These offerings are not all on one meter
This model is metered accelerator instance hour, token. Rates on different meters cannot be read across as cheaper or dearer. The values are still shown, because they are facts; what is withheld is the implication that they are rankable.
Offerings, side by side
Each column is one provider’s commitment to serve this model on its own terms. The reference is a deterministic path into the data — not a recommendation, and not the cheapest.
| Term | Oryventa ModelsHosted Open ModelReference offering |
|---|---|
| Channel | Hosted Open Model |
| Status | Active |
| Disclosure | Public List |
| Provider's name for it | Not recorded |
| Processing tiers | Standard, Economy |
| Capacity | Dedicated Compute, Shared On Demand |
| Commitment | Term Commitment — the committed amounts are not recorded, only the discount they carry |
| Separately-priced services | Tuning Service |
| Effective from | 1 Jan 2026 |
| Captured as of | 31 Aug 2026 |
| Rates recorded | 4 across 2 tiers — 1 tier's cost is not recorded, which is not the same as matching the default |
Who else sells something like this
3 external offerings from 3 providers, captured as of 2026-09-14. Each offering’s own page names the surface every one is sold on and links to the provider’s pricing. No price is shown on either: most of these providers do not publish one per unit.
Google · 1 offeringMicrosoft · 1 offeringNASA/IBM · 1 offering
This model’s pricing versions
Every recorded version of every offering’s pricing policy, with its amounts. A scheduled version is a published future price, not a forecast.
- Oryventa ModelsIn forceEffective from 1 Jan 2026 — no end date recordedCaptured as of 31 Aug 2026
- Input tokens $1.904 per 1M tokens
- Output tokens $4.08 per 1M tokens
The earliest recorded version, so there is nothing before it to compare with.