Google AI language model

Gemma 4 31B

Gemma 4 31B is the largest dense model in Google's Apache 2.0 open-weights Gemma 4 family, released April 2026. it ships alongside the 26B A4B mixture-of-experts model, takes text and images, and has an optional thinking mode.

01

price

$0.2/M

input tokens

$0.2/M

output tokens

per 1M tokens. octomind bills the listed provider price, with no markup, on every plan, including the free one (pay as you go from credits).

  • output bills at the same $0.20 per 1M tokens as input, so long generations cost no more per token than prompts.
  • no cached-input discount: repeated prompt tokens bill at the full input rate.

02

what real work costs

workloadcost
one chat reply 3,000 tokens in, 500 out $0.0007
summarise a 100-page report 50,000 tokens in, 1,500 out, one request $0.01
1,000 classification calls 500 tokens in, 20 out, per call $0.10
coding-agent session 2M tokens in across ~60 turns, 60,000 out, no cache $0.41

at the listed rate. token counts are typical sizes, not measurements.

03

specs

provider
Google AI
context window
128k tokens — ≈ 190 pages of prose or 13,000 lines of code
input
text, images
output
text
structured output
yes — JSON schema
reasoning mode
yes
weights
open — Apache 2.0
released
April 2, 2026
knowledge cutoff
January 2025
status
live
model id
gemma-4-31b

integration notes

  • thinking is switched on by a <|think|> token at the start of the system prompt; remove it to switch thinking off.
  • with thinking off, the 31B model still emits an empty thought block before its answer.
  • in multi-turn history, keep only final answers from earlier turns; keep thinking only on tool-call turns.

04

benchmarks

as published by the maker; not measured by octomind.

benchmarkscore
GPQA Diamondsource84.3%
AIME 2026 no toolssource89.2%
LiveCodeBench v6source80.0%
Tau2 (average over 3)source76.9%
MMMU Prosource76.9%

05

when to pick it

pick it if

  • want Apache 2.0 open weights you can also self-host
  • need image input and a thinking mode at a low flat rate
  • write long outputs, since output costs the same as input

06

where it sits on price

blended price $0.2/M at 3 input tokens per output token: 6 live language models cost less, 37 cost more.

  1. Beam $0.194/M
  2. Claude Haiku 5.5 $0.2/M
  3. Gemma 4 31B $0.2/M
  4. GPT-6 Luna $0.2/M
  5. Qwen3.8 Omni Flash $0.23/M

07

use Gemma 4 31B in octomind

one octohub login reaches every model in the catalog. pick this one per run, per session, or as your default.

run it
octomind login
octomind run -m octohub:gemma-4-31b
switch mid-session
/model octohub:gemma-4-31b
make it the default — config.toml
[model]
name = "octohub:gemma-4-31b"

08

compare Gemma 4 31B

all model comparisons →

09

similar models

nearest live models by listed input price.

view all models →

10

Gemma 4 31B faq

how much does Gemma 4 31B cost?

Gemma 4 31B lists at $0.2/M input, $0.2/M output per 1M tokens. a typical coding-agent session (2M tokens in, 60,000 out) comes to $0.41. octomind bills the listed provider price, with no markup, on every plan, including the free one (pay as you go from credits).

what is the context window of Gemma 4 31B?

128k tokens, ≈ 190 pages of prose or 13,000 lines of code.

does Gemma 4 31B accept images?

yes. it takes text and images as input and returns text.

is Gemma 4 31B open source?

its weights are open under Apache 2.0, so it can also be self-hosted.

when was Gemma 4 31B released?

April 2, 2026, by Google AI.

is Gemma 4 31B free?

no. Gemma 4 31B bills per token at $0.2/M input, $0.2/M output per 1M. octomind's free plan reaches it pay as you go from credits, so you pay only for the tokens you use, at the provider price. credits carry a 3% top-up fee.

how do i use Gemma 4 31B in octomind?

run octomind login once, then octomind run -m octohub:gemma-4-31b. inside a running session, /model octohub:gemma-4-31b switches to it without losing context.

use Gemma 4 31B in octomind

choose a model when you start, then switch mid-session when the work changes.