Sources
Where every number comes from, how fresh it is, and whether the last fetch worked. A failed source is shown as failed; its SKUs read "not verified since" and show no price. A source with no good fetch in the last 72 hours is treated the same way.
| Source | Kind | Pricing page | Status | Last good fetch | Last attempt | SKUs | Notes |
|---|---|---|---|---|---|---|---|
anthropic-pricingAnthropicPricing | first-party | docs.anthropic.com/en/docs/about-claude/pricing | ok | 2026-10-23T12:59:19+00:00 | 2026-10-23T12:59:19+00:00 | 7 | Anthropic's own pricing page (cell text parsed: leading $ amount). Long-context tier as *_in_long / *_out_long / *_cached_in_long. |
assemblyai-pricingAssemblyaiPricing | first-party | www.assemblyai.com/pricing | ok | 2026-10-23T16:44:50+00:00 | 2026-10-23T16:44:50+00:00 | 7 | AssemblyAI's own price page: pay-as-you-go speech-to-text listed per audio hour, converted to USD per minute (the hourly price is kept in prices.per_hour). |
aws-bedrock-offers-us-east-1BedrockOffers | first-party | aws.amazon.com/bedrock/pricing/ | ok | 2026-10-23T20:46:13+00:00 | 2026-10-23T20:46:13+00:00 | 134 | AWS Bedrock bulk offer files, us-east-1 (AmazonBedrock and AmazonBedrockFoundationModels), on-demand, USD per 1M tokens; prefers the in-region profile, the global cross-region profile is kept in prices. |
aws-bedrock-pricelistBedrockPricelist | first-party | ok | 2026-10-22T23:53:44+00:00 | 2026-10-22T23:53:44+00:00 | 0 | AWS Bedrock public price list, US East (N. Virginia), on-demand, USD per 1M tokens. Meter keys are resolved to model names with bedrock_skus.json (built from the AWS bulk offer file); unmapped meters are skipped. | |
azure-retail-openaiAzureOpenAI | first-party | azure.microsoft.com/en-us/pricing/details/c… | ok | 2026-10-23T12:28:14+00:00 | 2026-10-23T12:28:14+00:00 | 125 | Azure retail prices, Foundry Models, swedencentral (eastus2 as fallback). Standard text-token meters only, per 1M tokens in USD excl. tax; global deployment preferred, then data zone, then regional. |
azure-retail-vm-westeuropeAzureVm | first-party | azure.microsoft.com/en-us/pricing/details/v… | ok | 2026-10-23T12:28:14+00:00 | 2026-10-23T12:28:14+00:00 | 3565 | Azure Retail Prices API, Virtual Machines, region westeurope, Consumption, Linux. USD per hour, excl. tax; per_month = hourly x 730. Spot meters are separate -spot SKUs. Hardware specs are not in the API. |
baseten-pricingBasetenPricing | first-party | www.baseten.co/pricing/ | ok | 2026-10-23T17:00:12+00:00 | 2026-10-23T17:00:12+00:00 | 15 | Baseten's own Model APIs table, USD per 1M input and output tokens (cache-input column not captured). Fast variants keep their own model name; dedicated GPU deployments are not part of this source's fields. |
cerebras-modelsCerebrasModels | first-party | api.cerebras.ai/public/v1/models | ok | 2026-10-23T18:46:09+00:00 | 2026-10-23T18:46:09+00:00 | 2 | Cerebras Inference's public models API: USD per token converted to per 1M tokens, with context length and max completion tokens. |
coreweave-pricingCoreweavePricing | first-party | www.coreweave.com/pricing | ok | 2026-10-23T00:54:52+00:00 | 2026-10-23T00:54:52+00:00 | 9 | CoreWeave's own GPU pricing table (North America, on-demand): per-GPU price is the 'Inference Single GPU Price', the node price is kept as per_node_hour; spot prices are not archived. |
