Claude Sonnet 5.5 vs Gemma 4 31B
Compare Claude Sonnet 5.5 and Gemma 4 31B side-by-side.
Compare Claude Sonnet 5.5 vs Gemma 4 31B live
Run the same image across every model that supports a task and compare their outputs side-by-side.
These models don't share enough common tasks for a side-by-side demo. See the comparison table below for their capabilities.
Models in this comparison
Claude Sonnet 5.5 vs Gemma 4 31B on Vision Evals
Claude Sonnet 5.5 scores higher on 5 of the six Vision Evals tasks.
The widest gap is Counting, where Claude Sonnet 5.5 leads 79.3% to 51.4%.
Overall, Claude Sonnet 5.5 averages 83.8% (#7 of 60) against 67.0% (#35 of 60) for Gemma 4 31B.
Gemma 4 31B is cheaper ($0.0015 vs $0.0065 per sample), while Claude Sonnet 5.5 is faster (10.8s vs 34.4s per sample).
Claude Sonnet 5.5 vs Gemma 4 31B Comparison Table
Evals updated September 28, 2026Pricing updated September 28, 2026
| Property | Claude Sonnet 5.5 | Gemma 4 31B |
|---|---|---|
| Organization | Anthropic | |
| Category | closed | open |
| Modality | multimodal | multimodal |
| Release Date | Sep 2026 | Apr 2026 |
| Context Window | 1.0M | 256K |
| Parameters | 31B | |
| License | Proprietary | Apache 2.0 |
| Pricing per 1M tokens | ||
| Input $/1M | $0.090 | |
| Output $/1M | $0.340 | |
| Vision Tasks | ||
| Captioning | Demo | |
| Chart Question Answering | ||
| Classification | Demo | |
| Document Question Answering | ||
| Image Tagging | ||
| Multi-Label Classification | ||
| Object Detection | Demo | |
| OCR | Demo | |
| Vision Language | ||
| Visual Question Answering | Demo | |
| Model Features | ||
| Foundation Vision | ||
| LLMs with Vision Capabilities | ||
| Multimodal Vision | ||
Vision Evalsground-truth scores across 6 vision tasks, pooled at low effort | ||
| Overall | 83.8% | 67.0% |
| Quantizationsself-hosted | ||
| Avg cost / sample | $0.0065 | $0.0015 |
| Avg speed / sample | 10.78s | 34.36s |
| By task | ||
| Object Detection (low) | 74.3% ±0.9, Mean of 3 runs, range 73.5 to 75.3 | 47.5% ±0.6, Mean of 3 runs, range 46.8 to 48.0 |
| Object Detection (high) | 76.8% ±0.4, Mean of 3 runs, range 76.5 to 77.3 | – |
| Counting (low) | 79.3% ±0.7, Mean of 3 runs, range 78.4 to 79.7 | 51.4% ±2.7, Mean of 3 runs, range 48.6 to 54.0 |
| Counting (high) | 82.9% ±1.4, Mean of 3 runs, range 81.1 to 83.8 | – |
| Identification (low) | 91.7% ±3.1, Mean of 3 runs, range 87.5 to 93.8 | 79.2% ±1.6, Mean of 3 runs, range 78.1 to 81.3 |
| Identification (high) | 90.6% ±0.0, Mean of 3 runs, range 90.6 to 90.6 | – |
| OCR (low) | 90.6% ±0.9, Mean of 3 runs, range 90.0 to 91.7 | 90.8% ±0.6, Mean of 3 runs, range 90.2 to 91.5 |
| OCR (high) | 90.9% ±1.5, Mean of 3 runs, range 89.2 to 92.3 | – |
| Data Extraction (low) | 90.7% ±1.5, Mean of 3 runs, range 89.7 to 92.8 | 80.4% ±1.0, Mean of 3 runs, range 79.4 to 81.4 |
| Data Extraction (high) | 93.1% ±0.5, Mean of 3 runs, range 92.8 to 93.8 | – |
| Reasoning (low) | 76.4% ±0.7, Mean of 3 runs, range 75.5 to 76.8 | 52.8% ±1.3, Mean of 3 runs, range 51.7 to 54.3 |
| Reasoning (high) | 83.9% ±1.7, Mean of 3 runs, range 82.1 to 85.4 | – |
Claude Sonnet 5.5 vs Gemma 4 31B: Overview
Claude Sonnet 5.5 is a proprietary multimodal language model from Anthropic and the second release in the Claude 5.5 family, following Claude Opus 5.5. It accepts interleaved text and image input and returns text, operating with a 1M token context window and a maximum output of 128K tokens per request. The model uses adaptive thinking by default, allocating variable reasoning effort per request rather than exposing a manual extended thinking toggle, and its training data cutoff is June 2026. Anthropic positions it as a faster, lower cost complement to Opus 5.5 for well scoped everyday tasks, bug fixing, and producing documents, slides, and spreadsheets.
On visual and agentic evaluations reported at launch, Sonnet 5.5 scores 61.6% on Chartography, a chart recognition test, compared with 15.6% for Claude Sonnet 5, and 80.1% on OSWorld 2.1, a computer use benchmark measuring screenshot driven control of a desktop environment, compared with 57.0% for Sonnet 5. It reports 70.6% on Terminal-Bench 4.0 for agentic coding. Anthropic describes it as the first Sonnet model able to complete Pokemon Red from screenshots alone, and it generates output more than 30% faster than Sonnet 5 while using fewer tokens for equivalent work.
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, Claude Sonnet 5.5 performed better. It scores higher on 5 of the six vision tasks and averages 83.8% (#7 of 60) against 67.0% (#35 of 60) 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 Counting benchmark at low effort, Claude Sonnet 5.5 leads with 79.3% against 51.4%. 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 $0.0065. Actual costs depend on your image sizes, prompts, and output length.
Claude Sonnet 5.5 is faster. Across Roboflow's Vision Evals it averaged 10.8s per inference against 34.4s. Latency varies with image size, prompt length, and provider load, so treat these as relative rather than guaranteed figures.