram_per_vcpu": [
    null,
    null
   ],
   "ram_fixed_per_vcpu": null,
   "vcpu_options": null,
   "min_vcpu": null,
   "max_vcpu": 4,
   "max_ram_gib": 16,
   "sizes": null,
   "pool_tiers": null,
   "requires_always_on": false,
   "always_on_month_per_instance": null,
   "commit_note": null,
   "min_billed_seconds": 60,
   "granularity_s": 60,
   "start_fee": null,
   "boot_overhead_s": null,
   "multiplier": 1,
   "windows_vcpu_h": null,
   "gpu": {
    "L4": 0.8484
   },
   "note": " Source: https://docs.baseten.co/deployment/resources",
   "cpu_class": "shared"
  },
  {
   "key": "serverless-a10g-2",
   "label": "A10G, serverless (×1)",
   "flags": [
    "alt"
   ],
   "os": [
    "linux"
   ],
   "regions": null,
   "gpu_only": true,
   "features": {},
   "pricing": "resource",
   "vcpu_h": 0,
   "ram_gib_h": 0,
   "cpu_basis": "alloc",
   "ram_basis": "alloc",
   "active_floor": 0,
   "ram_per_vcpu": [
    null,
    null
   ],
   "ram_fixed_per_vcpu": null,
   "vcpu_options": null,
   "min_vcpu": null,
   "max_vcpu": 4,
   "max_ram_gib": 16,
   "sizes": null,
   "pool_tiers": null,
   "requires_always_on": false,
   "always_on_month_per_instance": null,
   "commit_note": null,
   "min_billed_seconds": 60,
   "granularity_s": 60,
   "start_fee": null,
   "boot_overhead_s": null,
   "multiplier": 1,
   "windows_vcpu_h": null,
   "gpu": {
    "A10G": 1.2072
   },
   "note": " Source: https://docs.baseten.co/deployment/resources",
   "cpu_class": "shared"
  },
  {
   "key": "serverless-a100-80gb-3",
   "label": "A100-80GB, serverless (×1)",
   "flags": [
    "alt"
   ],
   "os": [
    "linux"
   ],
   "regions": null,
   "gpu_only": true,
   "features": {},
   "pricing": "resource",
   "vcpu_h": 0,
   "ram_gib_h": 0,
   "cpu_basis": "alloc",
   "ram_basis": "alloc",
   "active_floor": 0,
   "ram_per_vcpu": [
    null,
    null
   ],
   "ram_fixed_per_vcpu": null,
   "vcpu_options": null,
   "min_vcpu": null,
   "max_vcpu": 12,
   "max_ram_gib": 144,
   "sizes": null,
   "pool_tiers": null,
   "requires_always_on": false,
   "always_on_month_per_instance": null,
   "commit_note": null,
   "min_billed_seconds": 60,
   "granularity_s": 60,
   "start_fee": null,
   "boot_overhead_s": null,
   "multiplier": 1,
   "windows_vcpu_h": null,
   "gpu": {
    "A100-80GB": 4.0002
   },
   "note": " Source: https://docs.baseten.co/deployment/resources",
   "cpu_class": "shared"
  },
  {
   "key": "serverless-h100-4",
   "label": "H100, serverless (×1)",
   "flags": [
    "alt"
   ],
   "os": [
    "linux"
   ],
   "regions": null,
   "gpu_only": true,
   "features": {},
   "pricing": "resource",
   "vcpu_h": 0,
   "ram_gib_h": 0,
   "cpu_basis": "alloc",
   "ram_basis": "alloc",
   "active_floor": 0,
   "ram_per_vcpu": [
    null,
    null
   ],
   "ram_fixed_per_vcpu": null,
   "vcpu_options": null,
   "min_vcpu": null,
   "max_vcpu": 16,
   "max_ram_gib": 118,
   "sizes": null,
   "pool_tiers": null,
   "requires_always_on": false,
   "always_on_month_per_instance": null,
   "commit_note": null,
   "min_billed_seconds": 60,
   "granularity_s": 60,
   "start_fee": null,
   "boot_overhead_s": null,
   "multiplier": 1,
   "windows_vcpu_h": null,
   "gpu": {
    "H100": 6.4998
   },
   "note": " Source: https://docs.baseten.co/deployment/resources",
   "cpu_class": "shared"
  },
  {
   "key": "serverless-h200-5",
   "label": "H200, serverless (×1)",
   "flags": [
    "alt"
   ],
   "os": [
    "linux"
   ],
   "regions": null,
   "gpu_only": true,
   "features": {},
   "pricing": "resource",
   "vcpu_h": 0,
   "ram_gib_h": 0,
   "cpu_basis": "alloc",
   "ram_basis": "alloc",
   "active_floor": 0,
   "ram_per_vcpu": [
    null,
    null
   ],
   "ram_fixed_per_vcpu": null,
   "vcpu_options": null,
   "min_vcpu": null,
   "max_vcpu": 16,
   "max_ram_gib": 200,
   "sizes": null,
   "pool_tiers": null,
   "requires_always_on": false,
   "always_on_month_per_instance": null,
   "commit_note": null,
   "min_billed_seconds": 60,
   "granularity_s": 60,
   "start_fee": null,
   "boot_overhead_s": null,
   "multiplier": 1,
   "windows_vcpu_h": null,
   "gpu": {
    "H200": 7.5
   },
   "note": " Source: https://docs.baseten.co/deployment/resources",
   "cpu_class": "shared"
  },
  {
   "key": "serverless-b200-6",
   "label": "B200, serverless (×1)",
   "flags": [
    "alt"
   ],
   "os": [
    "linux"
   ],
   "regions": null,
   "gpu_only": true,
   "features": {},
   "pricing": "resource",
