Find the right AI model for the job

Browse every model by price and intelligence, then test the best ones side by side on your own prompt.

Intelligence against blended price per 1M tokens (log scale). 29 models scored. Scores from Artificial Analysis.

Click any dot to test that model, or pickTap a dot to test that model, or pick several and chat with them side by side.

Pick models and compare

Top for everything

Best value: MiMo V2.6 Pro scores 46 for $0.54/1M.

Test the shortlist on your prompt

Keep exploring

All 30 models in one table

35Image modelsTop rated: GPT Image 2.5 Sunburst. Good and cheap: Muse Image 1.0 ($0.01/image)Compare them 13Voice modelsTop rated: Gemini 3.8 Flash TTS. Good and cheap: Fish Audio Drama 3 ($0.01/1k chars)Compare them 25Video modelsTop rated: Wan v3.0 Video. Good and cheap: MiniMax H3 ($0.08/sec)Compare them

Popular head-to-heads

What we found testing them 29/35Every AI model has the same “controversial” take on humansText and image models 96%Everyone else is three months behind Claude and GPTAI labs 33.7Your 5-year-old MacBook can now beat free ChatGPTLocal AI

MCP server for AI agents

Get a second opinion without leaving chat

Connect ModelCompare to the AI app you already use. Your agent tests the prompt on any of the 30 models and tells you which answer is best for the money. Same credits, same prices as the site.

Compare answersOne prompt on up to 6 models, each answer with its cost and time.
Get a second opinionHave the smartest models review your agent's plan before it builds.
Make images, speech and videoGenerate on several models at once and pick the best.

Works in

Claude
ChatGPT
Claude Code
Cursor
VS Code
Codex
Gemini CLI

claude mcp add --scope user --transport http modelcompare https://www.modelcompare.ai/mcp

Set up your agent

Can you check what GPT-6 Astra thinks of this launch plan?

modelcompare · comparegpt-6-astra · blunt startup advisor · 22s

Astra mostly agrees with the plan but is harsher on how you'd test it. (full answer on ModelCompare, $0.04)

Where it pushes back:

A waitlist proves interest, not payment. Ask for a small deposit instead.
The launch email promises three features. Lead with the one people asked for.
Two weeks is too short to judge churn. Decide now what number would make you stop.

I agree with the first two. Want me to update the plan?

An example chat in Claude

claude · ~/my-app

⏺Wrote plans/csv-exports.md (6 steps)

>Get a second opinion on this plan from the 3 smartest models

⏺modelcompare - list_models (MCP)(sort: "intelligence")

⎿{"models":[{"id":"claude-opus-5-5","intelligence":57.6,…

⏺modelcompare - compare (MCP)(prompt: "Review this implementation plan…", models: ["claude-opus-5-5", "gpt-6-astra", "gemini-4-argon"])

⎿{"ok":true,"total_cost_usd":0.0519,"prompts":[{"url":"https://www.modelcompare.ai/…
… +41 lines (ctrl+o to expand)

⏺

Each reviewer caught something the others missed:

claude-opus-5-53 issues · 1 only it found

gpt-6-astra4 issues · 2 only it found

gemini-4-argon3 issues · 1 only it found

I missed 4 of these. Updating the plan now. $0.05 · 13s

An example second opinion in Claude Code

An example second opinion in Claude CodeExample chats in Claude and Claude Code

Put AI to work in your company

We do forward-deployed engineering and AI consulting: we automate your processes, build AI into your products and take your team to the next level with AI. Picking the right models is just the start.

Tell us about your project