crusoe-cloud-pricingCrusoePricing | first-party | www.crusoe.ai/cloud/pricing | verdacht:ok | 2026-10-22T16:29:39+00:00 | 2026-10-23T16:44:34+00:00 | 6 | Crusoe Cloud's own GPU instances table, On-demand column, USD per GPU-hour; spot, reserved, managed inference and the Contact-sales rows (GB200, MI355X) are not archived. |
datacrunch-instancesDataCrunch | first-party | datacrunch.io/products | ok | 2026-10-23T23:39:39+00:00 | 2026-10-23T23:39:39+00:00 | 52 | DataCrunch instance types. Price per instance per hour (all GPUs of the type); per-GPU price is derived. CPU-only nodes are not listed. |
deepgram-pricingDeepgramPricing | first-party | deepgram.com/pricing | ok | 2026-10-23T16:44:46+00:00 | 2026-10-23T16:44:46+00:00 | 20 | Deepgram's own price page: speech-to-text USD per minute and text-to-speech USD per 1k characters, Pay As You Go and the prepaid Growth plan, with promotional streaming rates at their current price. |
deepinfra-modelsDeepInfra | first-party | deepinfra.com/ByteDance/Seed-1.8 | ok | 2026-10-23T23:39:40+00:00 | 2026-10-23T23:39:40+00:00 | 101 | DeepInfra model list, token-priced models only. Cents per token converted to USD per 1M tokens. Deprecated models are skipped. |
deepseek-pricingDeepSeekPricing | first-party | api-docs.deepseek.com/quick_start/pricing/ | ok | 2026-10-23T12:59:19+00:00 | 2026-10-23T12:59:19+00:00 | 2 | DeepSeek's own pricing page. Standard prices plus the off-peak discount as *_in_offpeak / *_out_offpeak / *_cached_in_offpeak. |
digitalocean-dropletsDigitaloceanDroplets | first-party | www.digitalocean.com/pricing/droplets | ok | 2026-10-23T00:54:24+00:00 | 2026-10-23T00:54:24+00:00 | 10 | DigitalOcean's Droplet pricing tables (Basic and the first CPU-Optimized plans): USD, monthly price plus hourly price; no tax shown, RAM and disk are GiB. |
digitalocean-gpu-dropletsDigitaloceanGpuDroplets | first-party | www.digitalocean.com/pricing/gpu-droplets | ok | 2026-10-23T16:44:37+00:00 | 2026-10-23T16:44:37+00:00 | 16 | DigitalOcean's own GPU Droplets page, USD per GPU-hour (a Droplet has 1 or 8 GPUs, billed per second): On-Demand plans, 12-month reserved and spot (interruptible) prices. |
elevenlabs-pricingElevenlabsPricing | first-party | elevenlabs.io/pricing | verdacht:ok | 2026-10-22T16:29:47+00:00 | 2026-10-23T16:44:42+00:00 | 3 | ElevenLabs' own plan page and FAQ: speech, transcription and music priced in shared credits, converted to USD at the Starter plan's monthly list price per credit (larger plans are cheaper per credit). |
exoscale-opencomputeExoscaleOpencompute | first-party | www.exoscale.com/pricing/ | ok | 2026-10-23T18:46:09+00:00 | 2026-10-23T18:46:09+00:00 | 26 | Exoscale compute instance types from the portal pricing feed: EUR per hour (monthly = 730 h); the feed lists no vCPU/RAM, zone or VAT status. GPU types are not emitted. |
fal-pricingComposite | first-party | fal.ai/pricing | verdacht:ok | 2026-10-22T00:54:28+00:00 | 2026-10-23T01:10:11+00:00 | 21 | fal.ai's own pricing page, compute table: the list price per GPU-hour (not the 'as low as' column). fal's own pricing page: list price per second of generated video or per image; per-megapixel models (FLUX, Z-Image) are not emitted because the ledger has no per-megapixel unit. GPU rows are read by the fal GPU normalizer. |
fireworks-pricingFireworksGpu | first-party | fireworks.ai/pricing | ok | 2026-10-23T01:10:11+00:00 | 2026-10-23T01:10:11+00:00 | 5 | Fireworks AI's own pricing page, On-demand deployments: USD per GPU-hour. |
