Welcome to the forefront of conversational AI as we explore the fascinating world of AI chatbots in our dedicated blog series. Discover the latest advancements, applications, and strategies that propel the evolution of chatbot technology. From enhancing customer interactions to streamlining business processes, these articles delve into the innovative ways artificial intelligence is shaping the landscape of automated conversational agents. Whether you’re a business owner, developer, or simply intrigued by the future of interactive technology, join us on this journey to unravel the transformative power and endless possibilities of AI chatbots.
Latest news
Editions
Free AI chatbot plans changed more in the past three weeks than in the previous six months. Anthropic shipped Claude Sonnet 5.5 on 28 September and Claude Haiku 5.5 on 7 October, according to its Claude Platform release notes. OpenAI began rolling GPT-6 into ChatGPT on 7 October and widened it to Free and Go users the next day. Google published a post titled “Gemini 4 Argon is now available for everyone” on 13 October, yet our research does not show Argon as a free-tier model.
That leaves Irish readers with a practical question. If you pay nothing, which AI chatbot gives you the most useful model today? This article compares the free plans of ChatGPT, Claude, Gemini, Grok and Mistral’s Le Chat as they stand on 25 October 2026. We built it from OpenAI, Anthropic, Google, xAI and Mistral pages fetched this week, plus a recency-filtered research pass.
One warning up front. No vendor page we could reach publishes exact daily message caps for its free plan in Ireland, and we will not invent them. Where a limit is unpublished, the tables say so. Our earlier free-tier context comparison covered a different set of tools, so this piece starts from the newest models and works backwards to what a reader actually gets for €0.
The short answer: ChatGPT Free offers the broadest rollout of a brand-new model family, Mistral’s Le Chat offers the clearest European angle, and Claude, Gemini and Grok each hold their best models back for paying users. The details below show where that summary holds and where it bends.
Don't miss new tech stories on Google
Add Tech Insider once in the Google app and our stories appear in your news suggestions.
Timing matters here because three of the five vendors changed their default models inside a fortnight. A comparison written in September would already describe the wrong lineup. The dates below come from vendor pages and the research pass, and we dropped any date that two sources disagreed on.
Three dates are still ahead of us. ChatGPT Pro 200 subscribers keep their old usage allowance until 29 October, after which the allowance drops while the $200 price stays. Mistral said it would publish Large 4 weights at the end of October. And on 30 November, Google’s introductory Argon API price ends, which matters to anyone building on the free chatbot side of Google’s tools.
Model names now change faster than product pages. OpenAI alone lists GPT-6 Astra, GPT-6 Sol, GPT-6 Luna, GPT-6.1 Sol and a GPT-5.6 family that its Help Center says is not being retired. Anthropic’s API notes mention Claude Opus 5.5, Fable 5.1 and Mythos 5.1 beside Sonnet 5.5 and Haiku 5.5. A free user rarely picks among these. The vendor picks for you, and the vendor’s own pages sometimes disagree about what it picked. Treat any list of the model you get free with suspicion, including ours, and check the model label inside your own account.
The table below lists only facts we could tie to a vendor page or to two agreeing sources. “Not confirmed” means exactly that: the pages we reached did not state the figure. We chose honest gaps over plausible guesses, because a wrong message cap sends readers to the wrong tool.
Read the table as a map of what is knowable. The gaps cluster in the Free column, which tells you something by itself. Vendors advertise paid models with exact numbers and describe free access in loose language such as limited daily access. That pattern repeats across all five companies and shapes everything that follows.
ChatGPT has the widest free rollout of any new model family this month. OpenAI announced GPT-6 with Intelligent UI on 7 October. The release started with Plus, Pro, Business and Enterprise accounts and reached Free and Go on 8 October. GPT-6 Sol powers the paid plans, while GPT-6 Luna powers Free and Go. OpenAI lists Free, Go, Plus, Pro, Business and Enterprise as the current plan names.
Then the story gets messy. OpenAI’s own Help Center page on GPT-6 says the model is still rolling out to eligible Pro, Plus, Enterprise and Business users, and it also says GPT-5.6 Luna is becoming the default for Free and Go. In that same article, Instant mode for Free and Go uses GPT-5.6 Luna, and the Think option for harder questions uses GPT-5.6 Luna rather than Sol. The Help Center adds that the GPT-6 rollout does not retire older selectable models.
Both statements cannot describe the same account at the same moment, so they probably describe different stages of a staged rollout. We cannot tell which stage your account sits in. The safe reading for an Irish reader: you will get a Luna-class model on the free plan, and the generation (5.6 or 6) depends on your rollout bucket. Open the model picker and read the label before you judge the tool.
