- vendor
- ScaledThought
- whatItIs
- Hosted API for newer open-weight models behind one OpenAI-compatible endpoint, with batch inference and fine-tuning jobs, and an agent-driven self-serve signup flow (a human approves each credit purchase through a payment link).
- hosting
- managed-only
- pricingModel
- per-token (rates per model listed per 1M tokens), billed from prepaid credits (top-ups $20 to $2,000); batch jobs at 50% of normal price with a 24-hour window
- modelProviders
- open-weight (newer open-weight models; specific list not verified)
- stateModel
- stateless
- toolModel
- OpenAI-compatible API
- longRunning
- per-request + batch jobs
- observability
- status page (scaledthought.com/status)
- maturity
- preview
- notes
- Page states prompts and completions are never stored or logged and are not used for training (except fine-tuning jobs the user runs). US region open; Canada region listed as coming soon. The page carries no GA/beta label and no launch date, so maturity 'preview' is inferred from the absence of any track record and the unreleased second region, not a vendor label. Closest tracked entries are Together AI, Fireworks, Featherless and SiliconFlow; the differentiators are agent-operated onboarding, zero data retention and regional handling, which is a thin distinction, so treat as a lower-confidence me-too inference host.
