What image tasks do these models do well?

Reading charts and tables, extracting text from scans and screenshots, describing photos, and reviewing UI states. Attach the image in chat and ask; files in Drive work as image context too.

Can these models edit or generate images?

No, reading and generating are different jobs. For creating and editing images, the image generation roundup covers the dedicated models, which run in the same workspace.

How do I pick between them?

Attach your actual image and run the same question across two or three picks in parallel tabs; vision quality varies more by task than text quality does, so your document is the benchmark.

All roundups

Best of

Models that read what you see.

Models that accept images and return analysis: screenshots, charts, scans, and photos, verified against live capability data.

Last reviewed July 16, 2026

Try idapt free Browse all models

Vision models turned images from decoration into input: a chart gets read, a scanned table becomes data, a UI screenshot gets reviewed like code. Every pick below accepts image input per the live capability registry, and because they share one composer in idapt, you can attach the same image once and re-ask across models until one reads it right.

How we pickedEvery pick's image-input capability is recorded in the live model registry (the same data the model picker enforces), filtered to current models and ordered by general benchmark standing. Where a pick has a specialty (documents, open weights, price), the entry names it; prices and context render live.

The picks

1Gemini 3.1 ProThe multimodal workhorse.

$2 in / $12 out per M tokens · 1.0M context

Image understanding attached to a million-token window: dozens of screenshots or scanned pages stay in one conversation with their analysis.

Best for: Document and chart analysis at scale.

2GPT-5.6 SolFrontier reasoning about images.

$5 in / $30 out per M tokens · 1.1M context

Top-lane reasoning applied to what it sees: strong at UI screenshots, diagrams, and questions where the image is evidence, not the answer.

Best for: Mixed text-and-image problem solving.

3Claude Opus 4.8The careful document reader.

$5 in / $25 out per M tokens · 1M context

The family's careful-analysis habits extend to figures: it distinguishes what a chart shows from what a caption claims, which matters in contracts and reports.

Best for: Contracts, reports, and figures that matter.

4Qwen3 VL 235B A22BThe open-weight vision flagship.

$0.2 in / $0.88 out per M tokens · 262K context

The strongest open-weight vision line: published weights at flagship scale, for teams whose vision workloads need open deployment options.

Best for: Open-weight requirements with vision.

5Gemini 3.5 FlashCheap vision at volume.

$1.50 in / $9 out per M tokens · 1.0M context

Image input at the speed line's price: the pick for reading hundreds of screenshots or scans where per-image cost decides feasibility.

Best for: Bulk image reading on a budget.

Frequently asked

Run them in one workspace

Multi-Model Chat

GPT, Claude, Gemini, Grok, Llama: pick mid-conversation, cheapest-first.

Drive

Folders, previews, and lightweight versioning for AI work.

Image Generation

DALL·E, Imagen, FLUX, Stable Diffusion: same composer, same files.

Keep comparing

The image models that earn their prompts.Models that cite what survives.Models that hold the whole thing.

Judge them on your own prompt

Every pick on this page runs in one idapt chat. Start free with a daily allowance, no credit card.

Try idapt free