Checked 2026-10-17 · TypeSafe Models
Jev rate limits: organization quotas, throughput and 429s
Your organization determines both ceilings.
Read tokens/s and requests/s in the developer console. Gateway quotas can differ. Without those numbers, we cannot tell you how long a day of work takes or how much concurrency reaches a limit.
Enter your quotas in the calculator →Throughput and concurrency use the same assumption
For a known quota, requests/s is the smaller of your request ceiling and your token ceiling divided by tokens per call. Multiply that rate by latency in seconds to estimate in-flight concurrency. These are ideal ceilings, not measured capacity guarantees.
Choose content tokens or billed tokens explicitly. The latter includes the historical ~260-token request overhead measured for Jev 1.13 via OpenRouter on 23 September 2026. Whether that overhead consumes a rate quota is unverified. We use the same counting choice for throughput and concurrency.
Historical published limits
Current policy first verified here on 17 October 2026: limits are per organization. The records below describe earlier documentation; they do not establish any account's current quota. Dates are when this site first observed a change, not TypeSafe release dates.
- 2026-10-03
100,000 tokens/s and 40 requests/s → 100,000 tokens/s and 80 requests/s
Historical source - 2026-09-30
250,000 tokens/s and 1,200 requests/min → 100,000 tokens/s and 40 requests/s
Historical source
Historical latency, not current throughput
Measured on 2026-09-23, Jev 1.13 (jev-1.13-20260917) through OpenRouter, one client and location, median of three on a reused connection. These observations do not predict Jev 1.14 or long-context performance. No concurrency conclusion is possible without your quotas.
| Billed tokens | Warm latency |
|---|---|
| 1,286 | 375 ms |
| 4,317 | 382 ms |
| 8,354 | 478 ms |
| 16,429 | 571 ms |
| 24,502 | 571 ms |
| 32,579 | 734 ms |
Fresh-connection latency for the small request: 1,630 ms, including handshake. Measurement data.
Handling 429 and overloads
TypeSafe's SDKs retry with increasing delays and honor the retry-after header when present. If you call HTTP directly, implement bounded retries and backoff. Retrying immediately on 429 increases contention. Higher limits are available on custom and enterprise plans; ask TypeSafe for your account's terms.
Jev rate-limit questions
What are Jev's rate limits?
TypeSafe sets tokens-per-second and requests-per-second limits per organization. Read your current quotas in the developer console. There is no published numeric quota that applies to every account. A request over either quota returns 429.
Which rate limit binds first?
Divide your organization tokens/s quota by its requests/s quota to get the crossover tokens per call. Below it requests bind, above it tokens bind. You must specify whether your estimate counts content or billed tokens; TypeSafe does not document that detail.
What happened to the 100K tokens/s and 80 requests/s figures?
They are historical published figures first seen on 3 October 2026, not current universal quotas. On 17 October we verified that the current documentation directs users to organization-specific console quotas.
Get an email when Jev’s price or limits change
We monitor official documentation. Changes require review before publication.