Every provider we add, every model that goes free (or stops being free) — logged automatically, no spin.
An Indonesian OpenAI-compatible gateway with a daily free allowance on selected open models, separate from its prepaid models. Free inference runs on Cloudflare Workers AI. A verified email is required, and free requests never spend prepaid
A Turkish OpenAI-compatible gateway whose terms say free chat models "can be called on a new account with zero balance before any top-up". Paid use is prepaid, with an 8% margin taken at top-up. The keyless /v1/models catalog shows these mo
SenseTime's SenseNova Token Plan has a "Free · 公测" (public beta) plan at ¥0/month. The official page says "公测期完全免费开放,付费档位即将上线" (fully free during the public beta, paid tiers coming soon). It covers the native multimodal SenseNova 6.8 Flash
Voyage AI gives each account the first 200M tokens free on each of its current embedding and reranking models (50M on finance/law/multilingual models), plus 150B pixels on multimodal. NOTE: embeddings and rerankers only, not chat. The free
Jina's search foundation APIs (embeddings, reranker, Reader) give every new visitor an auto-generated API key with free tokens usable on any Jina model. NOTE: these are embedding, rerank and reader models, not chat LLMs.
Express mode gives new Google Cloud users an API key for Vertex AI Gemini models, with "no billing information" needed, for up to 90 days within free-tier quotas. Billing can be added later to raise quotas. It uses the global endpoint, with
A low-cost inference API for open models, aimed at long-running agents, with optional "completion windows" (balanced/flex) that cut prices by trading latency. The Free plan gives "$5/month in free credits when you attach a payment method".
An open-source AI gateway (0% markup) whose Free plan includes a recurring "Free core" pool covering GPT-6 Luna, DeepSeek V4.1 Flash and Clef. The pool allows 10 RPM, 250K TPM and 1M tokens/day. It unlocks only after a temporary $1 card ver
An EU-based OpenAI-compatible chat API whose zero-price "free" tier models work without credit. The rate-limit endpoint reports free-model hourly and daily request and token counters separately from the general per-minute bucket.
An OpenAI-compatible inference platform with a large role-play and fine-tune catalog. The free Tier 1 needs only registration ("no credit card required") and gives daily request quotas on free models under 100B parameters. The keyless /v1/m
Gitee's model platform (formerly Gitee AI) serves 200+ open models through OpenAI- and Anthropic-compatible APIs. The official FAQ says that with the "免费体验访问令牌" (free experience access token) "每位用户每日拥有 100 次免费调用次数" (each user gets 100 free
A serverless open-model API priced per token. The official docs say "Free models do not require prepaid credit", while paid models return 402 on an empty balance. One zero-priced Qwen3.5 4B route is currently listed.
A serverless API covering tens of thousands of open models plus closed providers, with one credit currency. The Free plan grants $1 of credit every 4 weeks, with "Billing: None". It runs open models up to 7B parameters, one request at a tim
IBM's free tier ("Toolbox playground" / watsonx.ai Runtime Lite) includes up to 300,000 foundation-model tokens and 20 Capacity Unit Hours per month. Inference goes through the watsonx REST API with an IBM Cloud IAM token and a project or s
Zhipu's mainland-China platform (separate from z.ai international) marks a set of Flash models as "免费" (free) in its model overview, including GLM-4.7-Flash with a 200K context. It also offers free CogView-3-Flash image and CogVideoX-Flash
iFlytek's Spark Lite is a lightweight Chinese LLM that the official HTTP API docs describe as "支持免费使用" (free to use). The endpoint works with the OpenAI SDK, using the APIPassword from the console as the Bearer key.
Tencent Cloud's model marketplace grants new users a one-time trial of 1 million tokens on each language model, until 2026-12-31. You claim it per model in the Model Gallery, or it is claimed automatically on the first API call. The intl do
An OpenAI-compatible gateway for vision, OCR and language models. Calls work anonymously (no Authorization header, or "Bearer vlmrun") within per-IP limits. A free account adds 100 credits (about $1) and higher rate limits. Every call is st
HPC-AI.com, the company behind Colossal-AI, runs pay-per-token Model APIs for open models. Its launch post offers "$2 in free credits to all users", claimed as a voucher in the Models Console after registration.
Sarvam AI's API platform covers Indic-language LLM chat, speech and translation. New accounts get INR 100 of free credits at signup. Sarvam 105B costs INR 15 per 1M input tokens and INR 60 per 1M output tokens, so the credit buys several mi
An OpenAI-compatible inference service for open models and their fine-tunes, on flat monthly plans. The $0 Free plan gives access only to Qwen3.5-27B-Derestricted, one request at a time, with at most 12K context tokens.
A free, open-source, crowdsourced inference network where volunteers run the models. The oai.aihorde.net front puts an OpenAI-compatible chat completions API on top of the Horde's text workers. Requests without a key use the anonymous key 0
Agnes AI serves its own text, image and video models behind an OpenAI-compatible API, and the official site calls itself a "free multimodal AI platform". The team's HN launch post says "It is completely free. No need to submit any kind of p
A hosted OpenAI-compatible API with a no-card Free plan. You sign in with Google or GitHub, and the Free plan includes three API keys, the AUTO router and "baseline rate limits". The Pro plan costs EUR 29.99/month and raises resource limits
An OpenAI-compatible model gateway that lists about 35 zero-priced "-free" routes next to its paid catalog. Each free route's description sets per-account limits, usually 5 RPM, 100 requests/day and 1M tokens/day. AIHubMix also has an Anthr
Report a model that's no longer free directly from its provider page — it shows up here the moment we confirm it.
Browse the Directory