   "vcpu_h": 0,
   "ram_gib_h": 0,
   "cpu_basis": "alloc",
   "ram_basis": "alloc",
   "active_floor": 0,
   "ram_per_vcpu": [
    null,
    null
   ],
   "ram_fixed_per_vcpu": null,
   "vcpu_options": null,
   "min_vcpu": null,
   "max_vcpu": 16,
   "max_ram_gib": 224,
   "sizes": null,
   "pool_tiers": null,
   "requires_always_on": false,
   "always_on_month_per_instance": null,
   "commit_note": null,
   "min_billed_seconds": 60,
   "granularity_s": 60,
   "start_fee": null,
   "boot_overhead_s": null,
   "multiplier": 1,
   "windows_vcpu_h": null,
   "gpu": {
    "B200": 9.9798
   },
   "note": " Source: https://docs.baseten.co/deployment/resources",
   "cpu_class": "shared"
  },
  {
   "key": "serverless-rtx-pro-6000-7",
   "label": "RTX-PRO-6000, serverless (×1)",
   "flags": [
    "alt"
   ],
   "os": [
    "linux"
   ],
   "regions": null,
   "gpu_only": true,
   "features": {},
   "pricing": "resource",
   "vcpu_h": 0,
   "ram_gib_h": 0,
   "cpu_basis": "alloc",
   "ram_basis": "alloc",
   "active_floor": 0,
   "ram_per_vcpu": [
    null,
    null
   ],
   "ram_fixed_per_vcpu": null,
   "vcpu_options": null,
   "min_vcpu": null,
   "max_vcpu": 16,
   "max_ram_gib": 116,
   "sizes": null,
   "pool_tiers": null,
   "requires_always_on": false,
   "always_on_month_per_instance": null,
   "commit_note": null,
   "min_billed_seconds": 60,
   "granularity_s": 60,
   "start_fee": null,
   "boot_overhead_s": null,
   "multiplier": 1,
   "windows_vcpu_h": null,
   "gpu": {
    "RTX-PRO-6000": 4.0002
   },
   "note": " Source: https://docs.baseten.co/deployment/resources",
   "cpu_class": "shared"
  },
  {
   "key": "serverless-h100-mig-40gb-8",
   "label": "H100 MIG 40GB, serverless (×1)",
   "flags": [
    "alt"
   ],
   "os": [
    "linux"
   ],
   "regions": null,
   "gpu_only": true,
   "features": {},
   "pricing": "resource",
   "vcpu_h": 0,
   "ram_gib_h": 0,
   "cpu_basis": "alloc",
   "ram_basis": "alloc",
   "active_floor": 0,
   "ram_per_vcpu": [
    null,
    null
   ],
   "ram_fixed_per_vcpu": null,
   "vcpu_options": null,
   "min_vcpu": null,
   "max_vcpu": 8,
   "max_ram_gib": 59,
   "sizes": null,
   "pool_tiers": null,
   "requires_always_on": false,
   "always_on_month_per_instance": null,
   "commit_note": null,
   "min_billed_seconds": 60,
   "granularity_s": 60,
   "start_fee": null,
   "boot_overhead_s": null,
   "multiplier": 1,
   "windows_vcpu_h": null,
   "gpu": null,
   "note": " Source: https://docs.baseten.co/deployment/resources ESTIMATOR EXCLUDED: Fractional slice cannot be represented by a full-GPU map. Exact researched rate remains in gpu.json.",
   "cpu_class": "shared"
  },
  {
   "key": "cpu-only",
   "label": "CPU-only instance (dedicated deployment)",
   "flags": [
    "alt"
   ],
   "os": [
    "linux"
   ],
   "gpu_only": false,
   "pricing": "sizes",
   "sizes": [
    {
     "name": "1x2",
     "vcpu": 1,
     "ram_gib": 2,
     "disk_gib": null,
     "hour": 0.0348,
     "month_cap": null
    },
    {
     "name": "1x4",
     "vcpu": 1,
     "ram_gib": 4,
     "disk_gib": null,
     "hour": 0.0516,
     "month_cap": null
    },
    {
     "name": "2x8",
     "vcpu": 2,
     "ram_gib": 8,
     "disk_gib": null,
     "hour": 0.1038,
     "month_cap": null
    },
    {
     "name": "4x16",
     "vcpu": 4,
     "ram_gib": 16,
     "disk_gib": null,
     "hour": 0.2076,
     "month_cap": null
    },
    {
     "name": "8x32",
     "vcpu": 8,
     "ram_gib": 32,
     "disk_gib": null,
     "hour": 0.4146,
     "month_cap": null
    },
    {
     "name": "16x64",
     "vcpu": 16,
     "ram_gib": 64,
     "disk_gib": null,
     "hour": 0.8292,
     "month_cap": null
    }
   ],
   "min_billed_seconds": 60,
   "granularity_s": 60,
   "boot_overhead_s": 900,
   "note": "Per-minute list prices x60. Shared CPU instance types; cannot select a region. Truss model-serving containers; general agent sandboxes are the separate private-preview Sandboxes product (mode \"sandboxes\", unpriced).",
   "cpu_class": "shared"
  },
  {
   "key": "sandboxes",
   "label": "Sandboxes (private preview)",
   "flags": [
    "beta",
    "sales"
   ],
   "os": [
    "linux"
   ],
   "regions": null,
   "gpu_only": false,
   "features": {
    "pause_resume": true,
    "persistent_disk": true,
    "wake_on_request": true,
    "live_resize": false,
    "max_session_h": null,
    "max_session_unlimited": false,
    "mcp_server": true,
    "lang_node": true,
    "lang_go": true,
