Tight budget, worried about surprise costs?

Spending is always capped. AI Services charges against your prepaid balance in Cloudj, so you never spend more than you deposit. You can also set spending limits per API Key and account-wide, with optional email alerts when thresholds are reached.

Is response speed fast enough for real-time use?

Measure it yourself. The list includes high-speed Flash models for code suggestions, real-time translation, and agent loops. Try them in the Playground, then check the duration and TTFT of each request in Logs.

How are my prompts and outputs handled?

We do not train on them. Cloudj does not use your prompts or outputs to train models and never sells them. Logs only records request metrics (model, Tokens, cost, duration)—never the prompt or response content.

Want to run workloads on your own server?

Use cloud servers from Cloudj. Launch a Cloudj cloud server, install your app or agent, and call the API with your API Key. Preconfigured runtime images are on the roadmap.

Open-source AI model API: cut your AI bill by 80%+ vs closed models

Q: Will I get the model I asked for?

Yes, that exact model. Each request goes to the open-source model you choose — no silent swaps.

Q: How do I switch my app over?

Three steps. Fund your wallet, create an API Key, and update the Base URL in your app—pay only for the Tokens you use.

One API key, every model on the list — Open-source LLM inference for chat, code, reasoning, long context and images. Pay only for the tokens you use, capped by your prepaid balance.

Get my API key See pricing

Why pay closed-model prices for every task?

Leading open-source models are strong enough to handle most of the work your team sends them — often at a fraction of the cost of closed models.

Game Prototype

Prototype a two-player Snake game

Large Refactor

Port and clean up a large codebase

Bulk Translation

Translate 100 research papers into English

Image Tasks

Write alt text and descriptions for 1,000 product photos

Ticket Triage

Tag 1,000 support tickets and draft replies

Examples: same token volume for both sides × official list price per 1M tokens of the models named (checked 2026-10-29). Newer closed models may cost less; batch discounts are not included.

Create my API key

AI model API pricing per 1M tokens

How AI API pricing works

Input Price: Billed per 1M tokens
Cache Hit: Billed per 1M tokens
Output Price: Billed per 1M tokens

Charged against your prepaid balance — you never spend more than you deposit. Same open models, at a lower price.

Other providers: official price pages checked 2026-10-29 (select a name to open the source). Peak = peak-hour list price; DeepSeek and Alibaba Cloud charge 50% less off-peak, below Cloudj prices at those hours. — = no published cache price. Alibaba Cloud and Tencent Cloud: Singapore region. Product names are trademarks of their owners; Cloudj is not affiliated with them.

Open-source LLMs on one API. Test each on your own prompts.

Benchmarks are a starting point. Your own results matter more. One API key works for every model listed here, and you pay only for the tokens you use. Try chat, code, reasoning, long context and images yourself.