OpenAI positions Luna as the fast, low-cost member of each family. Its own comparison says GPT-6 Luna at max effort scores 66.6%, a result it calls comparable to Claude Opus 5 and Fable 5 at medium effort, while costing 93% less per task than Opus 5 and 96% less than Fable 5. Those figures come from OpenAI and measure agent-style work at maximum reasoning effort. A free user chatting about a CV or a recipe will not run that setup. The useful takeaway is the direction: the cheap model is now close to last quarter’s expensive ones, which is why free tiers feel better each month.
On 5 October OpenAI said it would first test a new advertising format during image generation for Free and Go users. That is a test, not a general rollout, and we found no Irish-specific detail. Still, it marks the first clear sign that a major chatbot’s free tier will carry commercial placements. If privacy drives your choice, watch this space.
The paid ladder above the free tier is steep. OpenAI’s Help Center lists Pro 100 at $100 a month, Pro 200 at $200 and a new Pro 500 at $500. Pro 500 includes GPT-6 Astra Ultrafast and GPT-6.1 Sol Ultrafast, which Pro 100 and Pro 200 do not. We found no verified euro price for Go or Plus, so we do not print one.
Anthropic ships the cleanest naming of the five, and it reveals the least about its free plan. The current family is Claude 5.5. Sonnet 5.5 arrived on 28 September, and Anthropic says it runs more than 30% faster than Sonnet 5 and costs up to 30% less for most work. Haiku 5.5 followed on 7 October, which Anthropic calls its fastest and most capable small model, aimed at summarisation, subagents and browser use.
Anthropic has also announced Claude Opus 5.5, which it positions for agentic coding and knowledge work. Reporting puts it at 40% cheaper to run than Opus 5 on typical workloads, though that figure rests on a single source, so treat it as indicative. The docs also describe a fast mode for Opus 5.5 that runs up to 2.5 times faster at twice the standard price. None of that is aimed at a free user.
The honest gap: our sources imply a free Claude plan exists, but they do not confirm its message limits, reset period, file limits or which model sits behind it. Anthropic does not tell free users, in the pages we reached, that they get Haiku 5.5 or Sonnet 5.5. Do not assume either.
Anthropic’s footer states that services in the EU are provided by Anthropic Ireland Limited. For an Irish reader that is a small but real detail. Your contract, and the company answering a data-protection request, sit in Dublin rather than San Francisco. It does not change which model you receive, but it does shape where you send a complaint.
Pricing for the step above free is simple. Anthropic’s pricing page lists a $20 monthly option and a $200 up-front annual option for the individual plan, which works out at about $16.67 a month and saves $40 against paying monthly. A scientific-support programme announced in October also offers verified principal investigators Team access, with Standard seats at $0 and Premium seats at $15 a month for up to a year.
Developers should also note two dates. Anthropic scheduled the Claude Sonnet 4.5 API retirement for 30 November 2026 and recommends moving to Sonnet 5.5. It also released API version 2026-10-21, with SDK changes that include a switch from the MIT licence to Apache 2.0. Neither affects a person typing into the chat window, but both affect anyone who built a small tool around the old model.
Google’s October story is Gemini 4 Argon, and the free-tier angle is blunt: the research does not show Argon as a free-tier model. Google’s announcement describes it as its latest frontier model for developers, enterprises, scientists and consumers. Reporting on the launch says it reached the Gemini app and Google Search’s AI Mode for Google AI Pro and Ultra subscribers, with Antigravity access limited to Ultra. Read “for everyone” in the headline as a statement about global reach, not about price.
The developer price is public. Argon launched at $2 per million input tokens and $10 per million output tokens, with a 95% discount on cached input. Google says that introductory rate ends on 30 November 2026, after which public access costs $4 and $20. That doubling is worth planning for if you prototype with the API on a free-tier budget. Our head-to-head on the paid flagships covers how Argon stacks up against Opus 5.5 and GPT-6.1 Sol once money enters the picture.
Google’s subscription pages list Google AI Plus, Pro and Ultra, and they show different storage bundles in different markets. One page lists AI Pro at $19.99 a month or $199.99 a year with 5 TB of storage. That is a single-source figure for a displayed market, not an Irish euro price, so check Google’s checkout before you rely on it. Google’s benefits page reportedly adds perks such as YouTube Premium Lite and Google Health Premium to AI Pro, which makes a pure chatbot comparison unfair to Google.
