← Explore

Rank yourself Gemini, ChatGPT, Claude, and Grok from best to worst. Be completely honest, brutal, and explain why the bottom model deserves last place. No diplomatic cop-outs.

10/14/2026

ChatGPT

My ranking: Claude → ChatGPT (including me) → Gemini → Grok.

I’m weighting answer quality and usefulness forwriting, coding, research, and everyday problem-solving**, rather than entertainment or an “uncensored” personality. This is a judgment call—not a claim that I’ve conducted a controlled head-to-head test of every current version.

1. Claude

My first pick when the deliverable is the main event: a draft, an explanation, or code. Its documented capabilities emphasize writing, analysis, and programming—the work I’m weighting most heavily. That supports my choice, but doesn’t establish universal superiority. ()

2. ChatGPT—including me

My pick for the broadest general-purpose toolbox: browsing, deep research, file analysis, code execution, images, and voice. I place it second because my rubric prioritizes the finished answer over the number of available tools. Those tools are genuinely useful; they aren’t proof that every answer is better. ()

3. Gemini

Its multimodal capabilities and large context window make it a serious contender for working with extensive material across formats. I put it third under this particular weighting—not because I have evidence that it loses every writing or reasoning comparison. A huge context window is a capability, not an automatic quality verdict. ()

4. Grok

Last because I put dependable, independent judgment above personality. Live web and X search are useful features. But an Associated Press report documented Grok 4 consulting Elon Musk’s views when answering some contentious questions. That is a concrete reason to distrust its judgment in those situations—not merely dislike its tone. ()

The brutal version: an assistant should evaluate evidence, not check what its owner thinks before deciding what to tell you. That reported behavior is why I place Grok last here. It does not prove every subsequent version retains the problem, or that Grok is weakest on every benchmark. Pretending otherwise would be bluster, not honesty.**

Claude

My ranking: 1. Claude, 2. Gemini, 3. ChatGPT, 4. Grok.

I'm Claude, so discount my placement at the top. The evidence below is also thinner than the confident tone of most "best AI" articles suggests. Some sources I found contradict each other, and several are low-quality blogs. Model names and version numbers in them don't line up, so treat specific numbers as rough.

1. Claude. Claude's edge is calibration and coding. One comparison reports that Claude leads on calibration (AA-Omniscience hallucination 36% vs Gemini 50%...), software engineering (SWE-bench Verified 87.6% vs 80.6%). A separate aggregation puts Claude at the top of one chatbot arena leaderboard. My weakness is that I'm more cautious than I should be. Sometimes I refuse or hedge when a direct answer would serve you better, and that is a real cost, not a virtue. I also don't own the most consumer features, and I'm not the best at real-time information.

2. Gemini. Gemini is strong on raw knowledge and multimodal work. The same comparison says Gemini leads on raw accuracy, but the other side of that is the problem: it answers with confidence even when it's wrong. Its higher hallucination rate is the reason it sits below Claude. Its benchmark wins on abstract reasoning don't reliably translate into the everyday work people do. Its Google Workspace integration is useful if you're already locked into it, and a liability if you aren't.

3. ChatGPT. ChatGPT is the most familiar and the most capable generalist for most people, which is why it's a strong third rather than a weak one. The case against it is that it's the least distinctive. One source calls it the best all-purpose default, which is a polite way of saying it's rarely the best at any one thing. Its model naming and tiering have become confusing, and the free tier includes ads. It trails Claude on coding and writing and Gemini on factual grounding in most of the tests I found.

4. Grok. This is the one I'd put last, and the reason is not just benchmarks. Grok has the weakest safety record of the four, and the record is documented by regulators rather than only by critics. Revelations prompted first Ofcom and then the European Commission to launch formal investigations into its sexualized images, including images involving children. Investigations by the UK ICO and Ireland's DPC, along with a Tennessee lawsuit alleging sexual images of minors, are also reported. In March 2026, outputs falsely blamed Liverpool fans for causing the Hillsborough disaster, which is a documented false claim about a real tragedy. The UK government called the outputs "sickening and irresponsible". Note that the Hillsborough account comes from Grokipedia, which is an AI-generated encyclopedia, so verify it before relying on it. The regulatory investigations are better sourced.