    "custom_image": true,
    "image_oci": true,
    "inbound_https": true,
    "isolation": null,
    "ssh": null,
    "gpu": null,
    "gpu_types": [],
    "snapshot": null,
    "snapshot_on_demand": null,
    "snapshot_auto": null,
    "fork": null,
    "fork_running": null,
    "memory_snapshot_fork": null,
    "volume_attach": null,
    "volume_shared": null,
    "egress_allowlist": null,
    "egress_http_rules": null,
    "usage_api": null,
    "agent_harnesses": null,
    "docker_inside": null,
    "root": null,
    "guest_firewall": null,
    "ingress_rules": null
   },
   "pricing": "resource",
   "vcpu_h": null,
   "ram_gib_h": null,
   "cpu_basis": "alloc",
   "ram_basis": "alloc",
   "active_floor": 0,
   "ram_per_vcpu": [
    null,
    null
   ],
   "ram_fixed_per_vcpu": null,
   "vcpu_options": null,
   "min_vcpu": null,
   "max_vcpu": null,
   "max_ram_gib": null,
   "sizes": null,
   "pool_tiers": null,
   "requires_always_on": false,
   "always_on_month_per_instance": null,
   "commit_note": null,
   "min_billed_seconds": null,
   "granularity_s": null,
   "start_fee": null,
   "boot_overhead_s": null,
   "multiplier": 1,
   "windows_vcpu_h": null,
   "gpu": null,
   "cpu_class": null,
   "note": "No sandbox price published; unpriced, never 0. \"Sandboxes is in Private Preview. Contact Baseten to request access.\" and \"A sandbox is an isolated environment where an agent runs commands and works with files.\" (https://docs.baseten.co/sandboxes/overview). Started from a container image (built-in Base Image or your own Dockerfile / registry import); the runtime isolation technology is not published. Sized by memory only: \"memory (which also sets its CPU allocation), and a region. Name, image, memory, and region can't change after creation.\" Default 4096 MiB; \"An organization can run up to 10 sandboxes at once.\" (https://docs.baseten.co/sandboxes/manage). Standby keeps state: \"`DEACTIVATED`: the sandbox is on standby. Its files and processes persist, and the next request resumes it.\"; \"By default, sandboxes expire 7 days after creation.\" (TTL_MAX_AGE / TTL_IDLE / DATE policies, list can't be empty); \"Deletion and expiration permanently remove a sandbox's local data.\" (https://docs.baseten.co/sandboxes/lifecycle). Ports reached through the API server's /port/<PORT>/ route; per-sandbox MCP endpoint https://<sandbox-url>/mcp; TypeScript (@basetenlabs/sandbox), Python and Go SDKs; CLI baseten sandbox create/exec/connect. Checked 2026-10-14."
  }
 ],
 "storage": {
  "included_disk_gib": null,
  "disk_gib_month": null,
  "disk_billed_when_stopped": null,
  "snapshot_gib_month": null,
  "paused_ram_billed": null
 },
 "network": {
  "egress_free_gib": null,
  "egress_gib": null,
  "ipv4_month": null,
  "ipv4_available": null,
  "internet": {
   "default": "configurable",
   "evidence": [
    {
     "quote": "By default, a deployment can make outbound requests to any host on the internet on TCP ports 443 and 80.",
     "url": "https://docs.baseten.co/development/model/egress-restrictions"
    },
    {
     "quote": "Egress restrictions must be enabled for your organization. Contact support to request access.",
     "url": "https://docs.baseten.co/development/model/egress-restrictions"
    }
   ]
  }
 },
 "plans": [
  {
   "name": "Basic",
   "fee": 0,
   "flags": [],
   "fee_is_credit": false,
   "included_usd": 0,
   "per_seat": null,
   "concurrency": null,
   "max_session_h": null,
   "max_vcpu": null,
   "max_ram_gib": null,
   "sizes_allowed": null,
   "note": "$0 per month, pay as you go. Invoiced when usage exceeds $50 or at month end until payment history is established. If credits run out with no payment method, models are deactivated."
  },
  {
   "name": "Pro",
   "fee": null,
   "flags": [
    "sales"
   ],
   "fee_is_credit": false,
   "included_usd": null,
   "per_seat": null,
   "concurrency": null,
   "max_session_h": null,
   "max_vcpu": null,
   "max_ram_gib": null,
   "sizes_allowed": null,
   "note": "Unlimited autoscaling and priority compute access; priority access to high-demand GPUs; volume discounts; quote only."
  },
  {
   "name": "Enterprise",
   "fee": null,
   "flags": [
    "sales"
   ],
   "fee_is_credit": false,
   "included_usd": null,
   "per_seat": null,
   "concurrency": null,
   "max_session_h": null,
   "max_vcpu": null,
   "max_ram_gib": null,
   "sizes_allowed": null,
   "note": "Verify quotas and billing agreement."