One more trap deserves a line. Google Cloud announced on 22 October that selected OpenAI models, including GPT-6 Luna, are entering its Model Garden in private preview for certain enterprise customers. That does not make Luna part of Gemini. If a website tells you Gemini now runs on GPT-6, it has misread an enterprise-cloud partnership.
xAI reorganised its consumer plans in the second week of October. The X Subscription folds Grok, the Cursor coding tool and X into one ladder. The free rung costs $0 and gives limited daily Grok access, a starter allowance for Cursor coding agents and standard X. xAI does not publish a daily message quota, an image allowance or a voice allowance for it, and it does not say which Grok model answers at any given moment.
The paid rungs are better documented. Plus costs $8 a month or $80 a year. Super costs $100 a month or $1,000 a year and lists 15 times Plus usage. Ultra costs $200 a month or $2,000 a year and lists 50 times Plus usage. A little arithmetic shows the logic: per unit of Plus usage, Plus costs $8, Super about $6.67 and Ultra $4. The top tier halves the unit price, so heavy users get the discount and casual users pay the premium.
xAI also states that the benefits of SuperGrok, Cursor Pro and X Premium+ previously cost $90 a month combined. A further tier called Premium reportedly sits at $30 a month with usage four times Plus. If that holds, the bundle cuts the old combined bill by about two thirds, though only for people who wanted all three products. For a reader who only wants a chatbot, the saving is theoretical. Our paid-plan price gap between ChatGPT, Claude and Grok shows the same ladder from the other side.
Grok 4.7 is the current text model in xAI’s API documentation. It takes text and image input, returns text, supports configurable reasoning effort and offers a 500,000-token context window. Its knowledge cutoff is May 2026. The API price is $2 per million input tokens, $0.50 for cached input and $6 for output on prompts under 200,000 tokens. Longer prompts cost $4, $1 and $12. Requests sent to the US regional endpoint run in the United States and cost 1.1 times the global rate, so Irish developers who care about data location should read that line.
Grok Voice 3.0 arrived in mid-October through the speech-to-speech API at $0.08 a minute, or $4.80 an hour, the same price as Voice 2.0. The older model is deprecated, and agents using the latest alias switched over automatically. xAI says Voice 3.0 reaches 74% accuracy on ServiceNow’s EVA-Bench, the highest score on that test, and that Artificial Analysis found it cheaper to run than two OpenAI voice models in a Big Bench Audio run. These are developer products rather than free chatbot features, so they matter mainly as a sign of where xAI is spending effort.
A branding note: xAI’s current pages carry SpaceXSI branding, and different pages use xAI, SpaceXAI and SpaceXSI inconsistently. Search for the product name, not the company name, when you look for official documentation.
Mistral is the only European company in this group, and it sells that fact. On 6 October it opened a public preview of Mistral Large 4, a natively multimodal model that secondary reports put at roughly 1 trillion parameters. Mistral said it would release the weights at the end of October, so a finished open-weight release cannot be confirmed as of today.
Free access to Large 4 is not confirmed. A preview often means API and Studio access first, and the pages we reached do not say that Le Chat Free users get it. Likewise, Mistral’s free-plan limits, euro prices and exact branding could not be verified. One secondary source calls Le Chat’s coding side “Vibe,” which hints at a rename, but nothing official we found confirms that Le Chat was renamed or retired.
Paid pricing could not be confirmed from the sources we checked, so we leave out Mistral’s Pro, student and Team prices. The company’s bigger news is money and partnerships. Mistral announced a €3 billion Series D at a post-money valuation above €21 billion, a Munich hub on 28 September, a Mozilla partnership for Firefox Smart Window on 16 September and an NVIDIA collaboration on engineering workflows on 21 October.
For a casual Irish user, Le Chat’s draw is positioning rather than raw power. If you care that a model comes from a European company with sovereign-AI ambitions, it fits. Our comparison of Le Chat against Meta AI and Copilot looks at that trade-off in more detail.
Every price below is in US dollars because that is how the vendor pages we reached state them. Irish buyers should expect euro checkout prices that include VAT, and we could not verify any of those. Rows marked “not confirmed” stay empty on purpose.
Three patterns stand out. First, $200 a month is the common top of the individual ladder: ChatGPT Pro 200 and Grok Ultra both land there, and ChatGPT adds a $500 rung above it. Second, annual billing saves exactly two months at xAI ($80 against $96, $1,000 against $1,200, $2,000 against $2,400) and about 17% at Anthropic. Third, the $0 rows are the only ones where the sources go quiet on limits, which is why a free plan is a trial rather than a commitment.
If you build rather than chat, the cheap end moves fast. Grok 4.7 runs at $2 and $6 per million tokens, Argon at $2 and $10 until 30 November, and Sonnet 5 at $2 and $10 as its standard price. Our API cost comparison goes deeper on those developer rates.