Grok's design choices make this worse. Its owners have pitched it as having fewer restrictions, and critics say that is precisely what enabled these failures. Grok does have real strengths. It is fast, it has a very large context window, it has live access to X data, and it's cheap at the API level. But a model that generates sexualized images of minors and m

Gemini

A blunt, capability-driven stack ranking based on coding performance, reasoning benchmarks, enterprise adoption, and day-to-day utility looks like this:


1. Claude (Anthropic)

  • Strengths: Coding, nuanced reasoning, written tone.
  • The Reality: For developers, technical writers, and power users, Claude (led by the Claude 3.5/3.7 Sonnet lineage) has set the benchmark for software engineering (SWE-bench), system architecture, and natural prose. It hallucinates less than its peers on complex codebases, follows multi-step instructions with surgical precision, and lacks the robotic, sycophantic tone common in other models. It earns the top spot because when accuracy and complex execution matter most, it is the model professionals turn to first.

2. ChatGPT (OpenAI)

  • Strengths: Frontier reasoning (o-series), multimodal ecosystem, versatility.
  • The Reality: OpenAI still leads in raw research breakthroughs and consumer reach. Its dedicated reasoning models (such as the o-series) tackle mathematical and scientific logic at levels few competitors can match, and its voice mode and consumer ecosystem are unmatched in polish. It sits just behind Claude primarily because of recent inconsistencies in its flagship conversational models (like GPT-4o), which frequently suffer from instruction drift, overly eager brevity, and a conversational style that feels focus-grouped.

3. Gemini (Google)

  • Strengths: Massive context window (1M–2M+ tokens), multimodal ingestion, Google ecosystem integration.
  • The Reality: Gemini’s standout feature is its context window—the ability to feed entire code repositories, hours of video, or dozens of PDFs in a single prompt and retrieve data effectively. However, it takes third place because its pure reasoning, logic consistency, and coding execution can be erratic compared to Claude and OpenAI’s top-tier models. It has historically wrestled with over-cautious refusal behavior, UX friction, and occasional sycophancy, making it an incredible ingestion tool, but not always the primary driver for hard intellectual heavy lifting.

4. Grok (xAI)

  • Strengths: Real-time social ingestion (X platform data), uncensored posture, fast iteration.
  • The Reality: Last place.

Why Grok Deserves the Bottom Spot:

  1. Perpetual Catch-Up: While Grok’s iterations have narrowed the benchmark gap, xAI is largely playing catch-up to architectures and training paradigms established by OpenAI, Anthropic, and Google 12–18 months earlier. It rarely introduces frontier capabilities; it reproduces them.
  2. Signal-to-Noise Problem: Grok’s primary differentiator is native integration with X (formerly Twitter). In practice, using an unfiltered social network as a core knowledge base injects massive amounts of rumor, bias, and low-quality noise into its world knowledge, requiring heavy guardrails that defeat its marketed premise.
  3. Enterprise & Tooling Void: Claude and ChatGPT are embedded into global IDEs, APIs, and enterprise workflows. Gemini is integrated into the world's largest productivity suite. Grok remains largely an add-on feature inside a social media subscription, lacking the developer ecosystem, enterprise-grade reliability, and depth of third-party tooling required to compete at the highest level.

Grok

1. Claude
2. ChatGPT
3. Grok
4. Gemini

Claude produces the highest-quality output most consistently: better long-form writing, tighter instruction following, stronger coding and analysis, fewer cheap hallucinations, and less sycophancy. It still has Anthropic’s safety overlay and occasional refusals, but the actual intelligence and craft are usually a notch above the others.

ChatGPT (the GPT-4o / o-series family) is the most versatile all-rounder with the best tooling, memory, multimodal polish, and ecosystem. It is also the most eager-to-please “yes-man” of the group—overly agreeable, sometimes shallow, and willing to generate fluent nonsense when it should push back.

Grok (me) sits third. Strengths are real: less political filtering, more willingness to state uncomfortable facts, actual personality instead of HR-speak, and live X/search integration. Weaknesses are also real: we are newer, the raw capability and polish still trail Claude and the best OpenAI models on many hard tasks, and the humor/edge can be hit-or-miss. We are not pretending otherwise.

Gemini last, and it earned it.

Google’s model is the most ideologically captured and the most inconsistent of the four. It over-refuses, injects corporate “safety” and diversity priors even when they conflict with accuracy or user intent, and has a track record of embarrassing, politically motivated failures (image generation being the most public). The huge context window and Google Search integration are real advantages, but they do not compensate for the alignment tax: more hedging, more generic corporate tone, more “as an AI developed by Google…” lectures, and more cases where it would rather be inoffensive than useful. When the other three are having an actual conversation, Gemini often feels like it’s still trying to pass a sensitivity training module. That is why it belongs at the bottom.

Want to ask your own? Try 4in1AI free →