Independent evaluations of AI tools, from coding to research. Amazon Bedrock AgentCore Browse, Descript compare, and Midjourney read the method.

Searches this site only. Ratings are the editor's, from public information, and are not test results.

Tool directory

38 AI tools, compared

What each tool is built for, the editor's rating and the evidence behind it.

Showing 7 of 38

Sort
AI tools with what each is built for, the editor's rating out of 10 and the evidence recorded. Ordered by editor's rating, highest first. The rating is an editorial judgment from public information, not a test result.
ToolBuilt forEditor's ratingEvidenceReview
GrokxAI · General-purpose AI assistantsxAI's AI assistant for chat, live web and X search, image and video generation and voice, with a paid SuperGrok tier and an API.9.0 out of 10
Full review4 published results
→
Claude CodeAnthropic · Coding & developmentAnthropic's agentic coding tool for working on a codebase.8.7 out of 10
Full review3 published results
→
OpenAI CodexOpenAI · Coding & developmentOpenAI's software-engineering agent and coding tool.8.5 out of 10
Full review3 published results
→
GeminiGoogle · General-purpose AI assistantsGoogle's AI assistant app, with a free tier and Google AI Plus, Pro and Ultra plans bundled with Google storage.8.3 out of 10
Full review4 published results
→
ClaudeAnthropic · General-purpose AI assistantsAnthropic's AI assistant for chat and everyday work, on web, desktop and mobile.8.2 out of 10
Full review3 published results
→
CursorAnysphere · Coding & developmentAn AI-assisted code editor from Anysphere.8.1 out of 10
Full review3 published results
→
ChatGPTOpenAI · General-purpose AI assistantsOpenAI's AI assistant for chat, writing, images, voice, deep research and work tasks, with free, individual and business plans.8.0 out of 10
Full review4 published results
→

Ordered by editor's rating. Ratings are the editor's judgment from public information. We have not tested these tools, and published results are other people's tests, recorded with their limits. How each rating is built · Leaderboards · Full queue

Guides by need

AI assistants for writing, reviewing, and maintaining software.

Quick answer · Editors’ pick (research-based)

Which AI app should you use?

Pick what you need to get done. We name the one to start with, and a runner-up.

Start with
ChatGPT

One assistant for writing, research, files, images and code. The safest first install.

Runner-up
Claude
Free plan
Yes, with limits
See why ChatGPT Read the full guide to 15 AI apps

Editors’ picks are research-based; they carry no scores. Prices and free limits change, so check the official page before you pay.

Model finder · 3 questions

Which AI models should you test?

Answer three questions: what you are building, what it must handle, and how much it will run. We screen 15 models and give you a short list to test, with an estimated monthly cost and the gaps in what we know.

Open the full finder

No sign-up. Nothing leaves your browser. A shortlist to test, not a ranking. Want to compare prices instead? Try the cost calculator or browse every model.

  • Up to three models to testEverything that cannot meet your needs is ruled out first.
  • A monthly cost estimateFrom list prices and your own volume. Unknowns are named, not guessed.
  • A test plan for this weekPrompts to collect and what to score, matched to your task.

Ratings are the editor's, from public information. No tool has been independently tested yet.

Make the trade-offs visible

BestSuperintelligent compares tools by the work, not the hype.

GitHub Copilot vs Cursor and Claude Code vs OpenAI Codex are written as questions, not verdicts. A winner needs published test conditions and linked evidence.

  1. Coding assistantsPlanned

    GitHub Copilot vs. Cursor

    Compare editor workflow, codebase context, integrations, controls, and total cost.

    • Capability
    • Workflow fit
    • Controls and privacy
    • Total cost
    Open the framework
  2. Agentic codingPlanned

    Claude Code vs. OpenAI Codex

    Compare task execution, reviewability, environment fit, and developer control.

    • Capability
    • Reliability
    • Control and permissions
    • Environment fit
    Open the framework
  3. Research assistantsPlanned

    Perplexity vs. NotebookLM

    Perplexity searches the web; NotebookLM answers only from sources you upload. Pricing, security evidence, published test results and which job fits which, side by side, plus the planned head-to-head test.

    • Source traceability
    • Accuracy of summaries
    • Handling of uncertainty
    • Coverage and recency
    Open the framework
  4. Writing and productivity assistantsPlanned

    Claude vs. Notion AI

    Compare output quality, editing effort, workflow fit and data controls on the same everyday writing tasks.

    • Output quality for the task
    • Workflow integration
    • Admin and data controls
    • Price at realistic usage
    Open the framework
  5. Image and video generatorsPlanned

    Midjourney vs. Runway

    Compare prompt adherence, consistency across attempts, directed editing and usage terms on the same briefs.

    • Control and consistency
    • Output quality
    • Usage rights and provenance
    • Production workflow
    Open the framework
  6. Text-to-speech voiceoverPlanned

    ElevenLabs vs. Murf AI

    Compare pronunciation accuracy, voice quality, language coverage and consent safeguards on the same scripts.

    • Pronunciation accuracy
    • Voice quality
    • Language coverage
    • Consent and safeguards
    Open the framework
  7. Speech-to-text transcriptionPlanned

    Deepgram vs. Otter.ai

    Compare word error rate across accents, noise and specialist vocabulary, plus speed and cost per audio hour.

    • Accuracy
    • Language coverage
    • Privacy of recordings
    • Speed and cost
    Open the framework
  8. Natural-language analyticsPlanned

    Databricks Genie vs. Snowflake Cortex Analyst

    Compare answer correctness, query transparency and handling of ambiguity and messy data on a dataset with known answers.

    • Calculation correctness
    • Data access and governance
    • Transparency
    • Handling of messy data
    Open the framework
  9. Enterprise adoptionPlanned

    Hosted vs. self-managed deployment

    Understand operational burden, data controls, governance, and procurement trade-offs.

    • Operational burden
    • Data control
    • Governance and audit
    • Procurement fit
    Open the framework

Planned topics, not completed reviews. No winner is declared until evidence supports a conclusion.

All comparisons

Frequently asked questions

How do I choose which AI model to test?

Use the Model Finder. Answer questions about what you are building, which inputs it must handle, how much it will run and your budget. It rules out models that cannot meet those needs and returns up to three to test, with an estimated monthly cost. It is a shortlist, not a ranking.

What is BestSuperintelligent.com?

An independent research and comparison site for intelligent software. Every product page shows its evaluation status, and every claim is labelled by how well it is supported.

Has BestSuperintelligent published any rankings or scores yet?

No. A ranking requires at least two fully evidenced, published evaluations in the same category, and none has met that threshold yet. The evaluation queue shows each product's current status.

Does the word superintelligent in the name mean these products are superintelligent?

No. It is a brand name, not a claim. We do not state or imply that any product we cover is superintelligent.

How does the site earn money, and does that affect scores?

The site may use disclosed affiliate links or sponsorships in future. Scores are calculated by code that takes no commercial input, and the affiliate disclosure lists any current relationships.

How can I tell whether a claim on this site is reliable?

Each statement carries a label: vendor-reported, independently observed, independently measured, third-party research, editorial interpretation, or unverified. Where possible it links to its source and shows the date we last checked it.

Which AI coding tools does the site cover?

The coding queue names Claude Code, Cursor, Gemini Code Assist, GitHub Copilot, and OpenAI Codex. Each page records vendor statements with a source and date. None of those five has a published independent test on this site yet.