Prices tracked
15
14 confirmed on the provider's feed or page at the last check
Sources
13
Every cited page, listed below
Market
Baseten is privately held, so there is no public market data to show.
Against the market
How Baseten prices compare
Baseten's lowest price for each GPU beside the lowest price every other tracked provider lists for the same GPU, sold the same way. Typical is the middle provider, one price each.
NVIDIA H100On demand
All NVIDIA H100 prices →Baseten$6.50per GPU-hour, on-demand
Rank18 of 2153% above typical
Typical$4.25middle of 21 providers
Lowest anywhere$1.99Voltage Park
Voltage Park1. H100 (single GPU, no contract)
$1.99
GMI Cloud2. NVIDIA H100
$2.00
Vultr3. NVIDIA HGX H100 — Cloud GPU
$2.30
Denvr Dataworks4. NVIDIA H100 SXM (8 GPUs, 208 vCPU)
$2.45
Cirrascale5. 8X NVIDIA H100 (16-Node+ Clustered)
$2.49
Massed Compute6. H100 (80 GB) ×1/×2/×4
$2.73
Runpod7. H100 PCIe / H100 NVL — Pods
$2.89
Hyperbolic8. H100 SXM
$3.19
Lambda9. NVIDIA H100 PCIe
$3.29
Together AI10. H100 (80 GB), 8–256 GPUs
$3.99
Baseten18. H100 (80 GiB) dedicated deployment / training
$6.50
NVIDIA B200On demand
All NVIDIA B200 prices →Baseten$9.98per GPU-hour, on-demand
Rank13 of 1623% above typical
Typical$8.12middle of 16 providers
Lowest anywhere$3.50Vultr
Vultr1. NVIDIA HGX B200 — Cloud GPU
$3.50
GMI Cloud2. NVIDIA B200
$4.00
Massed Compute3. B200 SXM6 ×8
$5.43
Hyperbolic4. B200
$5.99
Cirrascale5. 8X B200
$5.99
Lambda6. NVIDIA B200 SXM6
$6.69
Runpod7. B200 (180 GB, 28 vCPU) — Pods
$7.99
Google Cloud8. a4-highgpu-8g (A4)
$8.06
Together AI9. HGX B200 (180 GB)
$8.19
Verda10. Instant clusters, B200 SXM6 180GB
$8.44
Baseten13. B200 (180 GiB)
$9.98
Tracked prices
Baseten prices
Every price the Index tracks for Baseten, in the provider's own words beside a normalized figure, with the date it was last checked. What AI Costs compares them across the market.
| GPU | Offer | Terms | Published price | Per GPU-hour | Change | Checked | Source |
|---|---|---|---|---|---|---|---|
| NVIDIA H100 | H100 (80 GiB) dedicated deployment / training | on-demandList | $0.10833 per minute, H100 80 GiB VRAM, Dedicated Deployments | $6.50 | No change | On pageOct 26, 2026 | baseten.co ↗ |
| NVIDIA B200 | B200 (180 GiB) | on-demandList | $0.16633 per minute, B200 180 GiB VRAM, Dedicated Deployments | $9.98 | No change | On pageOct 26, 2026 | baseten.co ↗ |
| Seller | Model | Input | Cached input | Output | Revenue per GPU-hour at list | Other published prices | Change | Checked |
|---|---|---|---|---|---|---|---|---|
| Baseten | Kimi-K3Model by moonshotai | $3.00 | n.a. | $15.00 | n.a. | None published | No change | FeedOct 27, 2026huggingface.co ↗ |
| Baseten | GLM-5.2Model by zai-org | $1.40 | n.a. | $4.40 | n.a. | None published | No change | FeedOct 27, 2026huggingface.co ↗ |
| Baseten | GLM-5.3Model by zai-org | $1.40 | n.a. | $4.40 | n.a. | None published | No change | FeedOct 27, 2026huggingface.co ↗ |
| Baseten | Muse-Spark-Open-816B-A42BModel by meta-models | $1.25 | n.a. | $4.25 | n.a. | None published | New · Oct 16 | FeedOct 27, 2026huggingface.co ↗ |
| Baseten | DeepSeek-V4-Pro-0813Model by deepseek-ai | $1.32 | n.a. | $3.96 | n.a. | None published | No change | FeedOct 27, 2026huggingface.co ↗ |
| Baseten | Beam-1.0-501B-A23BModel by reflection | $0.40 | n.a. | $1.20 | n.a. | None published | −72.1% | FeedOct 27, 2026huggingface.co ↗ |
| Baseten | DeepSeek-V4.1-FlashModel by deepseek-ai | $0.30 | n.a. | $1.20 | n.a. | None published | No change | FeedOct 27, 2026huggingface.co ↗ |
| Baseten | DeepSeek-V4.1-Flash-1017Model by deepseek-ai | $0.30 | n.a. | $1.20 | n.a. | None published | New · Oct 20 | FeedOct 26, 2026huggingface.co ↗ |
| Baseten | GLM-5.3-FlashModel by zai-org | $0.15 | n.a. | $0.50 | n.a. | None published | No change | FeedOct 27, 2026huggingface.co ↗ |
