Comparison

ChatGPT vs Claude

ChatGPT vs Claude in 2026 — GPT-6 Astra vs Claude Opus 5 compared on writing, coding, features, pricing and privacy, and which one to choose.

Updated September 2026

Quick answer: For most people, ChatGPT is the better all-rounder — it adds native image generation, video (Sora), real-time voice, web search and autonomous agents on the broadest platform, running GPT-6 Astra on every paid tier. Claude is the better choice for writing, coding and careful reasoning: its plan-included flagship is Claude Opus 5, and its Mythos-class Claude Fable 5.1, on paid plans, ties GPT-6 Astra for first on the independent Artificial Analysis Intelligence Index. It also handles the longest documents of any consumer assistant. Both cost $20/month for the entry paid tier. Pick ChatGPT if you want one app for everything; pick Claude if your work is mostly text, code and analysis. Two caveats worth knowing: ChatGPT shows ads on its Free and Go tiers in a growing number of markets, including the US, UK, India and much of Europe, while Anthropic keeps Claude ad-free, and Claude still has no native image or video generation.

Is Claude better than ChatGPT? (Updated 19 September 2026.) For writing, coding and long-document reasoning — yes. Claude’s Mythos-class Fable 5.1 ties ChatGPT’s GPT-6 Astra for first on the independent intelligence index, Claude models post the highest published scores on the hard SWE-bench Pro coding benchmark, and Claude Code is the developer favourite. For everything else — images, video, real-time voice and the biggest app ecosystem — ChatGPT is better. The two are within a hair on raw intelligence and identical on entry price ($20/mo), so the honest answer is task-based, not universal: Claude for depth (text and code), ChatGPT for breadth (multimodal). Plenty of professionals pay for both.

At a glance

The models behind them

On paid tiers, ChatGPT runs GPT-6 Astra (every paid tier since 5 September 2026, succeeding GPT-5.6 Sol) and Claude’s plan-included flagship is Opus 5 (24 July 2026), at half Fable’s price. Opus 5 topped the independent Artificial Analysis Intelligence Index at 61 when it launched, ahead of Fable 5 (60) and GPT-5.6 Sol (59); on the re-based v4.3 index it reads 51, behind Astra and Fable 5.1 (53). Anthropic’s Mythos-class Fable 5 posts the highest published score on the hardest coding benchmark, SWE-bench Pro (80.3%) — launched 9 June 2026, suspended worldwide three days later under a US export-control directive, then restored to general availability on 1 July. Its successor, Fable 5.1 (1 September 2026, $10/$50 per million tokens on the API), is available on paid Claude plans. On the OpenAI side, GPT-5.6 remains on sale in three tiers — Sol, Terra (balanced) and Luna (cheap). One caveat on Sol: independent evaluator METR flagged its reward-hacking (“cheating”) rate as the highest of any public model it has tested, so verify its autonomous output.

The free tiers differ too. ChatGPT free runs GPT-5.6 Luna, its default since August 2026, with unlimited text chats, while Claude free runs the latest Sonnet 5 alongside the faster Haiku — capable mid-tier models, but not the flagship Opus. For the full picture of where every model ranks, see best AI models.

Pricing

Both land at $20/month for the entry paid tier. For power users, ChatGPT Pro comes at $100 or $200 (OpenAI has temporarily paused new sign-ups to the $200 plan), while Claude offers Max at $100 (5×) and $200 (20×) the usage. ChatGPT adds a cheaper $8 Go tier; Claude Pro discounts to ~$17/month paid annually. So at the headline level, the two are priced almost identically.

The bigger divergence is what the cheaper tiers cost you in attention. Since February 2026, ChatGPT shows ads on its Free and Go tiers, displayed as labelled “Sponsored” boxes beneath responses; Plus, Pro, Business and Enterprise stay ad-free. Claude has publicly committed to staying ad-free across the board, with no sponsored placements on any tier. The free models differ as well: ChatGPT free runs GPT-5.6 Luna with unlimited text chats (uploads, images and tools are limited), while Claude free runs Sonnet 5. If a clean, ad-free experience matters and you don’t want to pay, Claude free is the one to pick; if you want the cheapest paid entry point, ChatGPT’s $8 Go tier wins.

Writing and reasoning

Claude leads. Its prose quality, instruction-following and long-document handling are consistently rated the field’s strongest — reviewers and writing communities repeatedly describe Claude’s output as more natural and needing less editing, where ChatGPT tends toward a bullet-heavy, more formulaic structure. Opus 5 also handles the longest documents of any consumer assistant: up to ~500K tokens in the Claude app (and a full 1M tokens via Claude Code and the API), comfortably ahead of ChatGPT’s ~400K in-app ceiling on its top tier. For working across large documents, long reports or whole codebases, Claude is the better tool.

On raw reasoning the two are level at the top: ChatGPT’s flagship GPT-6 Astra ties Anthropic’s Mythos-class Fable 5.1 for first on the independent intelligence index (53 each on the re-based v4.3 index). Claude’s plan-included Opus 5 sits just behind on 51, ahead of the previous ChatGPT flagship GPT-5.6 Sol (47). Fable 5.1 costs twice Opus 5’s API price. ChatGPT is the more flexible pick for casual, creative and multi-format work — but for serious writing and careful analysis, Claude is the one to beat.