google-pricingGooglePricing | first-party | ai.google.dev/gemini-api/docs/pricing | ok | 2026-10-23T12:59:19+00:00 | 2026-10-23T12:59:19+00:00 | 8 | Google's Gemini API pricing page, paid tier. USD per 1M tokens; long-context tier as *_in_long / *_out_long / *_cached_in_long. |
groq-modelsGroqModels | first-party | console.groq.com/docs/models | ok | 2026-10-23T16:29:31+00:00 | 2026-10-23T16:29:31+00:00 | 10 | Groq's console model table: USD per 1M input/output tokens; Whisper (per hour of audio, shown per minute) and Orpheus TTS (per 1M characters, shown per 1k characters). Models listed as contact-sales have no price and are skipped. |
hf-inference-providersHfInferenceProviders | community cross-check | router.huggingface.co/v1/models | ok | 2026-10-23T18:46:09+00:00 | 2026-10-23T18:46:09+00:00 | 175 | Hugging Face's inference-provider router: per model, each provider's input/output price in USD per 1M tokens. An aggregator of other hosts' prices, so a community cross-check, not a first-party price. |
hostinger-vpsHostingerVps | first-party | www.hostinger.com/vps-hosting | ok | 2026-10-23T00:54:34+00:00 | 2026-10-23T00:54:34+00:00 | 4 | Hostinger KVM plans in USD excl. VAT: regular (renewal) price as primary, the first-term promo and the struck-through list price in prices; every price needs the 24-month term. |
hostwinds-vpsHostwindsVps | first-party | www.hostwinds.com/vps/linux | ok | 2026-10-23T00:54:40+00:00 | 2026-10-23T00:54:40+00:00 | 7 | Hostwinds Linux VPS table, first seven SSD Cloud plans: the price on each plan's Order button, USD; the page states no VAT. |
hyperstack-gpu-pricingHyperstackGpu | first-party | www.hyperstack.cloud/gpu-pricing | ok | 2026-10-23T00:54:58+00:00 | 2026-10-23T00:54:58+00:00 | 13 | Hyperstack's own 'On-Demand GPU Pricing' table, USD per GPU-hour, billed per minute; reservation and spot tables are not archived. |
ionos-cloud-pricesIonosCloudPrices | first-party | cloud.ionos.com/prices | ok | 2026-10-23T00:54:43+00:00 | 2026-10-23T00:54:43+00:00 | 9 | IONOS Cloud price list (US site, USD), Basic and Memory Cubes: hourly price and the page's 30-day price (24 x 30 x hourly); the page shows no VAT status. |
ionos-vpsIonosVps | first-party | www.ionos.com/servers/vps | ok | 2026-10-23T01:10:06+00:00 | 2026-10-23T01:10:06+00:00 | 5 | IONOS US-site VPS Plus plans in USD: regular monthly price as primary, the first-3-months promo (1-year term) in prices; the page says state and local sales taxes may apply (not included). |
kamatera-pricingKamateraPricing | first-party | www.kamatera.com/pricing/ | ok | 2026-10-23T00:54:43+00:00 | 2026-10-23T00:54:43+00:00 | 3 | Kamatera's three example server cards (Basic, Standard, Pro) in USD; real prices come from the configurator, so these are examples, not a price list. The page states no VAT. |
kling-pricingKlingPricing | first-party | kling.ai/dev/pricing | ok | 2026-10-23T11:12:21+00:00 | 2026-10-23T11:12:21+00:00 | 33 | Kling's developer API price page (prepaid packages, USD per second of video or per image; 1 Unit is a stated USD amount that differs for video and image). |
lambda-instancesLambdaInstances | first-party | lambda.ai/instances | ok | 2026-10-23T00:54:52+00:00 | 2026-10-23T00:54:52+00:00 | 14 | Lambda's own instances page, on-demand, USD per GPU-hour excl. VAT: ids without suffix are from the 8-GPU instance table, '-1x' from the single-GPU instance table. |
leaseweb-public-cloudLeasewebPublicCloud | first-party | www.leaseweb.com/en/products-services/cloud… | ok | 2026-10-23T01:10:36+00:00 | 2026-10-23T01:10:36+00:00 | 10 | Leaseweb Public Cloud price tables (General Purpose M3, Compute C3, Memory R3): EUR per hour and per month; the page does not state VAT; storage beyond the 5 GB root disk is billed separately. |
