Roboflow

Gemma 4 31B vs Grok 4.6

Compare Gemma 4 31B and Grok 4.6 side-by-side. See how these vision models stack up in Image Captioning, OCR, Open Prompt, Object Detection, and Classification.

Compare Gemma 4 31B vs Grok 4.6 live

Run the same image across every model that supports a task and compare their outputs side-by-side.

Detect and compare bounding boxes across models on the same image.

Open Object Detection in the full playground
GoogleGemma 4 31B
Run to compare this model.
GrokGrok 4.6
Run to compare this model.

Models in this comparison

Gemma 4 31B vs Grok 4.6 on Vision Evals

Grok 4.6 scores higher on 5 of the six Vision Evals tasks.

The widest gap is Object Detection, where Gemma 4 31B leads 47.5% to 23.8%.

Overall, Gemma 4 31B averages 67.0% (#36 of 61) against 68.7% (#33 of 61) for Grok 4.6.

Gemma 4 31B is cheaper ($0.0015 vs $0.0097 per sample), while Grok 4.6 is faster (17.5s vs 34.4s per sample).

Gemma 4 31BGrok 4.6

Gemma 4 31B vs Grok 4.6 Comparison Table

Evals updated September 29, 2026Pricing updated September 29, 2026

PropertyGemma 4 31BGrok 4.6
OrganizationGoogleSpaceXAI
Categoryopenclosed
Modalitymultimodalmultimodal
Release DateApr 2026Aug 2026
Context Window256K500K
Parameters31B
LicenseApache 2.0Proprietary
Pricing per 1M tokens
Input $/1M$0.090$2.00
Output $/1M$0.340$6.00
Vision Tasks
CaptioningDemoDemo
Chart Question Answering
ClassificationDemoDemo
Document Question Answering
Image Tagging
Multi-Label Classification
Object DetectionDemoDemo
OCRDemoDemo
Vision Language
Visual Question AnsweringDemoDemo
Model Features
Foundation Vision
LLMs with Vision Capabilities
Multimodal Vision
Vision Evalsground-truth scores across 6 vision tasks, pooled at low effort
Overall
67.0%
68.7%
Quantizationsself-hosted
BF1666.9%FP867.0%QAT-W4A1667.0%hardware →
Avg cost / sample$0.0015$0.0097
Avg speed / sample34.36s17.55s
By task
Object Detection (low)
47.5%
±0.6, Mean of 3 runs, range 46.8 to 48.0
$0
23.8%
±2.8, Mean of 3 runs, range 20.2 to 25.9
$0.013
Object Detection (high)–
24.0%
±1.0, Mean of 3 runs, range 23.1 to 25.1
$0.041
Counting (low)
51.4%
±2.7, Mean of 3 runs, range 48.6 to 54.0
$0
65.8%
±4.1, Mean of 3 runs, range 62.2 to 70.3
$0.0079
Counting (high)–
56.8%
±1.4, Mean of 3 runs, range 55.4 to 58.1
$0.027
Identification (low)
79.2%
±1.6, Mean of 3 runs, range 78.1 to 81.3
$0
84.4%
±3.1, Mean of 3 runs, range 81.3 to 87.5
$0.0055
Identification (high)–
85.4%
±1.6, Mean of 3 runs, range 84.4 to 87.5
$0.015
OCR (low)
90.8%
±0.6, Mean of 3 runs, range 90.2 to 91.5
$0
91.8%
±0.3, Mean of 3 runs, range 91.5 to 92.1
$0.0091
OCR (high)–
91.6%
±0.2, Mean of 3 runs, range 91.4 to 91.7
$0.023
Data Extraction (low)
80.4%
±1.0, Mean of 3 runs, range 79.4 to 81.4
$0
85.6%
±1.0, Mean of 3 runs, range 84.5 to 86.6
$0.0050
Data Extraction (high)–
85.6%
±1.0, Mean of 3 runs, range 84.5 to 86.6
$0.0090
Reasoning (low)
52.8%
±1.3, Mean of 3 runs, range 51.7 to 54.3
$0
61.1%
±1.3, Mean of 3 runs, range 59.6 to 62.3
$0.0093
Reasoning (high)–
63.8%
±2.0, Mean of 3 runs, range 62.3 to 66.2
$0.032

Gemma 4 31B vs Grok 4.6: Overview

Gemma 4 31B

Gemma 4 31B is the largest dense model in Google's Gemma 4 family, built from the same research as Gemini 3 and released as open weights under the Apache 2.0 license. It supports a 256K token context window with text and image input, configurable thinking mode for step-by-step reasoning, and multilingual support across 140+ languages. The unquantized model fits on a single 80GB GPU.

For vision tasks, Gemma 4 31B supports image understanding with variable aspect ratios and resolutions, and can output structured bounding boxes for UI element detection, making it useful for document parsing and UI understanding. Compared to Gemma 3, it delivers stronger reasoning and multimodal performance. It is part of a four-size family alongside the 26B A4B MoE variant and two on-device models (E2B, E4B), with the 31B dense variant optimized for output quality and fine-tuning over inference speed.

Grok 4.6

Grok 4.6 is a proprietary reasoning model from xAI aimed at long-running agentic workflows, coding, and knowledge work. It accepts text and image input and returns text, with a 500,000 token context window and a knowledge cutoff of February 1, 2026. The model exposes an adjustable reasoning budget with low, medium, high, and xhigh settings, where high is the default, and it supports function calling, structured outputs, web and X search, and code execution as documented tool behaviors. Its visual capability covers interpreting images supplied alongside text prompts, which places it in the visual question answering and document understanding family, and it can also return object detection boxes as text coordinates when prompted.

xAI characterizes Grok 4.6 as the result of an extended post-training run over the Grok 4.5 lineage rather than a new pretrained base. The described recipe combines curated model-generated reasoning and technical data, engineering data, a revised optimizer, regenerated supervised fine-tuning trajectories, and reinforcement learning across agent environments spanning knowledge work, coding, kernel optimization, web development, and computer-aided design. Parameter count and architecture specifics are not disclosed. Independent measurement from Artificial Analysis places the model at 61 on its Intelligence Index, five points above Grok 4.5.

Frequently Asked Questions

On Roboflow's Vision Evals, Grok 4.6 performed better. It scores higher on 5 of the six vision tasks and averages 68.7% (#33 of 61) against 67.0% (#36 of 61) for Gemma 4 31B. These scores measure vision capabilities only; either model may perform differently on work outside this benchmark's tasks.

Yes. On the Vision Evals Object Detection benchmark at low effort, Gemma 4 31B leads with 47.5% against 23.8%. This is the widest gap between the two models across the benchmark's tasks.

Gemma 4 31B is cheaper to run. Across Roboflow's Vision Evals task mix it averaged $0.0015 per sample against $0.0097. Actual costs depend on your image sizes, prompts, and output length.

Grok 4.6 is faster. Across Roboflow's Vision Evals it averaged 17.5s per inference against 34.4s. Latency varies with image size, prompt length, and provider load, so treat these as relative rather than guaranteed figures.

Yes. The comparison demo on this page runs both models on the same image side by side for image captioning and OCR in the free Roboflow Playground. You can try it instantly, and a free account unlocks unlimited runs.