AI infrastructure blog
The FastInfra blog publishes practical guides on AI infrastructure: how LLM API pricing works and how to cut it, how to integrate an OpenAI-compatible API, how to size GPU pods for a model, when dedicated servers beat cloud VMs, and how to run video and speech generation through an API. Every post draws on FastInfra's live catalog data and is reviewed against the current prices before publication.
11 published guides · Written by FastInfra · Figures update from the live catalog · RSS feed
LLM API pricing per million tokens explained
LLM API pricing per million tokens explained: input vs output costs, median rates, and how to estimate spend across 785 models. LLM API pricing per.
Text-to-video API: the async job pattern, polling, and keeping generation cost under control
Text-to-video API guide: why generation is an async job returning 202, how to poll for the MP4, what drives cost and latency, and a production checklist.
Server Auction: how to get dedicated bare metal for less, and what to check before you buy
Server auction explained: how FastInfra lists discounted dedicated servers with no setup fee, what the hardware is, how prices move, and how to pick one.
Dedicated server vs cloud VPS: which one fits the workload, and what each really costs
Dedicated server vs cloud VPS compared on cost, consistency, provisioning and scaling, with guidance on where AI and database workloads belong.
Cloud VPS for AI backends: sizing, setup and cost for the server that sits in front of the model
Cloud VPS for AI backends: how to size vCPU and RAM for an API that calls LLMs, what per-minute billing costs, and a hardened Ubuntu setup to deploy today.
Per-minute GPU billing: what a rented GPU pod really costs and how to keep idle time near zero
Per-minute GPU billing explained: how hourly rates become monthly cost, where idle time hides, stop vs terminate, and a break-even against per-token APIs.
How much VRAM do you need to run an LLM? A sizing guide with the actual arithmetic
How much VRAM to run an LLM: weights at fp16 vs int4, the KV cache per token, batch and context overhead, a per-model-size table and which GPU pod fits.
Streaming chat completions: handling server-sent events that survive production
Streaming chat completions over server-sent events: how delta chunks arrive, how to parse them, and the timeout, retry and cancellation rules that hold up.
Free LLM API models: what you get at $0, where the limits are, and when to pay
A free LLM API tier is real on FastInfra: open models at $0 through the OpenAI-compatible endpoint. What the free tier includes, its limits, when to pay.
LLM API pricing per million tokens: how to estimate what a feature will cost
LLM API pricing per million tokens explained: input vs output rates, how to count tokens, a worked cost-per-request formula and live FastInfra prices.
How to move an OpenAI SDK app to an OpenAI-compatible API in one change
Switch an existing OpenAI SDK app to FastInfra's OpenAI-compatible API by changing base_url and the key. Covers model ids, streaming and what differs.