leaseweb-vpsLeasewebVps | first-party | www.leaseweb.com/en/products-services/cloud… | ok | 2026-10-23T01:10:25+00:00 | 2026-10-23T01:10:25+00:00 | 6 | Leaseweb VPS 1-6 plan cards (EUR per month, the page prints 'Excl. VAT'); data-centre choice is configured at order time, so no region is asserted. |
linodeLinode | first-party | www.linode.com/pricing/ | ok | 2026-10-23T22:53:52+00:00 | 2026-10-23T22:53:52+00:00 | 45 | Linode/Akamai /v4/linode/types. price.monthly in USD; region_prices lists the regions that deviate (id-cgk, br-gru); the base price applies to all EU regions. |
litellmLiteLLM | community cross-check | github.com/BerriAI/litellm/blob/main/model_… | ok | 2026-10-23T22:53:51+00:00 | 2026-10-23T22:53:51+00:00 | 3253 | Community cross-check (MIT). Provider = litellm_provider; plan = key without the provider prefix. |
luma-pricingLumaPricing | first-party | lumalabs.ai/pricing | ok | 2026-10-23T16:44:48+00:00 | 2026-10-23T16:44:48+00:00 | 73 | Luma's own price page (Luma and third-party video, image and audio models, all in Luma credits), converted to USD at the Plus plan's monthly list price per credit. |
mistral-pricingMistralPricing | first-party | mistral.ai/pricing/ | ok | 2026-10-23T00:39:18+00:00 | 2026-10-23T00:39:18+00:00 | 1 | Mistral's own pricing page. USD per 1M tokens. |
modelsdevModelsDev | community cross-check | models.dev/ | ok | 2026-10-23T23:39:40+00:00 | 2026-10-23T23:39:40+00:00 | 8488 | Community cross-check (models.dev). Row id = <provider>/<model>; USD per 1M tokens as published, models without cost.input are skipped. |
nebius-pricesNebiusPrices | first-party | nebius.com/prices | ok | 2026-10-23T11:12:21+00:00 | 2026-10-23T11:12:21+00:00 | 7 | Nebius' own price page, on-demand NVIDIA GPU instances, USD per GPU-hour excl. VAT; L40S is the 'from' price of the smallest CPU/RAM configuration. |
netcup-vpsNetcupVps | first-party | www.netcup.com/en/server/vps | ok | 2026-10-23T00:54:30+00:00 | 2026-10-23T00:54:30+00:00 | 5 | netcup VPS plan cards: EUR including 19% VAT (price_ex_vat derived); primary is the one-month price, the 12-month contract price is kept in prices. |
oci-pricelistOciPricelist | first-party | www.oracle.com/cloud/price-list/ | ok | 2026-10-23T22:54:09+00:00 | 2026-10-23T22:54:09+00:00 | 12 | Oracle Cloud price list in EUR (list price, pay as you go). GPU compute meters only, priced per GPU-hour (gpu_count 1); software, Cloud@Customer and CPU-shape-named legacy GPU meters are skipped. |
openai-pricingOpenAIPricing | first-party | developers.openai.com/api/docs/pricing | ok | 2026-10-23T23:24:34+00:00 | 2026-10-23T23:24:34+00:00 | 3 | OpenAI's own pricing page. USD per 1M tokens; long-context tier as *_in_long / *_out_long / *_cached_in_long. |
openrouterOpenRouter | community cross-check | openrouter.ai/aion-labs/aion-2.0 | ok | 2026-10-23T22:53:51+00:00 | 2026-10-23T22:53:51+00:00 | 469 | Prices as charged through OpenRouter (the router's own list price). Vendor = first path segment of the id. |
openrouter-endpointsOpenRouterEndpoints | community cross-check | openrouter.ai/aion-labs/aion-2.0 | ok | 2026-10-23T22:54:10+00:00 | 2026-10-23T22:54:10+00:00 | 1143 | OpenRouter per-endpoint prices: the upstream provider serving a model through OpenRouter, with its quantization. USD per token converted to per 1M tokens. status is OpenRouter's endpoint status (0 = ok). |
