Skip to content
CallMissed

Explore CallMissed

Back to all models
LLM Chatnewfastvision
Google — gemma-4-31b logo

Gemma 4 31B

by Google · Released 2026

Google's Gemma 4 31B instruct model: 128K context, image input and tool calling. With thinking off (the default) it is fast enough for live voice agents.

Google — gemma-4-31b logo
LLM Chat

Gemma 4 31B

Powered by Google · Dense transformer (Gemma 4)

Context Window

128K

Parameters

31B

Max Output

128K (shared with input)

Category

LLM Chat

Overview

Gemma 4 31B is the dense 31-billion-parameter instruct model in Google's open Gemma 4 family. By default it answers without thinking, which keeps latency low: in our tests on a short phone-receptionist prompt the first word arrived in about 0.25 seconds, in English and in Hindi.

It accepts images alongside text. Send them as base64 data URIs in the standard `image_url` content part; remote image URLs are not accepted. It also supports OpenAI-compatible tool calling, so it can book, look up and hand off inside an agent loop.

The 131,072-token context window covers long conversations and sizeable documents. Sending `reasoning_effort` as `low`, `medium` or `high` switches thinking on for harder questions, at the cost of latency and extra output tokens; `none` keeps it off.

Pricing

MetricPrice
Input /1M tokens₹59.0000
Output /1M tokens₹148.0000

1 credit = ₹1 = $0.01 USD. Transparent per-call pricing — you pay only for what you call, with no seat fees or minimums.

Key Highlights

  • About 0.25s to the first word in our voice tests
  • Image input via base64 data URIs
  • Tool calling for agent workflows
  • 131,072-token context window

Benchmarks

BenchmarkScore
First word~0.25s
Context128K
VisionYes

Technical Details

  • Dense 31B instruct model
  • Context window: 131,072 tokens
  • Thinking off by default; reasoning_effort low / medium / high turns it on
  • Image input as base64 data URIs (remote URLs rejected)
  • Tool calling via OpenAI-compatible tools and tool_choice

Strengths

  • Very low latency for live conversation
  • Handles both text and images
  • Good value for everyday chat and agent work

Limitations

  • Thinking is on/off: the effort level does not reliably change its depth
  • Images must be base64, not URLs
  • Paid plans only

Use Cases

Voice agentsCustomer support chatImage understandingTool-using assistants

API Example

curl https://api.callmissed.com/v1/chat/completions \
  -H "Authorization: Bearer cm_YOUR_KEY" \
  -d '{"model": "gemma-4-31b", "messages": [{"role": "user", "content": "Write a short welcome message for our clinic"}]}'

Endpoint: POST /v1/chat/completions · Model ID: gemma-4-31b

Try Gemma 4 31B now

Get 1000 free API credits on signup. No credit card required.

Start free Book a DemoRead docs