Co-Founder & Lead Programmer of AcceleratedLogic AI
Claude Sonnet 5.5 (batch) appears separately in OpenRouter’s newest-first, all-variants catalog. The batch entry exposes discounted execution of Sonnet 5.5. A team evaluating it should consider the timing and job-management contract as carefully as the price.
Verified route specifications
OpenRouter route, checked October 3, 2026
Value
Model ID
anthropic/claude-sonnet-5.5:batch
Listing date
September 28, 2026
Context
1,000,000 tokens
Inputs
text, image, file
Output
text
Input price
$1 / 1M tokens
Output price
$5 / 1M tokens
The listed $1/M input and $5/M output rates are half the standard route’s $2/M and $10/M. Both entries advertise the same one-million-token context in the checked catalog. The batch variant was paid and was not run through live chat for this article. OpenRouter batch listing
How OpenRouter describes batch execution
OpenRouter’s Batch API is asynchronous. It allows a provider to complete work during a 24-hour window in exchange for a per-token discount. The service returns per-request results, so a failed row need not discard an entire job. Its announcement says images and files must use public URLs and lists audio, video, and OpenRouter’s web-search plugin as unsupported batch inputs or features. OpenRouter’s Batch API announcement
Choosing suitable work
A backlog of summaries, classifications, or evaluation prompts can be a good fit when the result does not need to arrive during a user interaction. Assign a stable application ID to each row so the output can be joined back to the original record even if completion order changes. Validate each returned object before applying it to the application’s data.
Interactive editing has a different constraint: the user may be waiting to approve the next step before the rest of the task can continue. Batch scheduling introduces a variable delay at that checkpoint. Treat the acceptable completion window as a requirement, rather than infer responsiveness from the underlying Sonnet model’s streaming performance.
A worked token example
One million input tokens and 100,000 output tokens cost $1.50 at the displayed batch rates, compared with $3 at the displayed standard rates. This example excludes caching and other billed features and assumes the requests are accepted on the discounted route. It does not measure the time to complete a batch.
No hands-on latency or quality result is claimed here. For the shared model’s announcement and effort discussion, see our Sonnet 5.5 API guide. The exact batch endpoint and supported request shape should be checked against the current provider documentation before integration.