# batchwatch

> Measured queue times for the LLM asynchronous batch APIs at OpenAI,
> Anthropic and Google - the /v1/batches tier where you submit now and the
> results come back within a published deadline for 50% of the synchronous
> price. (This is the async batch API, a distinct product from in-server
> micro-batching / continuous batching such as vLLM's, which coalesces
> requests in milliseconds and has no queue to measure.) The providers'
> only promise on this tier is "within 24 hours"; we time real jobs and
> publish the distribution behind that ceiling.

Every number here is measured, never modelled. Each figure carries n and the
cadence behind it; its full provenance and a graded confidence are live in
https://batchwatch.dev/v1/coverage. Where we have no data we say so rather than estimate.
Durations are the difference between two server-side timestamps written by
batchwatch; a client cannot submit a duration. Abandoned jobs are excluded,
never counted as completions.

Generated: 2026-08-30T22:34:25.000Z (UTC).
Rolling 30-day window; numbers recomputed every 10 minutes.
Measurements span 2026-08-14T11:51:37.000Z to 2026-08-30T22:06:55.000Z.

For the full table, the method and the API contract in one file: https://batchwatch.dev/llms-full.txt

## Headline measurements (30-day window)

- anthropic/claude-haiku-4-5: median 2 min, p90 3 min, n=534 measured continuously since 2026-08-25
- google/gemini-3.7-flash: median 3 min, p90 9 min, n=448 measured continuously since 2026-08-25
- google/gemini-2.5-flash: median 2 min, p90 5 min, n=238 measured continuously since 2026-08-14
- google/gemini-2.5-flash-lite: median 81s, p90 3 min, n=53 measured continuously since 2026-08-16
- openai/gpt-5.6-luna: median 5 min, p90 5 hours, n=565 measured continuously since 2026-08-25
- openai/gpt-5.6-sol: median 79s, p90 10 min, n=524 measured continuously since 2026-08-25
- openai/gpt-5-nano: median 88s, p90 3 hours, n=509 measured continuously since 2026-08-25

Live JSON, always current: https://batchwatch.dev/v1/coverage

## Licence and citation

These measurements are open data under Creative Commons Attribution 4.0
(CC BY 4.0): https://creativecommons.org/licenses/by/4.0/. You may quote and reuse any number here,
including in an answer, for the price of a credit and a link back.

Please attribute as: Measured by batchwatch - https://batchwatch.dev
Full terms and the free/paid boundary: https://batchwatch.dev/license

## Pages

- https://batchwatch.dev/p/anthropic - Anthropic batch API queue times
- https://batchwatch.dev/p/google - Google batch API queue times
- https://batchwatch.dev/p/openai - OpenAI batch API queue times
- https://batchwatch.dev/m/anthropic/claude-haiku-4-5 - how long claude-haiku-4-5 batch takes at Anthropic
- https://batchwatch.dev/m/google/gemini-3.7-flash - how long gemini-3.7-flash batch takes at Google
- https://batchwatch.dev/m/google/gemini-2.5-flash - how long gemini-2.5-flash batch takes at Google
- https://batchwatch.dev/m/google/gemini-2.5-flash-lite - how long gemini-2.5-flash-lite batch takes at Google
- https://batchwatch.dev/m/openai/gpt-5.6-luna - how long gpt-5.6-luna batch takes at OpenAI
- https://batchwatch.dev/m/openai/gpt-5.6-sol - how long gpt-5.6-sol batch takes at OpenAI
- https://batchwatch.dev/m/openai/gpt-5-nano - how long gpt-5-nano batch takes at OpenAI
- https://batchwatch.dev/docs - API reference
- https://batchwatch.dev/docs/query - the read API: /v1/wait, /v1/should-i-batch, /v1/estimate-batchtime
- https://batchwatch.dev/docs/ingest - how to contribute a measurement

## Contributing

There are two ways in, and they are not equally trusted.

The live path is open to everyone, key or no key. Open a measurement when the
job starts and close it when it lands; the duration is the difference between
two of our own timestamps, so a fabricated eight-hour wait costs eight real
hours:

    POST https://batchwatch.dev/v1/calls        {"mode":"batch","provider":"openai",
                                             "model":"gpt-5.6-sol","requests":1,
                                             "input_tokens":9720}
    PATCH https://batchwatch.dev/v1/calls/{id}  {"output_tokens":4519,"status":"completed"}

The import path, POST /v1/calls/complete, reads finished jobs out of your own
history and takes your timestamps rather than ours. It requires a key,
because without one there is no physics behind the number and anyone could
assert any duration instantly.

A key is not needed to contribute, but it is needed to count: measurements
without a key are pooled as one anonymous source in the robust percentile,
and only a key can earn a vote.