Service eligibility and regional restrictions
PuppyIP serves only compliant overseas businesses and their authorized personnel. Proxy services are not available in mainland China. The service may only be used for lawful business activities outside mainland China. Use of this service within mainland China is prohibited.
Hosting a proxy IP or server overseas does not change these restrictions. The service must not be provided to end users in mainland China through relaying, forwarding, sharing or resale. Before use, read the Terms of Service.
Key Takeaways
- AWS document history explicitly says Cross-Model Max Tokens Per Day replaces the old per-model daily limits for bedrock-runtime. It covers supported models per account and Region, not a separate daily allowance for every model.
- General Reference currently lists Cross-Model Account-Level Tokens Per Day as 700,000,000 per supported Region, with Adjustable set to No. This public reference does not prove your account's granted quota or callable request count.
- Consumption is calculated as estimated billed tokens based on on-demand inference prices, affected by model prices, input/output ratio and cache hits. Do not treat 700M as an unweighted input-plus-output token total.
- Check the daily cross-model threshold alongside per-model, per-Region token/request limits per minute. This Runtime table does not automatically cover bedrock-mantle, Batch or Provisioned Throughput.
- Official pages do not provide the complete supported-model set, normalization formula, account-level enforcement values or your team's remaining quota. PRODUCT_FIT=NONE: Changing IP cannot increase AWS account Service Quotas.
The direct answer: Old per-model daily limits become one cross-model threshold
The Amazon Bedrock User Guide document history dates the update September 21, 2026 and explicitly says bedrock-runtime replaces per-model daily token quotas with Cross-Model Max Tokens Per Day. The new metric covers supported Bedrock models per account and Region. Only a date is given, without an exact activation time or time zone.
This changes multimodel capacity planning. Do not sum the daily allowances from different models' old tables, or interpret unused per-minute quota for one model as proof of remaining account capacity for the day. First confirm that requests use bedrock-runtime, then check the account and Region. For an existing model-call guide, see the Bedrock Grok 4.6 dual-endpoint explanation: Mantle and Runtime use different model IDs and authentication paths.
700M is a public reference, not a guarantee of 700 million raw tokens
AWS General Reference currently lists Cross-Model Account-Level Tokens Per Day as 700,000,000 per supported Region, with Adjustable set to No. Its description emphasizes an overall cross-model threshold of estimated billed tokens, estimated using on-demand inference prices. Actual consumption varies with model price, input/output token ratio and cache-hit rate.
Do not divide 700 million by raw tokens per call to promise a request count, or treat the public reference as an observed value already enforced for every account, model and Region. Account administrators should check visible quotas in Service Quotas for the Region in use and current request records. If console values or enforcement conflict with the public table, contact AWS with account-level evidence instead of applying your own conversion formula.
Impact matrix: How daily thresholds and per-minute limits interact
The cross-model daily threshold applies to supported bedrock-runtime model totals per account and Region, measured using a price-related estimate. Per-model minute limits separately constrain token or request rates by model and Region; some models have different RPM rules, so check Service Quotas. Bursts may hit a minute limit even with daily budget remaining. Conversely, normal short-term requests do not prove sufficient cross-model capacity for the day.
bedrock-mantle has separate quota documentation. Batch inference, Provisioned Throughput and custom inference configurations must not be inferred from this Runtime daily metric merely because they are also Bedrock services. Cross-region inference also involves source Region, inference configuration and model support. Remaining capacity in one Region cannot simply be moved to another.
Distinguish endpoints, accounts and model combinations before planning
When a team uses several Bedrock models in the same account and Region for customer service, coding and retrieval, list actual endpoints, model IDs, Region, granted quotas, minute peaks and actual per-model billed usage first. Leave headroom under the currently published cross-model daily rule, and include cache hits and changes in input/output structure in budgets. Do not use 700M as a launch commitment before confirming account readings.
If capacity approaches the threshold, pause expansion of batch tasks and preserve request IDs, Region, models and visible quotas. Follow AWS documentation and Support procedures to confirm available capacity and increase options. Adjustable=No in General Reference does not mean every combined limit can be raised directly in the console. This article does not claim to have requested or verified an increase for any account. External-data access also requires permission and data-boundary checks beyond quota; see the Bedrock external Web Search access guide.
What still requires account evidence
AWS's public pages do not establish a particular account's current displayed quota, exactly which models count in the cross-model set, the price-normalization formula, reset time zone, live enforcement results or whether Support can raise that account's threshold. Keep these fields unknown. Choosing network egress or a fixed IP does not grant model access, Region eligibility or Service Quotas. Capacity decisions should return to the AWS account and official support channels.
Sources
Frequently Asked Questions
When was Amazon Bedrock's cross-model daily quota documented?
AWS document history marks September 21, 2026, replacing bedrock-runtime per-model daily quotas with Cross-Model Max Tokens Per Day aggregated by account and Region. The exact switch time is unpublished.
Does 700M mean 700 million tokens for every individual model?
No. AWS General Reference lists a cross-model reference threshold of 700,000,000 per supported Region, measured as estimated billed tokens. It is not an individual raw-token allowance for every model.
Why can 700M not be converted directly into a request count?
Consumption varies with model price, input/output ratio and cache hits. These public pages do not give a fixed raw-token conversion formula valid for all accounts.
With a cross-model daily allowance, must I still check per-model minute quotas?
Yes. Daily cross-model thresholds and per-model, per-Region TPM/RPM limits are separate dimensions. Short bursts can hit minute limits first.
Can I use the same Runtime table for bedrock-mantle or Batch calls?
Not directly. This public explanation concerns bedrock-runtime. Check the corresponding AWS quota documents and account data separately for Mantle, Batch and Provisioned Throughput.
Can changing proxy IP increase Bedrock's daily quota?
No. AWS account, Region, endpoint and service rules determine quotas. Network egress does not change Service Quotas.