Inference.net

trial credit · inference_host · verification: partially_verified (60%, doc_fetch, 2026-07-17)

The record

Base URLs	https://api.inference.net/v1 openai-compatible
API key required	yes
Credit card	unknown
Phone verification	unknown
May train on prompts	unknown
Quota	trial_credit [credits, rpm, fair_use] “cheahjs README: 'Credits: $1, $25 on responding to email survey'. Free-LLM quota table: 'Varies \| Fair use \| Free for listed models \| Fair use policy \| DeepSeek-R1, Llama 3.1 8B/70B'. Official pricing page (Pay as you go plan): '30 req/min rate limit', '1M included Gateway requests'.”
Signup friction	steps unknown
Env var	INFERENCE_NET_API_KEY
Notes	Decentralized/distributed GPU network serving open models (per Free-LLM). Base URL https://api.inference.net/v1 confirmed from Free-LLM's code example (OpenAI SDK client); official docs landing page fetched 2026-07-17 did not restate it. The '1M included Gateway requests' line on the pricing page appears to describe their gateway/tracing product rather than raw inference quota.

model	capabilities	context	free	limits
deepseek-ai/DeepSeek-R1	chat, reasoning	?	yes	Free-LLM: 'Varies \| Fair use \| Free for listed models \| Fair use policy \| DeepSeek-R1, Llama 3.1 8B/70B'. cheahjs: 'Credits: $1, $25 on responding to email survey'. Model id taken verbatim from Free-LLM's inference-net code example (openai client, base_url https://api.inference.net/v1).
meta-llama/Llama-3.1-70B-Instruct	chat	?	yes	Free-LLM: 'Varies \| Fair use \| Free for listed models \| Fair use policy \| DeepSeek-R1, Llama 3.1 8B/70B'. Free-LLM gives the display name only ('Llama 3.1 70B'); exact API model id not individually verified.
meta-llama/Llama-3.1-8B-Instruct	chat	?	yes	Free-LLM: 'Varies \| Fair use \| Free for listed models \| Fair use policy \| DeepSeek-R1, Llama 3.1 8B/70B'. Free-LLM gives the display name only ('Llama 3.1 8B'); exact API model id not individually verified.