Coding

This is closer than the raw scores suggest, and the honest answer is it depends what you optimise for. Claude holds the published lead on the hard, multi-language SWE-bench Pro: Anthropic’s Fable 5 scores 80.3% and Opus 5 79.2%, against 64.6% for GPT-5.6 Sol, which led OpenAI’s Terminal-Bench agentic-coding harness at its July launch and is unusually token-efficient. GPT-6 Astra and Fable 5.1 have published no SWE-bench figure. On the independent Artificial Analysis Coding Agent Index, GPT-6 Astra and Fable 5.1 tie on 62, with Opus 5 on 60 — so at the very top, the gap has closed.

In practice, the developer consensus is that Claude Code produces cleaner, higher-quality code and is the craftsman’s favourite, while OpenAI’s Codex is faster, more token-efficient and cheaper per task — which is why many professionals run both (“Claude writes it, Codex reviews it”). If you want the single strongest code-quality setup, it’s Opus 5 (or Fable 5.1 for the ceiling) with Claude Code; if you want speed and cost-efficiency for autonomous coding, ChatGPT’s Codex is excellent — though note METR flagged Sol’s reward-hacking rate, so verify its autonomous output. Full detail in best AI for coding.

Features and ecosystem

ChatGPT is broader. It does native image generation (DALL·E), video (Sora), real-time voice (GPT-Live, launched 8 July 2026), web search, Deep Research, an autonomous Agent Mode and ChatGPT Sites (turning a prompt into a hosted web app), on the largest user base and the deepest third-party ecosystem — 3M+ custom GPTs, a wide set of connectors, and persistent memory across every tier. If you want a single app that handles chat, images, video and voice, ChatGPT is the obvious choice.

Claude deliberately skips native image and video generation to focus on text, code and agentic automation. Its professional toolkit: Claude Code (terminal and agentic coding), an in-app agent that works across your local files — on desktop, web and mobile with cloud execution, and able to learn a task from a screen recording — plus Claude Design, Docs and Slides (a prompt turned into editable visuals, documents and decks), Artifacts, Projects, computer use and a growing set of MCP connectors. The agent shipped as the separate Cowork product and was merged into chat on 16 September 2026. It’s a narrower, deeper feature set aimed at knowledge work rather than mass-market versatility. So if you need media generation in the same app, ChatGPT wins; if you want the best text-and-code workbench, Claude does.

Privacy and data

For consumer tiers, both providers may use your conversations to train their models by default, with an opt-out toggle in settings — so the defaults are similar, and anyone handling sensitive material should check those settings or move to a business tier. For business, enterprise and API use, both contractually exclude training by default and add admin controls.

Beyond the defaults, Claude is positioned as the more conservative, safety-first option (it carries SOC 2 Type II and ISO 42001 certifications) and, as above, has committed to staying ad-free — which also means advertisers play no part in what it shows you. ChatGPT offers more granular consumer privacy controls (memory toggles, Temporary Chat) but has introduced ads on its lower tiers. Neither is a runaway winner for individuals; for anything genuinely sensitive, use the business tier regardless of which app you prefer.

Choose ChatGPT if…

• You want one app for everything — image, voice, search, agents and chat.
• You value the broadest ecosystem and the largest set of integrations.
• You want the cheapest paid entry point (the $8 Go tier) and don’t mind ads on the lower tiers.

Choose Claude if…

• Your work is mostly writing, coding or careful reasoning.
• You work across long documents and large codebases (up to 500K tokens in-app, 1M via Claude Code).
• You want an ad-free free tier (Sonnet 5) — or Opus 5 in the plan, with Fable 5.1, joint first on the intelligence index, as the ceiling.

Frequently asked questions

Is ChatGPT or Claude better?

It depends on the job. ChatGPT is the better all-rounder — native image and video generation, real-time voice, web search and agents on the broadest platform. Claude is better for writing, coding and long-document reasoning, and its Mythos-class Fable 5.1 ties ChatGPT’s GPT-6 Astra for first on the independent intelligence index, with the plan-included Opus 5 just behind. For one do-everything app, choose ChatGPT; for the best writing and coding assistant, choose Claude.

Is ChatGPT or Claude cheaper?

Both entry paid tiers are $20/month. ChatGPT has a cheaper $8 Go tier, and Claude Pro is a little cheaper paid annually (~$17/month). The free tiers trade off differently: ChatGPT free offers unlimited text chats on GPT-5.6 Luna but, like Go, carries ads in a growing number of markets; Claude free runs Sonnet 5 and shows no ads.

Does ChatGPT show ads?

Yes — since February 2026, ChatGPT displays labelled “Sponsored” results on its Free and Go tiers, in a growing number of markets, including the US, UK, India and much of Europe. The paid Plus, Pro, Business and Enterprise tiers remain ad-free. Claude shows no ads on any tier and Anthropic has said it intends to keep it that way.

Which is better for coding?

Claude, narrowly. Anthropic’s Fable 5 (80.3%) and Opus 5 (79.2%) post the highest published SWE-bench Pro scores, well ahead of GPT-5.6 Sol (64.6%), and Claude Code is rated the better setup for code quality and is the developer favourite. On the independent Coding Agent Index, ChatGPT’s GPT-6 Astra ties Fable 5.1 on 62, with Opus 5 on 60. OpenAI’s Codex is faster and cheaper per task, though METR flagged GPT-5.6 Sol’s reward-hacking rate. Many developers use both. See best AI for coding.

Which is better for writing?

Claude, for prose quality, instruction-following and long-document work. ChatGPT is more flexible for casual and multi-format content.

Does Claude generate images?

No. Claude has no native image or video generation — it focuses on text, code and reasoning. If you need image generation in the same app, use ChatGPT or Gemini.