ovh-catalog-cloudOvhCloudAddons | first-party | www.ovhcloud.com/en/public-cloud/prices/ | ok | 2026-10-23T23:09:29+00:00 | 2026-10-23T23:09:29+00:00 | 72 | OVHcloud Public Cloud instance flavors (NL subsidiary). Hourly consumption price; the .monthly.postpaid addon gives the monthly price where it exists. The number in a flavor name is its RAM in GB. Micro-cents EUR excl. VAT. |
ovh-catalog-vpsOvhVps | first-party | www.ovhcloud.com/en/vps/ | ok | 2026-10-23T22:44:17+00:00 | 2026-10-23T22:44:17+00:00 | 44 | OVHcloud order catalog, NL subsidiary. Micro-cents EUR excl. VAT; spec from the product description in the raw body. |
paperspace-pricingPaperspacePricing | first-party | www.paperspace.com/pricing | ok | 2026-10-23T16:44:34+00:00 | 2026-10-23T16:44:34+00:00 | 13 | Paperspace's own Core > Hourly 'Dedicated GPUs' cards, USD per single GPU-hour (multi-GPU machines are multiples); H100 is the on-demand promo price from the page footnote, H100 and A100-80G also appear with their 3-year commitment prices. |
replicate-pricingReplicateHardware | first-party | replicate.com/pricing | ok | 2026-10-23T16:44:44+00:00 | 2026-10-23T16:44:44+00:00 | 17 | Replicate's own hardware price table: billed per second, converted to USD per hour (x 3600); multi-GPU machines are one Sku with the price of the whole machine; CPU hardware and per-output model prices are not archived. |
runpod-gpu-typesRunPod | first-party | www.runpod.io/pricing | ok | 2026-10-23T23:39:39+00:00 | 2026-10-23T23:39:39+00:00 | 47 | RunPod GraphQL gpuTypes. USD per hour per single GPU: Secure Cloud price is the list price, Community Cloud (and spot, when listed) are secondary. |
runpod-pricingRunpodPricing | first-party | www.runpod.io/pricing | ok | 2026-10-23T01:10:13+00:00 | 2026-10-23T01:10:13+00:00 | 19 | RunPod's pricing page (Pods table, Secure Cloud as the page shows it by default), USD per GPU-hour; plans are prefixed 'page-' so they never collide with the runpod-gpu-types SKUs. |
runway-pricingRunwayPricing | first-party | runway.com/pricing | ok | 2026-10-23T16:44:41+00:00 | 2026-10-23T16:44:41+00:00 | 4 | Runway's own plan page: generated video priced in credits per clip, converted to USD per second at the Standard plan's monthly list price per credit. |
sambanova-modelsSambanovaModels | first-party | api.sambanova.ai/v1/models | ok | 2026-10-23T18:46:09+00:00 | 2026-10-23T18:46:09+00:00 | 6 | SambaNova Cloud's public /v1/models API: USD per token converted to per 1M tokens, with context length and max completion tokens. |
scaleway-gpuScalewayGpu | first-party | www.scaleway.com/en/gpu-instances/ | ok | 2026-10-23T22:54:08+00:00 | 2026-10-23T22:54:08+00:00 | 11 | Scaleway GPU instance products, zone fr-par-2. EUR excl. VAT. gpu_info carries model and per-GPU memory. |
scaleway-instancesScalewayInstances | first-party | www.scaleway.com/en/pricing/virtual-instances/ | ok | 2026-10-23T22:44:16+00:00 | 2026-10-23T22:44:16+00:00 | 124 | Scaleway instance products, zone fr-par-1. EUR excl. VAT. ram in bytes. |
scaleway-instances-nl-ams-1ScalewayNlAms1 | first-party | www.scaleway.com/en/pricing/virtual-instances/ | ok | 2026-10-23T00:24:07+00:00 | 2026-10-23T00:24:07+00:00 | 115 | Scaleway instance products, zone nl-ams-1 (same API as fr-par-1). EUR excl. VAT. Plans carry the zone suffix. |