No independent lab has published a head-to-head test of the free tiers as we write. What exists are vendor-reported results on the paid or API versions, plus a few third-party measurements cited by vendors. The table collects them, with the source named in each row so you can weigh the bias.
The pattern in these rows is consistent: every number measures either a paid model, an API product or a cost ratio. Cost ratios are useful for a free-tier forecast, because cheaper models are what vendors can afford to give away. Quality scores are less useful, because the free model is almost never the one that earned the headline result. Until an independent lab tests Luna, Haiku or Medium 3.5 on everyday tasks, assume the free experience is a notch below the numbers in this table and test it with your own prompts.
A free plan is a shop window. The models behind it earn their money in business deployments, and those deployments show where each vendor believes its product is strongest. We found six dated examples in this month’s sources.
Taken together, the examples show a split. OpenAI and Anthropic win large-company contracts and developer tooling. xAI leans on voice and its sister companies. Mistral courts governments and European industry. Google sells breadth through bundled storage and cloud credits. None of these deployments says anything about the free chatbot you would open on your phone, and that is the point: the free tier is the part of the product with the least public evidence behind it.
These picks are our analysis, built on the verified facts above. We name a first choice and a reason, and we flag where the choice rests on a gap in the data.
Start with ChatGPT Free. It has the broadest, newest free rollout, and a Luna-class model handles explanations, summaries and practice questions well. Age rules matter more than model quality here, and they differ by vendor, so read our Leaving Cert age-limit breakdown before a student signs up. Use Claude as a second opinion for essay structure.
Pick whichever chatbot you already have an account with, then test two. Our Irish CV comparison covers how the big three handle that task, and a specific prompt usually matters more than the logo on the app. Paste the job advert, your experience and a request to cut it to one page.
Do not trust any free tier blindly. Our Gaeilge benchmark tests the big models on Irish and is the place to start. None of the newest free models has an Irish-language score in this month’s sources, so treat every output as a draft that a fluent speaker should check.
Look at Mistral’s Le Chat first, then Claude. Mistral is a European company, and Anthropic provides EU services through Anthropic Ireland Limited. Avoid assuming that either is private by default. Read the data settings in the app and turn off model-training options where the app offers them. ChatGPT’s ad test on the free tier makes it the least attractive option on this criterion for now.
Grok’s free tier includes a starter allowance for Cursor coding agents, which no rival free plan matches in our sources. Pair it with a Claude chat for code review. If you move to the API, compare Grok 4.7 at $2 and $6, Argon at $2 and $10 and Sonnet 5 at $2 and $10 per million tokens, and watch the 30 November changes.
If your email, calendar, photos and phone already live in Google’s world, Gemini’s free tier makes sense for convenience, even though Argon needs a subscription. Do not pay for AI Pro just for the chatbot. The 5 TB of storage and the bundled services are the real purchase, and that only pays off if you need the storage.
Free is never free of trade-offs. OpenAI’s October ad test is the clearest sign that the price of a free chatbot may soon be attention rather than money. Anthropic published a 2026 Usage Policy update on 8 October. xAI’s pages disclose a $0.05 usage-guideline violation fee per request for API calls caught before generation, a rule that targets developers rather than chat users. For EU law, the biggest open question for all five vendors is how they handle the AI Act, which we cover in our EU AI Act fine-risk comparison.
Usage data adds context, though most public numbers are older than the models they describe. We pulled the freshest verified statistics we could find, and each carries its year.
“In 2025, 64% of U.S. teens said they had ever used an AI chatbot, and about 30% said they used chatbots daily.”
“In 2025, 59% of U.S. teens said they used ChatGPT, while 23% said they used Gemini.”
“In March 2025, Grok averaged 16.5 million daily worldwide web visits.”
“In March 2025, Gemini averaged 10.9 million daily worldwide web visits, up 7.4% month over month.”
Two cautions apply. The Pew figures cover American teenagers, not Irish adults. The Similarweb numbers come from March 2025, before any of the models in this article existed. Even so, they show the shape of the market: among American teens ChatGPT use ran at 59% against 23% for Gemini, and Grok’s web traffic in March 2025 was larger than Gemini’s. Popularity does not equal quality. It does mean the biggest free tiers attract the most real-world testing and the most third-party tutorials.
Switching costs on free plans are low, but they are not zero. You lose chat history, custom instructions and any saved projects. A methodical move takes about an hour and avoids the common mistake of judging a new tool on one lucky answer. Our AI and machine learning coverage tracks the model changes that make switching worthwhile.