  }
 ],
 "free": {
  "monthly_credit": null,
  "one_time_credit": null
 },
 "features": {
  "linux": true,
  "windows": false,
  "macos": false,
  "arm64": false,
  "gpu": true,
  "gpu_types": [
   "A100-80GB",
   "A10G",
   "B200",
   "H100",
   "H200",
   "L4",
   "RTX-PRO-6000",
   "T4"
  ],
  "snapshot": "none",
  "pause_resume": false,
  "fork": false,
  "persistent_disk": false,
  "volumes": null,
  "auto_stop_idle": true,
  "max_session_h": null,
  "custom_image": true,
  "docker_inside": false,
  "nested_virt": false,
  "root": null,
  "systemd": null,
  "browser": false,
  "desktop": false,
  "computer_use": false,
  "code_interpreter": null,
  "agent_harnesses": false,
  "ssh": true,
  "public_ipv4": false,
  "inbound_https": true,
  "custom_domain": null,
  "raw_tcp_inbound": null,
  "egress_allowlist": true,
  "static_egress_ip": false,
  "private_network": null,
  "self_host": true,
  "byoc": true,
  "open_source": false,
  "regions": [
   "us",
   "eu"
  ],
  "soc2": true,
  "hipaa": true,
  "sso": true,
  "lang_python": true,
  "lang_node": false,
  "lang_go": null,
  "lang_rust": null,
  "lang_java": null,
  "gpu_desktop": null,
  "credential_injection": null,
  "egress_http_rules": false,
  "wake_on_request": true,
  "live_resize": false,
  "fork_running": false,
  "memory_snapshot_fork": false,
  "webhooks": null,
  "scheduled_wakeups": null,
  "audit_logs": null,
  "eu_data_residency": true,
  "zero_data_retention": null,
  "scoped_api_keys": null,
  "spend_limits": false,
  "mcp_server": true,
  "snapshot_auto": false,
  "harness_api": false,
  "ingress_rules": true,
  "guest_firewall": false,
  "secret_proxy": false,
  "volume_attach": true,
  "volume_shared": true,
  "isolation": "container",
  "image_oci": true,
  "anti_bot": false,
  "captcha_solving": false,
  "egress_ip_type": "datacenter",
  "snapshot_on_demand": false,
  "own_agent_api": false,
  "max_session_unlimited": true,
  "open_internet": true,
  "usage_api": true
 },
 "perf": {
  "cpu_runs_s": null,
  "cold_start_ms": null,
  "burst_tti_ms": null,
  "dax_s": null
 },
 "knobs": [],
 "caveats": [
  "Bundled CPU/RAM; billed by minute. Deployment/load/scale time billed; scale-to-zero is not free warm replicas. H100 MIG is 3/7 compute and 1/2 memory, not half a full H100. Custom model packaging via Truss; model deployments are not a general secure code-exec sandbox (see the separate private-preview Sandboxes caveat).",
  "Any null count, CPU/RAM, billing increment or minimum is unverified, not unlimited/free.",
  "This is GPU-product research, not a claim that every feature of the entire vendor documentation was audited. Unverified features are null.",
  "GPU multi-count bundles, variant/region selection and fractional slices require a SKU-aware estimator; unsupported rows remain unpriced in strict cards.",
  "Modes with gpu=null are intentionally excluded, not free GPU modes. Never substitute 0 for a null rate. Flags alt indicate DIY/ML infrastructure, not the same thing as managed untrusted-code sandboxes.",
  "Unknown fee, storage or bandwidth means full monthly total may be unavailable. Rates exclude taxes.",
  "Larger CPU/RAM shapes on the same GPU cost more (T4x8x32 $0.01504/min, T4x16x64 $0.02408/min, A10Gx16x64 $0.03248/min); multi-GPU SKUs are linear (H100:8 $0.86664/min). Card models only the smallest single-GPU shape.",
  "Image builds are billed on every push; cold start/model load is billed; failed boots and image pulls are not; development deployments (--watch) are listed as not billed. Partial minutes round up.",
  "Default autoscaling scale-down delay is 900 s (configurable 0-3600 s): with defaults each burst keeps the replica billed ~15 more idle minutes, plus billed cold start/model load. Not modelled.",
  "Sandboxes (agent sandboxes, private preview, access by contacting Baseten) is modelled as the unpriced mode \"sandboxes\": no sandbox price, billing unit or isolation technology is published (https://docs.baseten.co/sandboxes/overview, checked 2026-10-14). Its 10-concurrent-sandbox organization quota (\"An organization can run up to 10 sandboxes at once.\", https://docs.baseten.co/sandboxes/manage) is per product and is not expressible in the card-level fleet fields.",
  "Default egress from a deployment is open only on TCP 443 and 80 (\"By default, a deployment can make outbound requests to any host on the internet on TCP ports 443 and 80.\", https://docs.baseten.co/development/model/egress-restrictions); other outbound ports are not reachable by default. The card's internet level (configurable) cannot express a port limit.",
  "Monthly budget enforcement covers Model APIs only; dedicated deployments and training keep running and billing past the budget (https://docs.baseten.co/organization/billing)."
