Every provider we cover, rated by our editors on free credits, free tier, pricing, billing safety and support. Cloud platforms sell compute and storage; AI APIs sell model inference by the token; two do both.
4.3/5
Cloud platform · since 2008 New Google Cloud customers get a $300 Welcome credit that expires 90 days after signup, plus an Always Free tier with one e2-micro VM, 5 GB of Cloud Storage and 1 TiB of BigQuery queries a month that never ends. Funded startups can apply for up to $200,000 over two years, or $350,000 if Google judges them AI-first. You’ll need a card, but nothing is charged until you activate.
- $300 Welcome credit, and nothing charged until you activate
- Always Free tier with a VM, storage and 1 TiB of BigQuery a month
- Scale tier pays 100% of year-one usage up to $250,000
4.1/5
AI API · since 2023 The Gemini API has a free tier that needs a Google account and no card. Gemini 3.8 Flash, 3.6 Flash, 3.5 Flash-Lite and the embedding models cost nothing on it. Paid usage moves to prepaid or postpaid tiers with a $250 cap at Tier 1. Free tier prompts may be used to improve Google products; paid tier prompts aren’t, which is the real price of free.
- No credit card for Flash and Flash-Lite on the free tier
- Gemini 3.1 Flash-Lite at $0.25 per million input tokens
- Batch API takes 50% off standard prices
4.0/5
AI API · since 2020 OpenAI’s API is prepaid. You load credits from $5 and the Build tier opens on that first purchase. There’s no free trial credit for new accounts; the only recurring free credit is up to $500 a month for ChatGPT Pro and Business subscribers. The Batch API halves token prices and cached input costs 10% of list, which is where the real savings are.
- Up to $500 a month in API credits with ChatGPT Pro or Business
- Batch API and Flex processing cut token prices by 50%
- Cached input bills at 10% of list on GPT-5.6 and later
4.2/5
AI API · since 2023 Anthropic sells Claude through prepaid credits on the Claude Console, which now lives at platform.claude.com. New users get a small free credit with no published amount, so don’t plan around it. Claude Max, Ultra and Team plans add $100 to $500 a month in API credits, batches run at half price, and cache reads cost 5% to 10% of input.
- Max, Ultra and Team plans return $100 to $500 a month in API credits
- Cache reads cost 5% of input on Opus 5.5 and Sonnet 5.5
- 50% batch discount stacks with prompt caching
4.2/5
Cloud + AI API · since 2006 AWS gives new accounts $100 in credits at sign-up and up to $100 more for five console activities, good for six months on the Free plan. Startups can apply for Activate credits from $1,000 to $200,000 with an investor Org ID. Amazon Bedrock bills per token for Claude, Nova, Llama and other models, with batch inference at half the on-demand rate, and the same credits cover all of it.
- $100 credit on sign-up, no payment method for most customers
- Free plan can’t incur charges, so there’s no bill shock
- Activate Portfolio credits reach $200,000 and cover Bedrock models
4.0/5
Cloud + AI API · since 2010 A new Azure account gets $200 to spend in its first 30 days, plus free monthly amounts of 20+ services for a year. Microsoft for Startups opens at $200 and climbs to $150,000 as Microsoft verifies real usage. Azure OpenAI models bill per token through Microsoft Foundry, where GPT-4.1 costs $2 per million input tokens on a global deployment, so the $200 goes a long way.
- $200 credit plus 12 months of free services on a new account
- OpenAI and Anthropic models in one Foundry catalog on one bill
- Startup credits grow from $200 to $150,000 with no investor needed
3.4/5
AI API · since 2024 xAI sells Grok models through a prepaid API at console.x.ai. Grok 4.7 costs $2 per million input tokens and $6 per million output, and grok-4.3 costs $1.25 and $2.50, which means $10 of credit covers 5 million input tokens on the flagship. There’s no free tier and no sign-up credit today. The savings that exist are the 20% Batch API discount and cached input at $0.20 to $0.50.
- Grok 4.7 output tokens at $6 per million, below most frontier rivals
- grok-4.3 gives you a 1M context window at $1.25 input
- 20% Batch API discount on every Grok 4.x text model
4.2/5
AI API · since 2023 OpenRouter resells inference through one API key and passes each provider’s list price through without markup. Its cut comes when you buy credits, 5.5% on the Standard plan and 8% on Business, so $100 of credits costs $105.50. Free model variants are capped at 50 requests a day until you’ve bought $10 of credits, after which the cap jumps to 1,000, and a startup program offers up to $5,000.
- Provider list prices with no markup on inference
- Free model variants at 1,000 requests a day after $10 purchased
- Batch API at roughly 50% of per-token price across 70+ models
4.0/5
Cloud platform · since 2011 DigitalOcean gives every new team an automatic $5 signup credit that expires after 90 days, which covers about a month of the cheapest Droplet. Its referral program pays the referrer $25 once the new account spends $25. The Startups program (formerly Hatch) offers 12 months of credits but has paused applications until early 2027. GPU Droplets start at $0.76 per GPU hour and serverless inference is prepaid only.
- $4 Droplets billed per second with a 60-second minimum
- 500 GiB of outbound transfer bundled with the smallest Droplet
- GPU Droplets from $0.76 per GPU hour with a 5-minute minimum
4.1/5
AI API · since 2016 Groq runs open models on its own LPU hardware and lets anyone call the API on a Free plan with no card, capped at 200,000 tokens and 1,000 requests a day on gpt-oss-120b. The Developer plan bills in arrears at $0.15 input and $0.60 output per million tokens on that model, so a million-token day costs under a dollar. Batch and cached input are both 50% off. No trial credit, no startup program.
- Free plan needs no card and resets every day
- gpt-oss-120b at $0.15 input and $0.60 output per million
- Batch jobs priced 50% below synchronous calls