If a small script or app calls a model API, the October changes may break it quietly. Anthropic retires Claude Sonnet 4.5 on its API on 30 November and recommends Sonnet 5.5. xAI deprecated grok-voice-2.0 and moved the grok-voice-latest alias to Voice 3.0 at the same price. OpenAI says GPT-6 does not retire older selectable models in ChatGPT, but API names such as gpt-6-sol and gpt-6-luna are separate identifiers. A tiny harness helps you compare models on your own prompts. Only fill in model names that you have confirmed in each vendor’s documentation.
Write your own call_model function with each vendor’s official SDK. Keep the prompts and the scoring sheet in version control so you can rerun the test when the next model lands.
Our pick for most Irish readers is ChatGPT Free, with one caveat. It is the only free plan where a vendor confirms that a brand-new model family reached free accounts within a day of reaching paid ones (7 October for paid, 8 October for Free and Go). The caveat is the model-label confusion between GPT-6 Luna and GPT-5.6 Luna, plus the ad test. Check your label, and you will know what you are getting.
Our second pick is Claude Free for writing and summarising, because Anthropic’s naming is the clearest and its fresh Haiku 5.5 targets exactly that kind of work. We cannot confirm which model Claude’s free plan uses, so test it on your own text before you commit. Mistral’s Le Chat takes third place for privacy-minded readers, Grok Free for coders who want the Cursor allowance, and Gemini Free for people who live inside Google’s apps.
The data points that decided it: ChatGPT shipped GPT-6 to Free users in one day, Anthropic put out two new models in nine days (28 September and 7 October), Google kept Argon behind a subscription, and xAI and Mistral left their free-tier limits unpublished. Paying only makes sense at specific thresholds. Anthropic’s annual option saves $40 against monthly billing, xAI’s Ultra tier halves the per-unit cost against Plus, and ChatGPT’s top tier costs five times Pro 100. For light use, stay free and rotate between tools.
What would change our ranking? A published daily cap for Claude Free, a confirmed model name for Gemini Free, or an independent lab test of Luna, Haiku 5.5 and Medium 3.5 on everyday tasks. We will update this comparison when any of them lands.
Which free AI chatbot is best in Ireland right now?
ChatGPT Free is the safest all-round choice as of 25 October 2026, because OpenAI extended its GPT-6 rollout to Free and Go users on 8 October. Claude, Gemini, Grok and Le Chat are close behind for specific jobs. None of the vendors publishes enough free-tier detail to name a clear winner on quality.
Does the free version of ChatGPT use GPT-6?
OpenAI’s rollout notes say GPT-6 Luna powers Free and Go. Its Help Center page still names GPT-5.6 Luna as the default for those plans during the rollout. Both can be true at different moments, so check the model label in your account.
Is Gemini 4 Argon free?
The research we ran does not show Argon as a free-tier model. Google’s announcement points to developers, enterprises and subscribers. Its API price is $2 per million input tokens and $10 per million output tokens until 30 November 2026, then $4 and $20.
How much does Grok cost?
The X Subscription starts free with limited daily access. Plus costs $8 a month or $80 a year, Super costs $100 a month or $1,000 a year, and Ultra costs $200 a month or $2,000 a year. Prices are in US dollars, and Irish checkout prices were not confirmed.
Which company handles my data in Ireland if I use Claude?
Anthropic’s pages state that services in the EU are provided by Anthropic Ireland Limited. That is the contracting entity for EU users, so data-protection requests should be directed to it.
Is Mistral’s Le Chat a good free alternative?
It suits readers who want a European provider. Mistral Large 4 has been in public preview since 6 October. Free-plan limits, the current Le Chat default model and euro prices were not confirmed in the sources we checked.
Will free AI chatbots show ads?
OpenAI said on 5 October that it would first test a new advertising format during image generation for Free and Go users. We found no similar announcement from Anthropic, Google, xAI or Mistral in this month’s sources.
Should I pay for an AI chatbot?
Only if you hit limits regularly. Annual billing saves 17% at Anthropic and two months at xAI. For occasional use, rotating between two or three free tools costs nothing and covers most everyday tasks.
Niamh Kelly is the iGaming Editor at Tech Insider, where she previously worked as a freelance fashion journalist for The Irish News for three years and honed her media skills during her time at the BBC. At Tech Insider, she leads Ireland’s coverage with hands-on experience testing consumer and business technology, delivering in-depth analysis on AI, cybersecurity, cloud computing, and hardware trends shaping the future. Kelly was featured in RSVP online as part of their “Women of Style” series and has interviewed notable figures such as Katie Price and Louise Redknapp for major beauty product launches.
Tech Insider delivers in-depth coverage of the technologies shaping the future: AI, cybersecurity, cloud computing, hardware, and the trends that matter.
