How to get higher rate limits from AI providers

What decides your rate limits?

Mostly money and time. How much you’ve paid a provider, and for how long.

Each provider puts your account in a tier, and the tier sets per-model caps on requests and tokens per minute. Some add a monthly spend cap on top. The caps belong to the organization or project, not the API key, so a second key won’t get you anywhere. Go over one and you get a 429 back. It usually comes with headers saying how much room is left, and a retry-after telling you how many seconds to wait.

How do OpenAI usage tiers work?

There’s a free tier and three paid ones. You move up on your own as your total credit purchases grow.

Tier	Total credit purchases	Monthly usage limit
Free	Allowed geography	$100
Build	$5	$500
Launch	$100	$5,000
Grow	$500	$200,000

Limits differ by model, and a few models share one pool. OpenAI counts requests and tokens per minute and per day, and whichever runs out first is the one that stops you. If you’d rather not wait for the automatic move, there’s an Upgrade tier button in the Usage Tiers section. Big enterprise accounts that keep hitting the ceiling get pointed to Scale Tier, or to Reserved Tier for GPT-5.6 and newer. The steps are in how to raise your OpenAI rate limits.

How do Anthropic usage tiers work?

Anthropic doesn’t publish deposit thresholds. It places you based on usage history and account standing, and brand new accounts can start in a lower Evaluation tier until they’ve built some history.

Tier	Monthly spend cap	Claude Opus 5.5 RPM, input TPM, output TPM
Start	$500	4,000, 2M, 400K
Build	$1,000	10,000, 5M, 1M
Scale	$200,000	13,000, 10M, 2M
Custom	None	Set with sales

To move up, open the Rate limits page in the Claude Console and click Request tier increase.

Watch the spend cap more closely than the rate limits. Hit it and the API stops until the first of next month unless you get a higher limit sooner, and that 429 comes with no retry-after. Sudden traffic spikes can also trip what Anthropic calls acceleration limits, so ramp up slowly when you launch something. More on both in how to raise your Anthropic rate limits.

What about Gemini, Azure OpenAI and Bedrock?

Google’s Gemini API is the most explicit of the three. A linked billing account gets you Tier 1. Tier 2 needs $100 paid and 3 days since your first payment, and Tier 3 needs $1,000 and 30 days. Past that there’s a request form, with no promises attached.

Azure OpenAI has seven quota tiers. It moves subscriptions up automatically based on usage, payment history and whether you have an Enterprise Agreement, and its quota request form adds quota without changing your tier.

On Amazon Bedrock you ask through the Service Quotas console, per model and per Region. AWS says it gives priority to accounts already using most of their quota.

The Gemini, Azure and Bedrock details come from each provider’s quota docs, linked under Sources and checked October 30, 2026. Each one has its own guide too: Gemini, Azure OpenAI and Bedrock.

Can you get more out of the limits you have?

On Claude, start with prompt caching. Cache reads don’t count toward input tokens per minute on most models. Anthropic’s own example is a 2M limit with an 80% cache hit rate handling roughly 10M input tokens a minute.

Batch jobs help on both big providers. Anthropic’s Message Batches API has its own separate limits, and OpenAI caps batch work by the input tokens queued for each model. If several teams share one Anthropic organization, give each workspace its own cap so one runaway job can’t eat everyone’s quota.

When you do ask for more, say exactly what you need. Anthropic’s checklist for Claude Platform on AWS is a good one to borrow. It asks for the models, your peak input and output tokens per minute for each, and roughly how much of your input is cached.

Changelog

• October 30, 2026. First version. OpenAI’s paid tiers now go by Build, Launch and Grow after its 2026 rename, so older tier numbers won’t match. Anthropic no longer sells new Priority Tier commitments, so it isn’t covered here.