 ],
 "sources": [
  "https://docs.baseten.co/deployment/resources",
  "https://www.baseten.co/pricing/",
  "https://docs.baseten.co/organization/billing",
  "https://docs.baseten.co/development/model/egress-restrictions",
  "https://docs.baseten.co/sandboxes/overview",
  "https://docs.baseten.co/sandboxes/lifecycle",
  "https://docs.baseten.co/sandboxes/manage"
 ],
 "audit_2026_09_29": {
  "applied": 6,
  "source": "checked against the provider's live pricing and docs pages",
  "reviewed": 6
 },
 "fleet": {
  "default_max_instances": null,
  "default_max_vcpu": null,
  "default_max_gpus": null,
  "raise": "ticket",
  "raise_time_days": null,
  "verification": "payment method after credits; org verification for regional deployments",
  "tiers": "Pro advertises 'Unlimited autoscaling and priority compute access'; per-deployment max_replica defaults to 1 (user-adjustable); organization-level concurrent GPU limits exist (docs example: 16 H100) and are raised by contacting Baseten",
  "notes": "Default org GPU quota for a new Basic workspace is not published. Private-preview Sandboxes: \"An organization can run up to 10 sandboxes at once. To raise this quota, contact Baseten support with your organization ID.\" (https://docs.baseten.co/sandboxes/manage, 2026-10-14).",
  "sources": [
   "https://docs.baseten.co/deployment/autoscaling/overview.md",
   "https://docs.baseten.co/llms-full.txt",
   "https://www.baseten.co/pricing/",
   "https://docs.baseten.co/sandboxes/manage"
  ],
  "researched": "2026-09-29"
 },
 "feature_evidence": {
  "snapshot_auto": {
   "value": false,
   "source": "https://docs.baseten.co/deployment/autoscaling/cold-starts",
   "quote": "cold start = container pull + weight load + engine initialization",
   "note": "Scale-to-zero shuts replicas down; on wake a fresh replica is started (image pull, weight load, engine init). No state checkpoint/restore of replicas is documented. Speed-ups come from caches (BDN weight cache, b10cache torch.compile cache), not machine snapshots.",
   "checked": "2026-09-29"
  },
  "harness_api": {
   "value": false,
   "source": "https://docs.baseten.co/inference/model-apis/coding-agents",
   "quote": "Use [Baseten Switch](/reference/cli/baseten-switch) to route requests from Claude Code or Codex CLI to Model APIs.",
   "note": "Baseten's coding-agent features are an LLM routing CLI (Baseten Switch) and an MCP/skill to manage the workspace. The private-preview Sandboxes product runs an agent's tool calls (exec, files, per-sandbox MCP tools: \"To have your coding agent work in it, let the agent run the same CLI commands.\", https://docs.baseten.co/sandboxes/overview), but there is no first-party API that runs or drives an agent harness inside a sandbox.",
   "checked": "2026-10-14"
  },
  "egress_allowlist": {
   "value": true,
   "source": "https://docs.baseten.co/development/model/egress-restrictions",
   "quote": "Egress restrictions limit a deployment's outbound traffic to an allow list of IP ranges and domains that you define in its config.yaml, so you can lock down a sensitive model without affecting any other deployment in your organization.",
   "note": "Per-deployment runtime.egress_restrictions with ip_allow_list (public IPv4 / CIDR only, no IPv6) and fqdn_allow_list (wildcards anywhere in a label); both empty blocks all egress. Gated: \"Egress restrictions must be enabled for your organization. Contact support to request access.\" Fixed per deployment version (\"Allow lists can't be updated in place on an existing deployment\"). Model deployments only, not Chains or training jobs, and traffic through the BDN proxy bypasses it. Enterprise single-tenant can also run replicas without direct internet access (https://docs.baseten.co/hosting-options/single-tenant). Not documented for Sandboxes.",
   "level": "both",
   "checked": "2026-10-14"
  },
  "ingress_rules": {
   "value": true,
   "source": "https://docs.baseten.co/hosting-options/single-tenant",
   "quote": "You can also restrict inference to a specific IP range or to traffic originating within your VPC.",
   "note": "All Baseten Cloud endpoints are API-key authenticated (environment-scoped / model-scoped inference-only keys available); single-tenant adds PrivateLink / Private Service Connect and IP-range restriction. SSH to replicas uses short-lived certs limited to users who can manage the model.",
   "checked": "2026-09-29"
  },
  "guest_firewall": {
   "value": false,
   "source": "https://docs.baseten.co/inference/ssh",
   "quote": "Your container runs as the non-root `app` user.",
   "note": "Kubernetes containers (Calico/Cilium network isolation); 'Workloads and nodes run with minimal privileges'. No root or NET_ADMIN, so iptables/nftables inside is not possible.",
   "checked": "2026-09-29",
   "derived": true
  },
  "secret_proxy": {
   "value": false,
   "source": "https://docs.baseten.co/development/model/secrets",
   "quote": "Secrets are passed as keyword arguments to the `Model` class.",
   "note": "Workspace/team secret store (UI + Management API); secrets are delivered into the container (kwargs / env for custom servers, training SecretReference). Also OIDC short-lived tokens and AWS AssumeRole for cloud access. No injecting proxy.",
   "checked": "2026-09-29"
  },
  "volume_attach": {
   "value": true,
   "source": "https://docs.baseten.co/development/model/bdn",
   "quote": "Mount existing BDN volumes into your model container at runtime.",
   "note": "BDN volumes: versioned, immutable file trees, mounted read-only ('The mount is read-only.'); `baseten volume` CLI is PRE-RELEASE. Also b10cache (network volume at /cache/org and /cache/model, read-write, 'treat it as a cache') and the training cache ($BT_PROJECT_CACHE_DIR / $BT_TEAM_CACHE_DIR). No generic block volumes.",
   "checked": "2026-09-29"
  },
  "volume_shared": {
   "value": true,
   "source": "https://docs.baseten.co/development/model/runtime-caching",
   "quote": "It's a volume mounted over the network onto each of your replicas, with two scopes:",
