ChatGPT vs Gemini
GPT-6 Sol in ChatGPT Plus and Gemini 4 Argon in Google AI Pro are the matchup a $20 subscriber gets in each chat app. As of November 13, 2026, Gemini is ahead of ChatGPT on public benchmark score: Gemini 4 Argon scores 79.03 at #8, against GPT-6 Sol at 77.41 and #10. The lead is not settled: Gemini 4 Argon's 90% interval (75.4–82.7) overlaps GPT-6 Sol's (74.0–80.8). By job, Gemini 4 Argon has the higher coding estimate and GPT-6 Sol the higher knowledge estimate.
We compare the models each app serves on its free and paid plans. Benchmarks test the model, not the app, so upload limits and tools still deserve a test of their own.
Both $20 plans now have a flagship model
Google added Gemini 4 Argon to the Gemini app for Google AI Pro and Ultra subscribers on October 13, 2026. That changed this comparison. Until then, Argon was a model most Gemini users couldn’t open, and a subscriber’s fair matchup was a step down Google’s lineup. ChatGPT Plus runs GPT-6 Sol in Chat. Google AI Pro gets Gemini 4 Argon, plus Google’s Pro and Flash models.
On the overall reading, Gemini 4 Argon is ahead at 79.03, with GPT-6 Sol at 77.41, ranked #8 and #10. An overall score blends many benchmarks, and it hides the split you actually care about: the two models lead on different jobs. GPT-6 Astra ranks higher, at 85.65 and #4, but on Plus it is an option in ChatGPT Work and Codex, not in Chat.
Free plans are a thinner contest. ChatGPT Free and Go run a Luna model, but OpenAI’s pricing page lists GPT-6 Luna and its help center names GPT-5.6 Luna, so we don’t match either plan to a ranked row. Gemini’s free plan runs what Google calls “our Flash-Lite model” without naming the version. That leaves no free-tier winner for us to call.
Which app wins depends on the job
| Job | ChatGPT Plus chat model | Google AI Pro model | Reading as of November 13, 2026 | Basis |
|---|---|---|---|---|
| Coding | GPT-6 Sol67.3 · #13 · Supported | Gemini 4 Argon71.8 · #9 · Supported | Gemini 4 Argon has the higher point estimate. Conditional ranges do not establish rank confidence. | Like-for-like |
| Agents and automation | GPT-6 Sol65.1 · #18 · Supported | Gemini 4 Argon74.9 · #6 · Supported | Gemini 4 Argon has the higher point estimate. Conditional ranges do not establish rank confidence. | Like-for-like |
| Research and knowledge | GPT-6 Sol80.2 · #10 · Estimated | Gemini 4 Argon79.2 · #8 · Supported | GPT-6 Sol has the higher point estimate. Conditional ranges do not establish rank confidence. | Directional |
| Reasoning | GPT-6 Sol72.3 · #14 · Supported | Gemini 4 Argon76.5 · #7 · Supported | Gemini 4 Argon has the higher point estimate. Conditional ranges do not establish rank confidence. | Like-for-like |
Coding
On coding, Gemini 4 Argon scores 71.8 at #9, and GPT-6 Sol scores 67.3 at #13. The coding score ranges are unvalidated, so read that order as a lean, not a settled rank. Our coding table scores benchmark tasks run in each lab’s own setup. It doesn’t measure the editor, repository access, or agent you wrap around the model, so the test that settles it for you is one real change, made in the tool you already use. If you code in Codex instead of the chat window, Plus also includes GPT-6 Astra there, which ranks #5 on coding at 78.7.
Research and long documents
On research and knowledge, GPT-6 Sol scores 80.2 at #10, and Gemini 4 Argon scores 79.2 at #8. The research and knowledge score ranges are unvalidated, so read that order as a lean, not a settled rank. Our knowledge table scores answers to hard factual questions. It doesn’t check whether a research answer cites the right source. For long files, the plan’s input limit can decide before the model does. Google lists a 1 million token context window for Gemini on AI Pro. OpenAI lists 256K for reasoning models on ChatGPT Plus.
Writing and images have no measured winner yet
We don’t have a public rank for both models on writing or reading images and PDFs, so we don’t name a winner there. GPT-6 Sol ranks #8 on reading images and PDFs, and Gemini 4 Argon has no public rank there yet. Neither GPT-6 Sol nor Gemini 4 Argon has a public rank on writing yet. A missing public rank is a coverage limit: too few published results for us to place the model. It doesn’t mean the model failed.
