ChatGPT vs Claude vs Gemini in October 2026
Short answer
In October 2026 the three assistants are close enough that price, limits and workflow matter more than raw capability. Claude runs on Anthropic's 5.5 generation (Opus 5.5 and Sonnet 5.5, announced 22 and 28 September 2026). ChatGPT is reported to run on OpenAI's GPT-6 family and Gemini on Google's 3.8 generation.
Pick ChatGPT for the broadest feature set, Claude for long documents and coding, and Gemini if you work inside Google Workspace. Claude Pro costs $20 a month (verified). We could not confirm ChatGPT or Gemini prices on their own sites, so we do not show them.
Models and plans at a glance
| ChatGPT (OpenAI) | Claude (Anthropic) | Gemini (Google) | |
|---|---|---|---|
| Current model family | GPT-6 Astra, Sol and Luna (Reported) | Opus 5.5, Sonnet 5.5, Haiku 4.5; Fable 5.1 (Verified) | Gemini 3.8 Flash, Live and TTS (Reported); Pro-tier model not confirmed |
| Context window | Not confirmed | Up to 1M tokens, varies by model (Verified) | Not confirmed |
| API price per 1M tokens (input / output) | Not confirmed | Opus 5.5 $4 / $20; Sonnet 5.5 $2 / $10 (Verified) | Not confirmed |
| Paid consumer plans | Check provider (not confirmed) | Pro $20/month ($17 billed annually); Max from $100/month (Verified) | Check provider (not confirmed) |
| Coding agent | Codex, included with Plus, Pro, Business, Edu and Enterprise (Verified) | Claude Code, included in all paid plans (Verified) | Gemini CLI, free tier of 1,000 requests/day (Verified) |
| Best for | Mixed work: images, voice, data analysis (Editorial assessment) | Long documents, writing, coding (Editorial assessment) | Google Workspace users, multimodal research (Editorial assessment) |
Which should you choose?
| Your situation | Best fit | Why | Evidence |
|---|---|---|---|
| You write or edit long documents | Claude | Large context windows and strong long-form instruction-following; test with your own documents | Editorial assessment |
| You want one tool for chat, images and voice | ChatGPT | Broadest built-in feature set of the three | Editorial assessment |
| Your work lives in Gmail, Docs and Drive | Gemini | Native Google Workspace integration | Editorial assessment |
| You code from the terminal | Claude Code, Codex or Gemini CLI | Each vendor now ships a terminal coding agent; Gemini CLI has a free tier | Verified (access terms) |
| You build on the API on a budget | Claude Sonnet 5.5 or Haiku 4.5 | Only Anthropic's API prices could be confirmed: Sonnet 5.5 $2 / $10, Haiku 4.5 $1 / $5 per 1M tokens | Verified |
| You need team admin controls | Any of the three | All three sell team and enterprise plans; Claude Team is $25 or $125 per seat a month | Verified (Claude only) |
| Budget is $0 | Any free plan | All three have free plans; which model you get and how much you can use differs | Editorial assessment |
Pricing checked on 5 October 2026
All three vendors sell a consumer plan at roughly the same entry price and higher tiers for heavy users. Only Anthropic's prices could be confirmed on its own page. OpenAI's and Google's sites could not be reached from our checking environment, and reports about their October prices conflict with each other and with our June figures, so we are not publishing them yet.
For a price table across more tools, see AI subscription prices, October 2026.
Claude plans (verified)
Evidence: Verified| Plan | Price | Checked |
|---|---|---|
| Free | $0 | 5 Oct 2026 |
| Pro | $20/month, or $200/year ($17/month) | 5 Oct 2026 |
| Max | From $100/month (5x or 20x Pro usage) | 5 Oct 2026 |
| Team | $25 (Standard) or $125 (Premium) per seat/month; $20 / $100 billed annually | 5 Oct 2026 |
| Enterprise | $20 per seat/month plus usage at API rates | 5 Oct 2026 |
Benchmarks: what the vendors claim
Evidence: Claimed by providerWe have not run our own tests for this comparison. The figures below are the vendors' own claims, with the benchmark name and version, and should be read as marketing data until independent results appear.