scaleway-instances-pl-waw-1ScalewayPlWaw1 | first-party | www.scaleway.com/en/pricing/virtual-instances/ | ok | 2026-10-23T00:24:07+00:00 | 2026-10-23T00:24:07+00:00 | 40 | Scaleway instance products, zone pl-waw-1 (same API as fr-par-1). EUR excl. VAT. Plans carry the zone suffix. |
stability-pricingStabilityPricing | first-party | docs.stability.ai/pricing | ok | 2026-10-23T16:44:55+00:00 | 2026-10-23T16:44:55+00:00 | 6 | Stability AI's own API price page: image generation in credits, converted to USD at the 1 credit = $0.01 the page states. |
together-gpu-clustersTogetherGpuClusters | first-party | www.together.ai/gpu-clusters | ok | 2026-10-23T16:44:37+00:00 | 2026-10-23T16:44:37+00:00 | 7 | Together AI's own GPU Clusters page, USD per GPU-hour: self-serve on-demand clusters and the reserved 'Starting at' price (lowest tier, commitment up to 6 months paid upfront); the same on-demand cluster prices appear on together.ai/pricing (together-pricing). |
together-pricingComposite | first-party | www.together.ai/pricing | ok | 2026-10-23T01:10:06+00:00 | 2026-10-23T01:10:06+00:00 | 16 | Together AI's own pricing page, USD per GPU-hour: Dedicated Inference (single-tenant) and GPU Clusters (on-demand pay as you go); preemptible and reserved prices are not archived. Together AI's own serverless chat-model table, USD per 1M tokens (input and output; cached input not captured). GPU cluster and dedicated rows are read by the Together GPU normalizer. |
transip-vpsTransipVps | first-party | www.transip.nl/vps/ | ok | 2026-10-23T00:54:34+00:00 | 2026-10-23T00:54:34+00:00 | 5 | TransIP VPS V1-V5 'from' prices, EUR excl. VAT (page footer: prices excl. btw), placed in Amsterdam (AMS0) or Alblasserdam (RTM0); dedicated vCPU is a EUR 10 add-on. |
upcloud-pricingUpcloudPricing | first-party | upcloud.com/pricing | ok | 2026-10-23T00:54:45+00:00 | 2026-10-23T00:54:45+00:00 | 9 | UpCloud Starter plans in EUR excl. VAT ('Prices exclude any applicable local taxes'), billed by the starting hour with a monthly maximum of 28 days. |
vast-asksVastAsks | marketplace asks, not list prices | vast.ai/pricing | ok | 2026-10-23T18:16:06+00:00 | 2026-10-23T18:16:06+00:00 | 9 | Vast.ai marketplace: median ask per GPU model (rentable offers with one GPU), USD per hour. min / p10 / p90 / count describe the same asks. A market distribution, never a list price. |
vultrVultr | first-party | www.vultr.com/pricing/ | ok | 2026-10-23T22:44:16+00:00 | 2026-10-23T22:44:16+00:00 | 139 | Vultr /v2/plans. monthly_cost in USD, excl. tax. locations = data-centre codes. |
vultr-metalVultrMetal | first-party | www.vultr.com/products/bare-metal/ | ok | 2026-10-23T22:54:08+00:00 | 2026-10-23T22:54:08+00:00 | 23 | Vultr bare metal plans (/v2/plans-metal). USD excl. tax. GPU servers: price per server per hour, per-GPU price derived; other servers: price per month. |
xai-modelsXaiModels | first-party | docs.x.ai/docs/models | ok | 2026-10-23T00:39:20+00:00 | 2026-10-23T00:39:20+00:00 | 1 | xAI's own models page. USD per 1M tokens. |
Cross-checks and gaps
- Community cross-checks (LiteLLM, OpenRouter, models.dev, gpuhunt) are labelled as such wherever their numbers appear. They are never presented as the provider's own price.
- Text-only archives: some provider pages are archived and diffed as text but not yet parsed into prices; they show no price here.
- Blocked: Contabo refuses requests from the ledger's server, so it is a gap.
- Marketplaces (Vast.ai, capped at 64 offers per query) are a separate series and never mixed with list prices.
The full rules are on the methodology page. Machine-readable: /v1/sources.