   "note": "b10cache is read-write and shared: /cache/org/ 'Shared across every replica you deploy in your organization', /cache/model/ per deployment; not durable storage. BDN volumes are shared read-only across deployments; training cache is shared by all jobs in a project/team on one cluster.",
   "checked": "2026-09-29"
  },
  "isolation": {
   "value": "container",
   "source": "https://docs.baseten.co/inference/ssh",
   "quote": "You're connected when you see a shell prompt inside the model container.",
   "note": "Kubernetes-orchestrated containers, not full VMs; card already models isolation=container.",
   "checked": "2026-09-30"
  },
  "image_oci": {
   "value": true,
   "source": "https://docs.baseten.co/development/model/custom-server",
   "quote": "Set `base_image` to your image and use `docker_server` to specify how to start it",
   "note": "Custom Docker/OCI base images supported via docker_server/base_image config.",
   "checked": "2026-09-30"
  },
  "docker_inside": {
   "value": false,
   "source": "https://docs.baseten.co/inference/ssh",
   "quote": "Your container runs as the non-root `app` user.",
   "note": "not offered: outside the product's scope. Model containers are single-purpose serving containers with no root/privileged access; running Docker-in-Docker is not documented.",
   "checked": "2026-09-30"
  },
  "nested_virt": {
   "value": false,
   "source": "https://docs.baseten.co/inference/ssh",
   "quote": "Your container runs as the non-root `app` user.",
   "note": "not offered: outside the product's scope. No /dev/kvm or nested virtualization documented for a non-root Kubernetes pod.",
   "checked": "2026-09-30"
  },
  "desktop": {
   "value": false,
   "source": "https://docs.baseten.co/deployment/resources",
   "quote": "Every deployed model runs on an instance, a dedicated set of CPU, memory, and optional GPU resources.",
   "note": "not offered for model deployments. Baseten is a model-serving/inference platform, not a desktop/VNC provider. The private-preview Sandboxes pages (overview, lifecycle, manage) document none either.",
   "checked": "2026-10-14"
  },
  "browser": {
   "value": false,
   "source": "https://docs.baseten.co/deployment/resources",
   "quote": "Every deployed model runs on an instance, a dedicated set of CPU, memory, and optional GPU resources.",
   "note": "not offered for model deployments. No managed browser product is documented. The private-preview Sandboxes pages (overview, lifecycle, manage) document none either.",
   "checked": "2026-10-14"
  },
  "computer_use": {
   "value": false,
   "source": "https://docs.baseten.co/deployment/resources",
   "quote": "Every deployed model runs on an instance, a dedicated set of CPU, memory, and optional GPU resources.",
   "note": "not offered for model deployments. No screenshot/click desktop-driving API is documented. The private-preview Sandboxes pages (overview, lifecycle, manage) document none either.",
   "checked": "2026-10-14"
  },
  "anti_bot": {
   "value": false,
   "source": "https://docs.baseten.co/deployment/resources",
   "quote": "Every deployed model runs on an instance, a dedicated set of CPU, memory, and optional GPU resources.",
   "note": "not offered for model deployments. Baseten serves models, not stealth/anti-bot browsing infrastructure. The private-preview Sandboxes pages (overview, lifecycle, manage) document none either.",
   "checked": "2026-10-14"
  },
  "captcha_solving": {
   "value": false,
   "source": "https://docs.baseten.co/deployment/resources",
   "quote": "Every deployed model runs on an instance, a dedicated set of CPU, memory, and optional GPU resources.",
   "note": "not offered for model deployments. No CAPTCHA-solving feature documented; not relevant to a model-serving GPU cloud. The private-preview Sandboxes pages (overview, lifecycle, manage) document none either.",
   "checked": "2026-10-14"
  },
  "egress_ip_type": {
   "value": "datacenter",
   "source": "https://docs.baseten.co/development/model/egress-restrictions",
   "quote": "By default, a deployment can make outbound requests to any host on the internet on TCP ports 443 and 80.",
   "note": "Standard cloud egress runs from Baseten's/cloud-provider's datacenter IP ranges; no residential/mobile egress product exists.",
   "checked": "2026-10-14"
  },
  "ssh": {
   "value": true,
   "source": "https://docs.baseten.co/inference/ssh",
   "quote": "SSH into any running model deployment on Baseten and get a full terminal inside the model container",
   "note": "Requires org to enable Inference SSH and enabling it per deployment in config.yaml.",
   "checked": "2026-09-30"
  },
  "snapshot": {
   "value": "none",
   "source": "https://docs.baseten.co/deployment/autoscaling/cold-starts",
   "quote": "cold start = container pull + weight load + engine initialization",
   "note": "Scale-to-zero starts a fresh replica from the image on wake; no filesystem or memory snapshot/restore of a replica is documented. Caches (BDN weight cache, b10cache) speed cold starts but are not full machine snapshots.",
   "checked": "2026-09-30"
  },
  "snapshot_on_demand": {
   "value": false,
   "source": "https://docs.baseten.co/deployment/autoscaling/cold-starts",
   "quote": "cold start = container pull + weight load + engine initialization",
   "note": "No API/CLI to take an on-demand snapshot/checkpoint of a running deployment is documented.",
   "checked": "2026-09-30"
  },
  "fork": {
   "value": false,
   "source": "https://docs.baseten.co/deployment/resources",
   "quote": "Every deployed model runs on an instance, a dedicated set of CPU, memory, and optional GPU resources.",
   "note": "not offered for model deployments. Baseten scales by adding fresh replicas from the same image, not by cloning a running machine's state. The private-preview Sandboxes pages (overview, lifecycle, manage) document none either.",
   "checked": "2026-10-14"
  },
  "fork_running": {
   "value": false,
   "source": "https://docs.baseten.co/deployment/resources",