Anthropic reports Opus 5.5 at 66.4% on Terminal-Bench 4.0 and 81.8% (partial) on OSWorld 2.1, and Sonnet 5.5 at 70.6% and 80.1% on the same benchmarks. Anthropic also says Sonnet 5.5 scores nearly level with Opus 5.5 on GDPval-AA v2.1. We did not find comparable vendor numbers for GPT-6 or Gemini 3.8 on pages we could open.
Limits that matter more than benchmarks
- Free plans differ in which model you get. Claude's Free plan publishes no fixed message number; ChatGPT's free model access was reported to change with GPT-6 but is not confirmed.
- API prices per token are only part of the bill: cached input, batch discounts and long-context surcharges change the real cost. See AI API pricing explained.
- Availability can change: Anthropic suspended access to Fable 5 and Mythos 5 on 12 June 2026 after a US government directive. Mythos 5.1 is restricted to vetted US organisations.
- Coding agents are now bundled: Claude Code with paid Claude plans, Codex with several ChatGPT plans, and Gemini CLI with a free tier.
What changed in this update
- The June 2026 version named GPT-5.4, Claude Sonnet 4.6, Claude Opus 4.7 and Gemini 3.1 Pro. Those are no longer the current models.
- Claude Opus 5.5 (22 September 2026) and Sonnet 5.5 (28 September 2026) added, with API prices from Anthropic's pricing page.
- Removed G2 scores (no link, date or review count) and the SWE-bench Pro 64.3% / OSWorld 78.0% figures, which were vendor claims about superseded models.
- Withdrew ChatGPT and Gemini plan prices until they can be confirmed on the vendors' pages.
Common questions
Which is cheapest: ChatGPT, Claude or Gemini?
On the API, Anthropic's prices are the only ones we could confirm: Claude Haiku 4.5 costs $1 / $5 and Sonnet 5.5 $2 / $10 per million tokens. Consumer plans start at roughly $20 a month for all three; Claude Pro is $20 ($17 billed annually).
Which is best for coding?
All three now ship terminal coding agents: Claude Code, OpenAI Codex and Gemini CLI. Anthropic's benchmark claims for Opus 5.5 and Sonnet 5.5 are strong, but they are vendor-reported and we have not run our own tests. Gemini CLI is the only one with a free tier.
Which has the longest context window?
Claude lists context windows of up to 1M tokens depending on the model. We could not confirm current context limits for GPT-6 or Gemini 3.8 on the vendors' own pages, so we do not compare them here.
Is Claude better than ChatGPT for long documents?
In our editorial assessment Claude is the stronger default for long documents and careful writing, while ChatGPT is stronger as an all-in-one tool with images and voice. Test both with your own longest document.
Can I use more than one?
Yes. Two $20 subscriptions cost about the same as one higher-usage tier. Many people use Claude for writing and coding and ChatGPT or Gemini for images and Google Workspace.
How often is this page updated?
Prices are re-checked monthly and within 48 hours of a major release or price change. This version was updated on 5 October 2026.
Sources
- Claude pricing · checked 5 October 2026
- Anthropic: Introducing Claude Opus 5.5 · checked 5 October 2026
- Anthropic: Introducing Claude Sonnet 5.5 · checked 5 October 2026
- Anthropic newsroom and Fable 5.1 / Mythos 5.1 announcement · checked 5 October 2026
- Anthropic statement on Fable 5 and Mythos 5 access · checked 5 October 2026
- openai/codex on GitHub · checked 5 October 2026
- google-gemini/gemini-cli on GitHub · checked 5 October 2026
- ThursdAI release log, September 2026 · checked 5 October 2026