Why x-ratelimit-limit-tokens-per-day too little?

Question

I am testing the Kimi K3 model (and also verified the same behavior on K2.6) via Serverless Inference. After only two rounds of conversation, my requests started getting rejected. Upon inspecting the response headers, I discovered that my account is hitting a daily token limit:

x-ratelimit-limit-tokens-per-day: 50000
x-ratelimit-remaining-tokens-per-day: 0 (after minimal usage)

The Problem: A 50,000 token per day limit is extremely restrictive and makes meaningful testing or development work practically impossible. A single moderately complex prompt with context can consume 30,000–40,000 tokens, meaning I can barely complete one task per day before being locked out. My Questions:

What is the specific reason for this 50,000 daily token cap? Is this a Tier 1 restriction, a new account safeguard, or related to my current prepayment balance?
How can this limit be removed or significantly increased? Does upgrading to Tier 2 lift this daily cap, or is additional prepayment required?
If prepayment is the key, what is the minimum amount needed to unlock a reasonable daily quota (e.g., 500K+ tokens/day)?

I have already added a small prepayment to my account, but the 50K daily limit remains unchanged. This is blocking my work entirely. Please advise on the exact steps to resolve this. Best regards.

KFSys · Answer

Hey, That 50K/day cap isn’t a bug or a billing glitch — it’s the entry-tier guardrail on Serverless Inference. A quick rundown of how the system works, since it explains all three of your questions: Serverless Inference is prepaid and tier-based . Your tier is driven by your prepaid balance / spend history, and each tier sets your rate limits (requests-per-minute, tokens-per-minute, and the daily token cap you’re hitting). New/low-balance accounts land in the lowest tier, which is where that very tight 50K/day limit comes from — it’s an abuse/safety guardrail for fresh accounts, not something specific to Kimi K3/K2.6. 1. Why the 50K cap? It’s the lowest-tier daily safeguard, tied to your account being new / low prepaid balance — not the model itself. Worth knowing: Tier 1 and Tier 2 are also the most restricted on model access (they don’t get the Anthropic/OpenAI models), so moving up tiers matters for more than just token volume. 2. How to lift it? Two levers: Move up tiers by increasing your prepaid balance / accumulating usage. Higher tiers jump the per-minute and daily limits substantially. Request an increase directly. If you’re a team owner or resource modifier, go to the Resource Limits page in the DigitalOcean Control Panel and request a limit increase there. For production-level quotas, contact support with your use case. 3. Minimum prepayment for ~500K/day? Here’s the honest part: DigitalOcean doesn’t publish the exact prepayment-amount → daily-cap mapping, and the specific daily numbers per tier aren’t in the public docs. So there’s no reliable figure I can quote you — that threshold is account-specific. On why your small prepayment didn’t change anything: tier upgrades don’t always apply instantly after adding a bit of balance, and a hard daily cap in particular usually needs an explicit bump. Since it’s blocking you entirely, the fastest fix is: Open the Resource Limits page in the Control Panel and request an increase, and Contact DigitalOcean support directly — tell them you’re on Serverless Inference, hitting x-ratelimit-limit-tokens-per-day: 50000, and state the daily quota you need. They can see your tier and tell you exactly what balance/steps unlock the quota you’re after. That support ticket is really the only way to get the precise prepayment number and to get the cap raised quickly, since it’s all account-specific.

alexdo · Answer

Worth adding some specifics though, since DigitalOcean’s Inference Limits documentation does publish a concrete tier table for requests-per-minute and tokens-per-minute: Tier 1 and Tier 2 both sit at 120 requests per minute and 500K to 750K tokens per minute. Tier 3 and Tier 4 move up to 600 requests per minute and 800K to 2M tokens per minute. Tier 5 goes up to 4,500 requests per minute and 3.5M to 70M tokens per minute. The docs also state directly that moving from Tier 1 to Tier 2 requires completing a prepayment in the Control Panel, which also unlocks access to additional models, Tier 1 and Tier 2 both lack access to Anthropic models and most OpenAI models. The x-ratelimit-limit-tokens-per-day header you are hitting is not part of that documented tier table at all, so I cannot confirm from the docs whether completing the Tier 2 prepayment actually raises your daily cap the same way it raises the per-minute limits, or whether the daily cap is a separate guardrail entirely. That is specifically what is worth asking support to confirm rather than just requesting a generic limit increase, since you want to know whether the daily token cap moves with tier, not just the per-minute numbers. Regards

Bobby · Answer

Hi there, I think that the 50K daily token limit is a Tier 1 restriction applied to new accounts as an abuse prevention measure. It is not related to your prepayment balance specifically but to your account tier. To get the limit raised you need to move to a higher tier. The exact requirements and token limits per tier are documented here: https://docs.digitalocean.com/products/inference/details/limits/ If your account does not automatically move tiers after meeting the requirements, open a support ticket at https://cloud.digitalocean.com/support and ask specifically for your inference rate limits to be reviewed. Mention your use case and that you have already added a prepayment. That usually speeds things up.

Become a contributor for community

DigitalOcean Documentation

Resources for startups and AI-native businesses

The developer cloud

Start building today

Report this

Why x-ratelimit-limit-tokens-per-day too little?

Related questions

Comparing model latency and cost on DigitalOcean Inference before you commit to one

Tier 2 serverless inference: Daily token bucket caps at 500K

Tier 2 serverless inference: Daily token bucket caps at 500K

What are the best AI frameworks and tools for adding AI features to mobile apps?

Report this

Why x-ratelimit-limit-tokens-per-day too little?

Related questions

Comparing model latency and cost on DigitalOcean Inference before you commit to one

Tier 2 serverless inference: Daily token bucket caps at 500K

Tier 2 serverless inference: Daily token bucket caps at 500K

What are the best AI frameworks and tools for adding AI features to mobile apps?