DeepInfra
Low-cost inference API for open-source models with simple per-token pricing.
why this verdict
Keep it for the workflow, not the AI
OpenRouter covers the headline feature free, and the workflow around it is why you would still open this.
- A named launch, not a vibe named, unlinked
- OpenRouter aggregating every inference provider and competing them down on price. OpenAI, September 25, 2025 — no announcement link recorded yet.
- How much of the job it covers headline only editorial call
- The headline feature, but not the workflow, data or integrations around it.
- Is there a free way to do it? yes
- 1 of 3 listed replacements have a usable free tier: OpenRouter.
- What the call is worth nothing to cancel
- No paid entry tier tracked, so there is no subscription to cancel.
- Threatened by
- OpenAI
- Since
- September 25, 2025
- List price
- free
- Per year
- —
The backstory
DeepInfra serves open models at some of the lowest per-token prices available, with an OpenAI-compatible endpoint that makes switching trivial. Trivial switching is precisely the problem. When OpenRouter lists every provider side by side and routes to whoever is cheapest this minute, serving open weights becomes an undifferentiated commodity with margins to match. DeepInfra competes credibly on price and breadth, but price is the only axis left and it belongs to whoever is willing to earn least.
Escape hatches
Broader model catalogue and fine-tuning support
together.ai open_in_newStrong performance tuning for production serving
fireworks.ai open_in_newRoutes across providers so you never pick one
openrouter.ai open_in_new1 of 3 replacements have a usable free tier.