   "quote": "Every deployed model runs on an instance, a dedicated set of CPU, memory, and optional GPU resources.",
   "note": "not offered for model deployments. No clone-a-running-machine capability documented. The private-preview Sandboxes pages (overview, lifecycle, manage) document none either.",
   "checked": "2026-10-14"
  },
  "memory_snapshot_fork": {
   "value": false,
   "source": "https://docs.baseten.co/deployment/autoscaling/cold-starts",
   "quote": "cold start = container pull + weight load + engine initialization",
   "note": "No memory-inclusive clone/fork feature exists; new replicas always cold-start from the image.",
   "checked": "2026-09-30"
  },
  "pause_resume": {
   "value": false,
   "source": "https://docs.baseten.co/deployment/resources",
   "quote": "Choose 'Enabled' switch to pause it without removing it.",
   "note": "The 'Enabled' toggle removes/disables a deployment (scales to zero replicas) rather than pausing a live machine with reduced billing state preserved; pause/resume documented elsewhere refers to rolling-deployment traffic promotion, not machine pause. No true machine-level pause/resume with preserved state is documented. Model deployments only: the private-preview sandboxes mode does pause (https://docs.baseten.co/sandboxes/lifecycle: \"`DEACTIVATED`: the sandbox is on standby. Its files and processes persist, and the next request resumes it.\"), set in its mode features.",
   "checked": "2026-10-14"
  },
  "persistent_disk": {
   "value": false,
   "source": "https://docs.baseten.co/development/model/runtime-caching",
   "quote": "It's a volume mounted over the network onto each of your replicas",
   "note": "Container filesystem itself is ephemeral per replica; durable storage is via BDN volumes (versioned/read-only) or b10cache (explicitly a cache, not durable), not a generic persistent root disk that survives stop/start of a given machine. Model deployments only: sandbox files persist on standby until deletion or expiry (default 7 days after creation), set in the sandboxes mode features.",
   "checked": "2026-10-14"
  },
  "live_resize": {
   "value": false,
   "source": "https://docs.baseten.co/deployment/resources",
   "quote": "Every deployed model runs on an instance, a dedicated set of CPU, memory, and optional GPU resources.",
   "note": "not offered for model deployments. Changing instance type/resources requires a redeploy, not a live in-place resize. Sandboxes: \"Name, image, memory, and region can't change after creation.\" (https://docs.baseten.co/sandboxes/manage)",
   "checked": "2026-10-14"
  },
  "agent_harnesses": {
   "value": false,
   "source": "https://docs.baseten.co/inference/model-apis/coding-agents",
   "quote": "Use Baseten Switch to route requests from Claude Code or Codex CLI to Model APIs.",
   "note": "Baseten Switch is a local gateway that routes an already-installed coding agent's requests to Baseten Model APIs; no coding agent runs preinstalled inside a Baseten sandbox/container.",
   "checked": "2026-09-30"
  },
  "own_agent_api": {
   "value": false,
   "source": "https://docs.baseten.co/deployment/resources",
   "quote": "Every deployed model runs on an instance, a dedicated set of CPU, memory, and optional GPU resources.",
   "note": "not offered for model deployments. Baseten is a model-inference/training platform; it has no first-party browser/computer-use/research task agent product. The private-preview Sandboxes pages (overview, lifecycle, manage) document none either.",
   "checked": "2026-10-14"
  },
  "mcp_server": {
   "value": true,
   "source": "https://docs.baseten.co/llms-full.txt",
   "quote": "Add the backend MCP server at https://api.baseten.co/mcp with the header \"Authorization: Bearer $BASETEN_MCP_KEY\". Add the docs MCP server at https://docs.baseten.co/mcp.",
   "note": "Backend MCP server (workspace management) plus docs MCP server; each private-preview sandbox also exposes its own MCP endpoint at https://<sandbox-url>/mcp with tools for commands and file edits (https://docs.baseten.co/sandboxes/manage).",
   "checked": "2026-10-14"
  },
  "lang_python": {
   "value": true,
   "source": "https://docs.baseten.co/development/model/model-class",
   "quote": "Baseten installs your Python dependencies, packages the `Model` class into a container, and deploys it.",
   "note": "Python is the primary first-party runtime (Truss Model class, Python SDK).",
   "checked": "2026-09-30"
  },
  "lang_node": {
   "value": false,
   "source": "https://docs.baseten.co/llms-full.txt",
   "quote": "To switch to Baseten with the JavaScript SDK, change `baseURL` and `apiKey` when initializing the client",
   "note": "Node/JavaScript is only offered as a client SDK for calling deployed models over HTTP; there is no first-party Node.js runtime for packaging/running your own model code inside a Baseten container (that path is Python/Truss, or bring-your-own Docker image).",
   "checked": "2026-09-30"
  },
  "public_ipv4": {
   "value": false,
   "source": "https://docs.baseten.co/hosting-options/single-tenant",
   "quote": "Private networking: You control inference ingress through your load balancer, security groups, and provider-native networking controls.",
   "note": "Inference is reached via Baseten-issued HTTPS API endpoints; no documented dedicated/public IPv4 address per model deployment.",
   "checked": "2026-09-30"
  },
  "egress_http_rules": {
   "value": false,
   "source": "https://docs.baseten.co/development/model/egress-restrictions",
   "quote": "Each allowed domain is enforced by resolving it to its current IP addresses and permitting traffic to those IPs; the hostname on the outbound request itself isn't inspected.",
   "note": "Allow-lists are IP/CIDR and resolved-FQDN only; no allow/deny by HTTP method, path or header.",
   "checked": "2026-10-14"
  },
  "static_egress_ip": {
   "value": false,
   "source": "https://docs.baseten.co/development/model/egress-restrictions",