To settle writing for your own work, give both apps one real brief with its source material, then count the edits each draft needs before you’d send it. For images and PDFs, ask about a chart’s units, a scanned page, and a question the file can’t answer. Our multimodal table covers reading files, not generating images or video.
The $20 plans cost the same and bundle different limits
US prices and plan contents, checked on October 23, 2026.
| Plan level | ChatGPT | Gemini |
|---|---|---|
| Free | $0. A Luna model, with limited uploads, image creation, and deep research. | $0. Google’s Flash-Lite model and 15 GB of storage. |
| Entry | Go, $8 a month. A Luna model, with more messages and uploads. May include ads. | Google AI Plus, $4.99 a month. Google’s Flash model, twice the free plan’s usage, and 400 GB of storage. |
| Around $20 | Plus, $20 a month. GPT-6 Sol in Chat, with a 256K context window for reasoning. GPT-6.1 Sol and GPT-6 Astra in ChatGPT Work and Codex. | Google AI Pro, $19.99 a month. Gemini 4 Argon, Google’s Pro and Flash models, four times the free plan’s usage, a 1M-token context window, and 5 TB of storage. |
| Top | Pro, from $100 a month, in $100, $200, and $500 tiers. Pro reasoning on GPT-6 Astra, with a 400K reasoning context window. | Google AI Ultra, from $99.99 a month, or $199.99 for 20 times AI Pro’s usage. Highest access to Gemini 4 Argon, and from 20 TB of storage. |
At the $20 level the price is a wash: one cent apart. What differs is the bundle. Google AI Pro includes 5 TB of cloud storage and a context window about four times ChatGPT Plus’s reasoning limit. ChatGPT Plus adds GPT-6.1 Sol and GPT-6 Astra in ChatGPT Work and Codex, outside the chat window.
Gemini is cheaper at the entry level. Google AI Plus costs $4.99 against $8 for ChatGPT Go, and OpenAI says Go may include ads. Both top plans start near $100, and ChatGPT Pro runs up to $500 a month.
One clarification, though: these are app plans. They don’t include API use, which is billed per token and priced further down.
ChatGPT vs Gemini: model evidence
These are the model rows behind each app’s lineup. Scores and ranks come from our main leaderboard, and the preference column shows a human-preference rating from an independent index where one is published.
| Model | Score | Preference rating | Context |
|---|---|---|---|
| GPT-6 SolOpenAI | 77.41 | 1456.8 | 1.05M |
| GPT-6 AstraOpenAI | 85.65 | 1474.1 | 1.05M |
| GPT-6.1 SolOpenAI | 84.6 | 1482.1 | 1.05M |
| GPT-6.1 TerraOpenAI | 79.64 | — | 1.05M |
| GPT-6 LunaOpenAI | 61.13 | 1440.8 | 1.05M |
| Gemini 4 ArgonGoogle | 79.03 | 1522.4 | 1M |
| Gemini 4 CarbonGoogle | 89.15 | 1526.0 | 1M |
| Gemini 3.8 FlashGoogle | 64.47 | 1497.5 | 1M |
| Gemini 3.1 ProGoogle | 63.11 | 1486.7 | 1M |
Gemini 4 Argon reached the Gemini app on October 13, 2026, for Google AI Pro and Ultra subscribers, according to Google's announcement. Google’s release notes describe that rollout as global.
For the benchmark-by-benchmark detail, see Gemini 4 Argon vs GPT-6 Sol.
API rates for the same models
If you call these models through an API instead of using the apps, you pay per token, and the price order can flip. Gemini's top-ranked model is the cheaper one: Gemini 4 Carbon at $2/$10 per million input/output tokens, against GPT-6 Astra at $10/$50. Gemini 4 Carbon's listed rate carries a condition: introductory rate through Nov 30, 2026; $4 / $20 from Dec 1. At the budget tier the order flips: GPT-6 Luna ($0.10 / $0.50) undercuts Gemini 3.8 Flash ($0.75 / $3.75). Context tiers, caching, and tool charges change the final bill. Full rate cards are on the OpenAI API pricing and Google API pricing pages.
