Gemini 3.6 Flash vs Gemma 4 31B
Compare Gemini 3.6 Flash and Gemma 4 31B side-by-side. See how these vision models stack up in Open Prompt, Classification, Image Captioning, OCR, and Object Detection.
Compare Gemini 3.6 Flash vs Gemma 4 31B 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.
Upload an image
Drag and drop an image here, or click to browse
Models in this comparison
Gemini 3.6 Flash vs Gemma 4 31B on Vision Evals
Gemini 3.6 Flash scores higher on all six Vision Evals tasks.
The widest gap is Reasoning, where Gemini 3.6 Flash leads 77.7% to 49.0%.
Overall, Gemini 3.6 Flash averages 83.0% (#6 of 52) against 67.1% (#29 of 52) for Gemma 4 31B.
Gemma 4 31B is cheaper ($0.0011 vs $0.0032 per sample), while Gemini 3.6 Flash is faster (14.7s vs 29.7s per sample).
Gemini 3.6 Flash vs Gemma 4 31B Comparison Table
Evals updated September 3, 2026Pricing updated September 4, 2026
| Property | Gemini 3.6 Flash | Gemma 4 31B |
|---|---|---|
| Organization | ||
| Category | closed | open |
| Modality | multimodal | multimodal |
| Release Date | Jul 2026 | Apr 2026 |
| Context Window | 1.0M | 256K |
| Parameters | 31B | |
| License | Proprietary | Apache 2.0 |
| Pricing per 1M tokens | ||
| Input $/1M | $0.750 | $0.090 |
| Output $/1M | $3.75 | $0.340 |
| Vision Tasks | ||
| Captioning | Demo | Demo |
| Chart Question Answering | ||
| Classification | Demo | Demo |
| Document Question Answering | ||
| Image Tagging | ||
| Multi-Label Classification | ||
| Object Detection | Demo | Demo |
| OCR | Demo | Demo |
| Vision Language | ||
| Visual Question Answering | Demo | Demo |
| Video Classification | ||
| Model Features | ||
| Foundation Vision | ||
| LLMs with Vision Capabilities | ||
| Multimodal Vision | ||
Vision Evalsground-truth scores across 6 vision tasks, pooled at low effort | ||
| Overall | 83.0% | 67.1% |
| Quantizationsself-hosted | ||
| Avg cost / sample | $0.0032 | $0.0011 |
| Avg speed / sample | 14.66s | 29.72s |
| By task | ||
| Object Detection (low) | 57.1% ±1.7, Mean of 3 runs, range 55.9 to 59.4 | 48.0% |
| Object Detection (high) | 70.7% ±0.4, Mean of 3 runs, range 70.3 to 71.2 | – |
| Counting (low) | 80.2% ±2.0, Mean of 3 runs, range 78.4 to 82.4 | 52.7% |
| Counting (high) | 79.3% ±2.7, Mean of 3 runs, range 77.0 to 82.4 | – |
| Identification (low) | 99.0% ±1.6, Mean of 3 runs, range 96.9 to 100.0 | 84.4% |
| Identification (high) | 100.0% ±0.0, Mean of 3 runs, range 100.0 to 100.0 | – |
| OCR (low) | 88.2% ±0.3, Mean of 3 runs, range 87.9 to 88.4 | 85.6% |
| OCR (high) | 89.5% ±0.0, Mean of 3 runs, range 89.5 to 89.6 | – |
| Data Extraction (low) | 95.9% ±1.0, Mean of 3 runs, range 94.8 to 96.9 | 82.7% |
| Data Extraction (high) | 94.8% ±1.0, Mean of 3 runs, range 93.8 to 95.9 | – |
| Reasoning (low) | 77.7% ±2.0, Mean of 3 runs, range 76.2 to 80.1 | 49.0% |
| Reasoning (high) | 81.0% ±2.0, Mean of 3 runs, range 79.5 to 83.4 | – |
Gemini 3.6 Flash vs Gemma 4 31B: Overview
Gemini 3.6 Flash is a multimodal language model from Google DeepMind, positioned as the workhorse tier in the Gemini 3.x family. It accepts text, image, video, audio, and PDF inputs with a 1 million token context window and produces up to 64,000 output tokens. The model builds directly on Gemini 3.5 Flash, incorporating developer and customer feedback to improve token efficiency, coding quality, and knowledge work performance. According to the Artificial Analysis Index, it consumes 17% fewer output tokens than its predecessor, and on some benchmarks such as DeepSWE, token reduction reaches up to 65%. It supports function calling, structured output, search as a tool, and code execution, and includes computer use as a built-in capability in the Gemini API and Gemini Enterprise.
On coding benchmarks, Gemini 3.6 Flash scores 49% on DeepSWE versus 37% for 3.5 Flash, and 63.9% on MLE Bench versus 49.7%. Computer use performance on OSWorld-Verified improves from 78.4% to 83%, and knowledge work scores on GDPval-AA v2 rise from 1349 to 1421. The model carries a knowledge cutoff of March 2026 and ships with enhanced Frontier Safety safeguards covering chemical, biological, radiological, nuclear, and cyber offense domains, with training to minimize refusals for beneficial uses. It is a proprietary, closed-weights model available in preview through the Gemini API via Google AI Studio, Android Studio, Google Antigravity, Gemini Enterprise, and the Gemini app.
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.
Frequently Asked Questions
On Roboflow's Vision Evals, Gemini 3.6 Flash performed better. It scores higher on all six vision tasks and averages 83.0% (#6 of 52) against 67.1% (#29 of 52) 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 Reasoning benchmark at low effort, Gemini 3.6 Flash leads with 77.7% against 49.0%. 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 $0.0011 per sample against $0.0032. Actual costs depend on your image sizes, prompts, and output length.
Gemini 3.6 Flash is faster. Across Roboflow's Vision Evals it averaged 14.7s per inference against 29.7s. 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 open prompts and image classification in the free Roboflow Playground. You can try it instantly, and a free account unlocks unlimited runs.