Gemma 4 31B IT: a public model name without a public deployment spec
The 31B IT model is public, but its deployment details are not. Do not let the name stand in for limits, pricing, or results.
By the benchr team · · Changelog · Provider-published facts rechecked against the official sources on July 28, 2026
GOOGLE · GEMMA431B · OPEN
Benchr model field plateGemma 4 31BOpen-weight · no inferred spec
Editorial imageA BenchR editorial illustration of a larger open-weight architecture, where a suffix alone does not complete the specification.
Released2 Apr2026
Model ID31BIT
Output limit—Not published
Benchmarks—Not published
Google released gemma-4-31b-it on April 2 alongside the 26B A4B IT model. That shared launch does not make the two records interchangeable.
The suffix is a name, not a contract
gemma-4-31b-it is useful when you look for the endpoint. It does not tell you the input limit, output cap, deployment hardware, or observed performance. Get those details from a model-specific provider page, not from the name.
Do not copy the sibling's numbers
Google announced the 31B and 26B models together. Unpublished figures are still separate for each ID. benchr records the shared availability and leaves the unlisted limits, price, and benchmarks empty on both records.
Ask six questions before you reserve capacity
Confirm endpoint access, rate limits, the pricing unit, input and output limits, license terms, and the acceptance test for your workload. The launch note gets this model onto a watchlist. It does not answer the production questions.
benchr