| Baseten | gpt-oss-120B | $0.10 | n.a. | $0.50 | n.a. | None published | No change | ResearchOct 21, 2026baseten.co ↗ |
| Baseten | gpt-oss-120bModel by openai | $0.10 | n.a. | $0.50 | n.a. | None published | No change | FeedOct 27, 2026huggingface.co ↗ |
| Baseten | NVIDIA-Nemotron-3.5-Super-VL-120B-A12B-NVFP4Model by nvidia | $0.20 | n.a. | $0.50 | n.a. | None published | New · Oct 16 | FeedOct 27, 2026huggingface.co ↗ |
| Baseten | DeepSeek-V4-Flash-0731Model by deepseek-ai | $0.13 | n.a. | $0.26 | n.a. | None published | No change | FeedOct 27, 2026huggingface.co ↗ |
Revenue per GPU-hour is SAVRN's calculation, not a price any seller charges: the 6.83 million input and 0.85 million output tokens one Vera Rubin GPU handles in an hour in CoreWeave's measured test, billed at this model's list prices. It puts a token price beside a GPU rental price; it says nothing about how fast this model actually runs.
Sources
Sources for Baseten
Every source cited on this page, once each.
- baseten.cobaseten.co
- huggingface.cohuggingface.co
- huggingface.cohuggingface.co
- huggingface.cohuggingface.co
- huggingface.cohuggingface.co
- huggingface.cohuggingface.co
- huggingface.cohuggingface.co
- huggingface.cohuggingface.co
- huggingface.cohuggingface.co
- huggingface.cohuggingface.co
- huggingface.cohuggingface.co
- huggingface.cohuggingface.co
- huggingface.cohuggingface.co
GPU cloud
More in the Index
Astra IntellectusCirrascaleDellDenvr DataworksDigitalOceanFireworks AI (dedicated)GMI CloudHPEHyperbolicImpossible CloudLightning AIMassed ComputeRunpodTensorWaveTheAI CloudVast.ai (marketplace medians)Verda
Capital figures come from the Capital Atlas, refreshed every Tuesday, Thursday and Saturday. Site commitments overlap reported capital expenditure and overlap one another, so each figure appears as its source states it and none are added into a single deployment number. Informational, not investment advice.
Sources
The SAVRN reference set
Seven references, one system
Capital, delay, policy, water, grid, permits and scarcity are one system seen from seven sides. Each reference is primary-sourced, refreshed on a schedule, and links into the others, so a project can be followed from the money committed to the docket that stopped it.
Capital Atlas
Who is funding AI data centers: capital providers, named sites, financing transactions, and the blocked-and-delayed record, state by state and company by company.
Open the atlas →
Delay Watchlist
Project-level slippage: named projects, jurisdictional actions, utility and federal dockets, SEC disclosures, and the interconnection queue.
Open the watchlist →
Moratorium Tracker
The policy layer: every state measure and local moratorium mapped and cited, with one page per state.
Open the tracker →
Water Tracker
Every federal, state, local and interstate water measure that touches a data center, plus a fifty-state water ledger.
Open the water tracker →
Grid Operator Watchlist
ERCOT, PJM, MISO, CAISO, SPP, NYISO, ISO-NE and FERC: the large-load rules, filings and queue positions that decide whether a site can energize.
Open the watchlist →
Permits and Power Development
Data center and substation permits, large loads in the interconnection queues, and the power plants planned to serve them, refreshed hourly from public records.
Open the permits tracker →
SAVRN Scarcity Tracker
Eleven primary-sourced indicators of AI infrastructure scarcity, one composite index, read against the token index, SAVRN's own cost basis and the frontier ceiling.
Open the scarcity tracker →
Have a specific address? Site Intelligence reads all seven references against one location, county, utility and grid operator included, and cites every finding.
Check an address →