| Model | Input / 1M tokens | Output / 1M tokens | Note |
|---|---|---|---|
| GPT-6 Sol | $2.00 | $10.00 | — |
| GPT-6 Astra | $10.00 | $50.00 | — |
| GPT-6.1 Sol | $2.00 | $10.00 | — |
| GPT-6.1 Terra | $1.00 | $5.00 | — |
| GPT-6 Luna | $0.10 | $0.50 | — |
| Gemini 4 Argon | $2.00 | $10.00 | Introductory rate through Nov 30, 2026; $4 / $20 from Dec 1 |
| Gemini 4 Carbon | $2.00 | $10.00 | Introductory rate through Nov 30, 2026; $4 / $20 from Dec 1 |
| Gemini 3.8 Flash | $0.75 | $3.75 | Introductory rate through Dec 31, 2026; $1.50 / $7.50 after |
| Gemini 3.1 Pro | $2.00 | $12.00 | Rises to $4 / $18 above 200K-token prompts |
What each plan includes: ChatGPT, Gemini, and Microsoft Copilot
Scores compare the models. The plan decides what you can do with them, so this table lists what each maker's own plan page includes and the lowest plan it starts on.
Microsoft Copilot is in the last column because many readers already pay for it through Microsoft 365. It has no score on this page. Microsoft doesn’t name the model Copilot runs, so no benchmark result can be attached to it. Its own comparison is on ChatGPT vs Copilot.
| Plans and features | ChatGPT | Gemini | Microsoft Copilot |
|---|---|---|---|
| Free plan | Yes. | Yes. | Yes. |
| Cheapest paid plan | Go, $8 a month. | Google AI Plus, $4.99 a month. | Microsoft 365 Personal, $9.99 a month. |
| Plan nearest $20 | Plus, $20 a month. | Google AI Pro, $19.99 a month. | Microsoft 365 Premium, $19.99 a month. |
| Models the plan names | GPT-6, with “advanced reasoning models” on Plus. | Gemini 4 Argon on Google AI Pro. | None named. Microsoft says Copilot uses “models from Microsoft, OpenAI, and other model developers”. |
| Deep research | Free, with limits. Expanded on Plus. | Free. | From Microsoft 365 Premium, as limited “advanced research and analytics”. |
| Voice chat | Free, with limits. | Free, as Gemini Live. | Not named. Paid plans list AI-generated audio (English only) and help from your screen or camera. |
| Image creation | Free, with limits. | Free, with editing. | Free. Paid plans raise the limit. |
| Memory across chats | Free, with limits. | Not on the plan page. Google’s privacy hub describes a memory summary you can turn off. | Not on the plan page. Microsoft’s privacy page describes Copilot Memory, which you can turn off. |
| Projects | Listed on Plus, with scheduled tasks and custom GPTs. | Not listed. Free includes Gems and Canvas. | Not listed on the plan page. |
| Scheduled tasks | Listed on Plus. | Not listed below Ultra. Ultra lists first access to “Tasks in Gemini”, in some countries. | Not listed on the plan page. |
| Coding agent | Codex: limited on Free, expanded on Plus. | Google Antigravity, from Google AI Pro at entry rate limits. | Not listed. GitHub Copilot is a separate product. |
| Inside office apps | Extensions for Excel, Word, PowerPoint, and Google Sheets, limited on the lower plans. | Gmail and Vids from Google AI Plus. Docs from Google AI Pro. | Word, Excel, PowerPoint, Outlook, and OneNote, from Microsoft 365 Personal. |
| Cloud storage included | None listed. | 15 GB on Free, 400 GB on Google AI Plus, 5 TB on Google AI Pro. | 1 TB on Microsoft 365 Personal, 2 TB on Premium. |
Each cell is what the maker’s own plan page lists and the lowest plan it lists it on. “Not listed” means the page doesn’t mention the feature, which isn’t proof the app lacks it. Plan prices were read October 16, 2026 for ChatGPT, November 10, 2026 for Copilot, and November 11, 2026 for Gemini.
Sources, read November 10, 2026: ChatGPT pricing; Google AI plans, Gemini Apps privacy hub; Microsoft Copilot plans for individuals, Microsoft Copilot transparency note, Microsoft Copilot: your privacy controls.