   "quote": "Egress restrictions limit a deployment's outbound traffic to an allow list of IP ranges and domains that you define in its config.yaml, so you can lock down a sensitive model without affecting any other deployment in your organization.",
   "note": "Egress controls cover destinations only; no fixed/static outbound source IP is documented.",
   "checked": "2026-10-14"
  },
  "windows": {
   "value": false,
   "source": "https://docs.baseten.co/deployment/resources",
   "quote": "Every deployed model runs on an instance, a dedicated set of CPU, memory, and optional GPU resources.",
   "note": "not offered for model deployments. All model deployment compute is Linux containers; Windows only appears as a client OS for the CLI/SSH tooling. The private-preview Sandboxes pages (overview, lifecycle, manage) document none either.",
   "checked": "2026-10-14"
  },
  "macos": {
   "value": false,
   "source": "https://docs.baseten.co/deployment/resources",
   "quote": "Every deployed model runs on an instance, a dedicated set of CPU, memory, and optional GPU resources.",
   "note": "not offered for model deployments. No macOS compute product; macOS only appears as a client OS (CLI, Baseten Switch) for developers. The private-preview Sandboxes pages (overview, lifecycle, manage) document none either.",
   "checked": "2026-10-14"
  },
  "arm64": {
   "value": false,
   "source": "https://docs.baseten.co/deployment/resources",
   "quote": "Every deployed model runs on an instance, a dedicated set of CPU, memory, and optional GPU resources.",
   "note": "not offered for model deployments. Baseten's deployment compute (GPU and CPU-only instances) is x86_64; ARM only appears for client-side CLI binaries (macOS arm64 laptops), not server-side compute. The private-preview Sandboxes pages (overview, lifecycle, manage) document none either.",
   "checked": "2026-10-14"
  },
  "open_source": {
   "value": false,
   "source": "https://docs.baseten.co/llms-full.txt",
   "quote": "The `config.yaml` follows Truss, Baseten's open-source model packaging format.",
   "note": "Truss (model packaging format/CLI) is open source, but the core Baseten serving platform/control plane itself is closed-source SaaS.",
   "checked": "2026-09-30"
  },
  "sso": {
   "value": true,
   "source": "https://docs.baseten.co/organization/sso-and-scim",
   "quote": "Baseten supports any SAML 2.0 IdP and any SCIM 2.0 directory provider through WorkOS, including: Okta, Microsoft Entra ID (Azure AD), Google Workspace.",
   "note": "SSO and SCIM available on the Enterprise plan.",
   "checked": "2026-09-30"
  },
  "max_session_h": {
   "value": "none",
   "source": "https://docs.baseten.co/deployment/resources",
   "quote": "Bundled CPU/RAM; billed by minute.",
   "note": "No documented maximum runtime for a deployment/replica; deployments run continuously (subject to autoscaling/scale-to-zero) until removed, billed per minute like an always-on VM.",
   "checked": "2026-09-30"
  },
  "cpu_class": {
   "value": {
    "cpu-only": "shared",
    "serverless-t4-0": "shared",
    "serverless-l4-1": "shared",
    "serverless-a10g-2": "shared",
    "serverless-a100-80gb-3": "shared",
    "serverless-h100-4": "shared",
    "serverless-h200-5": "shared",
    "serverless-b200-6": "shared",
    "serverless-rtx-pro-6000-7": "shared",
    "serverless-h100-mig-40gb-8": "shared"
   },
   "source": "https://docs.baseten.co/llms-full.txt",
   "quote": "Deployments on shared CPU instance types can't select a region. ... Models run on a multi-tenant infrastructure layer with namespace- and container-level isolation, and GPUs are never shared across customers.",
   "note": "Docs explicitly call the non-GPU tier 'shared CPU instance types' (high confidence for cpu-only). For GPU-attached tiers, Baseten Cloud is described as multi-tenant with namespace/container isolation and states only that 'GPUs are never shared across customers' -- it does not extend that dedicated guarantee to vCPUs, and all instances (GPU and CPU-only) are Kubernetes containers on shared nodes (isolation=container). Applying the same shared-vCPU characterization to GPU tiers is a reasonable inference (medium confidence) rather than a directly-quoted per-tier statement.",
   "checked": "2026-09-30"
  },
  "usage_api": {
   "value": true,
   "was": false,
   "source": "https://docs.baseten.co/organization/billing",
   "quote": "The CLI and API separate dedicated inference, Model APIs, and training usage and include per-resource and daily breakdowns. A query can cover up to 31 days and cannot start before January 1, 2026 UTC.",
   "note": "granularity: per-deployment; cost: yes; GET /v1/billing/usage_summary (and `baseten org billing usage --output json`) returns, per billable_resource (kind MODEL_DEPLOYMENT, name, instance_type, environment), subtotal, compute_cost, surcharge_cost, minutes and a daily breakdown. Usage updates hourly. Sandboxes are not among the listed usage categories, so the sandboxes mode leaves usage_api unknown. The earlier quote (\"per-replica or per-deployment cost breakdown is not exposed via API\") is no longer on the page.",
   "confidence": "high",
   "checked": "2026-10-14"
  },
  "spend_limits": {
   "value": false,
   "source": "https://docs.baseten.co/organization/billing",
   "quote": "Enforcement applies only to Model APIs. A budget never stops dedicated deployments or training jobs; they continue to run and accrue charges regardless of this setting.",
   "note": "Monthly budget sends email at 75/90/100%; \"Enforce budget\" only rejects Model API requests, so compute spend is never capped.",
   "checked": "2026-10-14"
  }
 },
 "locations": [
  {
   "country": "US",
   "state": null,
   "city": null,
   "provider_region": "us",
   "self_serve": true,
   "choosable": true
  },
  {
   "country": "DE",
   "state": null,
   "city": null,
   "provider_region": "eu",
   "self_serve": true,
   "choosable": true
  }
 ],
 "locations_source": "https://docs.baseten.co/deployment/regional-deployments",
 "locations_quote": "Region slug Location us United States eu European Union",
 "locations_checked": "2026-10-01"
}