What each app does with your chats
Read the training row before you paste in anything private. The defaults differ, and so does the place where you change them.
| Your chats | ChatGPT | Gemini | Microsoft Copilot |
|---|---|---|---|
| Do your chats train the models? | Yes by default on personal plans. | Yes while Keep Activity is on, and human reviewers read a subset of chats. | No. Microsoft says prompts, responses, and file contents “aren’t used to train foundation models”. |
| How to stop it | Turn off “Improve the model for everyone” under Data controls. Voice mode and Codex have their own settings. | Turn off Keep Activity. Chats already reviewed are kept for up to three years. | Nothing to turn off for training. Personalization and memory have their own switches. |
| Chats kept out of training | Temporary chats. They may be kept up to 30 days for safety. | Temporary chats. They are kept with your account for 72 hours. | All chats, on Microsoft’s statement above. |
| Ads | The Go plan “may include ads”. | Google says chats “are not being used to show you ads”. | Possible without a Microsoft 365 subscription. Ad personalization can be turned off. |
These terms cover personal accounts. Work, school, and API accounts run under different terms, and each maker can change a default at any time, so open the source before you paste in anything sensitive.
Sources, read November 10, 2026: OpenAI: data controls in ChatGPT, ChatGPT pricing; Gemini Apps privacy hub; Microsoft Copilot: your activity history, Microsoft Copilot: your privacy controls.
Which should you pick?
Pick ChatGPT Plus if…
- Research questions with a single right answer are most of your work: GPT-6 Sol scores 80.2 on knowledge, against 79.2 for Gemini 4 Argon.
- You also use ChatGPT Work or Codex, where Plus adds GPT-6.1 Sol and GPT-6 Astra to the GPT-6 Sol you get in Chat.
- It’s the wrong pick for very long files. Reasoning models on Plus stop at 256K tokens of context, and Gemini on AI Pro lists 1M.
Pick Google AI Pro if…
- Coding happens in the chat window: Gemini 4 Argon ranks #9 on coding at 71.8, against #13 for GPT-6 Sol, the model in ChatGPT Plus chat.
- You’d use 5 TB of storage anyway, or you need a long document in a single prompt.
- It’s the wrong pick if factual research is the main job, where GPT-6 Sol has the higher knowledge score. If neither plan fits, see our Gemini alternatives.
We read each plan's model from the public leaderboard
We pick the model each plan serves, then read its rows from our public leaderboard: the overall score, the task lanes, and the evidence label. Supported means we have enough independent evidence to stand behind the row. Estimated means the rank is visible and the uncertainty is wider. Plan prices and model lists come from OpenAI’s and Google’s plan pages, checked on October 23, 2026. API rates come from each provider’s rate card as recorded in our pricing data. A model without a public rank shows as unranked instead of getting an old score.
Questions
Is Gemini better than ChatGPT?
Yes, on our overall table as of November 13, 2026: Gemini 4 Argon scores 79.03 at #8, and GPT-6 Sol scores 77.41 at #10. It depends on the job, though. Gemini 4 Argon has the higher coding estimate and GPT-6 Sol the higher knowledge estimate. For writing, we have no measured winner.
Is Gemini or ChatGPT more truthful?
We don’t measure truthfulness in either app. The closest evidence is our knowledge table, which scores answers to hard factual questions. GPT-6 Sol has the higher score there, 80.2 to 79.2. That tells you which model got more test questions right. It doesn’t tell you how often either app states something false in your own chats.
Is Gemini cheaper than ChatGPT?
By one cent at the $20 level: Google AI Pro costs $19.99 a month, and ChatGPT Plus costs $20. Entry plans differ more, with Google AI Plus at $4.99 against $8 for ChatGPT Go. API rates are a separate bill. Only at the top of each lineup. Gemini 4 Carbon, Google's best-ranked model here, costs $12 per million input plus output tokens against $60 for GPT-6 Astra, while GPT-6 Luna, at $0.60, is the cheapest model on the page.
Which is better for coding, ChatGPT or Gemini?
Gemini, on the public coding ranking: Gemini 4 Carbon ranks #2 at 83.88, against GPT-6 Astra at #5 with 78.7. Gemini 4 Carbon leads by point estimate; its evidence label is Supported. The conditional ranges do not establish rank confidence. In ChatGPT Plus chat, GPT-6 Sol ranks #13 on coding.
Can I use both ChatGPT and Gemini?
Yes. Both apps have free plans, so you can run the same task in each before paying for either. Keep the source material, instructions, and review the same, and pick whichever app’s output needs fewer edits. Paying for both $20-level plans costs $39.99 a month.