ChatGPT, Claude & Gemini: Every Major Update as of October 2026

If you last checked in on the big three AI assistants over the summer, a lot of what you know is already out of date. In the space of about six weeks, OpenAI rolled a new GPT-6 generation into ChatGPT, Anthropic replaced its main Opus, Sonnet, and Haiku models with a 5.5 generation and made its document and design tools free, and Google unveiled its next frontier model while quietly shrinking what free users can access.

This is a practical rundown of what changed at each company as of October 12, 2026, what it costs, and what it means depending on whether you use these tools for free, for work, or as a developer.

ChatGPT: GPT-6 Arrives, and Prices Fall

OpenAI’s GPT-6 rollout happened in stages. The flagship, GPT-6 Astra, came first in early September. On September 22, OpenAI followed with two cheaper models built on the same architecture: GPT-6 Sol, aimed at coding and agent-style work, and GPT-6 Luna, a lightweight model for high-volume tasks like summarizing and extracting information. A week later, on September 29, OpenAI introduced GPT-6.1 Sol as an upgrade to Sol.

The most notable thing about Sol and Luna isn’t a new benchmark record — it’s price. Sol costs $2 per million input tokens and $10 per million output tokens, half of what GPT-5.6 Sol cost. Luna dropped to $0.10 and $0.50, with both models offering a roughly million-token context window and up to 128,000 tokens of output. Sol also exposes a range of reasoning-effort settings, from none up to max, so developers can trade depth for speed and cost on each request. Some independent reviewers have pointed out that the gains here are mostly about cost rather than leaps in top-end scores, which fits OpenAI’s decision to keep Astra as the premium option.

What changed inside ChatGPT itself is what most people will notice. On October 7, GPT-6 began powering regular ChatGPT chat: Sol for Plus, Pro, Business, and Enterprise users, with Luna reaching Free and Go users starting October 8. Alongside that came what OpenAI calls an “Intelligent UI,” where answers can include interactive elements such as charts, buttons, and calculators instead of plain text alone. OpenAI also announced “dots” on September 29, an always-available assistant that keeps working even when you aren’t actively chatting, currently limited to Pro and Business Premium users.

Who should care: free and Go users get a meaningful quality jump for the first time in months, and developers get a much cheaper route to capable coding models.

Claude: The 5.5 Generation and Free Creation Tools

Anthropic had two big waves of news. The first was the Claude 5.5 model generation.

Claude Opus 5.5 launched on September 22. It costs $4 per million input tokens and $20 per million output tokens, about 20% below Opus 5’s list price, and Anthropic says it costs roughly 40% less than Opus 5 on typical workloads while generating output more than 30% faster. It supports a million-token context and 128,000 tokens of output, and defaults to a medium effort setting. Anthropic’s own benchmark table shows it matching or beating the pricier Fable 5.1 on most measures at a bit over a third of the price. Anthropic also says it improved the clarity of Opus’s writing, addressing a common complaint about the previous version.

Claude Sonnet 5.5 followed at $2 and $10 per million tokens, positioned as the best mix of speed and intelligence, and Anthropic cut its cache-read price to $0.10 per million tokens on October 7. Claude Haiku 5.5 is the new low-cost, fastest tier, starting at $0.10 and $0.50 per million tokens. At the top, Claude Fable 5.1 (alongside the restricted Mythos 5.1) remains the premium option at $10 and $50.

One detail developers should not skip: Opus 5.5 introduced breaking changes versus Opus 5. Extended thinking can no longer be switched off, forcing a specific tool call now returns an error, thinking blocks are tied to the specific model and conversation, and an older computer-use tool version is no longer accepted on the Claude API and Google Cloud. Anyone with production code on Opus 5 should read the migration notes in Anthropic’s Opus 5.5 documentation before switching.

The second wave arrived on October 8, and it matters more to everyday users. Anthropic took Claude Docs, Slides, and Design out of beta and made them available on every plan, including Free. Anthropic says people had already created more than 45 million documents, decks, and designs in Claude. The same day it launched Claude Dashboards (beta, paid plans) for tracking live numbers and Claude Motion (beta, Team and Enterprise) for turning ideas into short animated explainers. Anthropic also merged Claude Cowork and standard chat into a single Claude experience, and the standalone Claude Design site is scheduled to close on December 14 as it folds into the main app. One caveat: all of this draws from the same usage pool as everything else you do in Claude, so a large deck can eat a free session quickly.

Who should care: free users can now draft documents, presentations, and designs without paying, and developers get a cheaper top-tier option in Opus 5.5.

Gemini: A Frontier Model Behind a Gate, and a Tighter Free Tier

Google’s news cuts in two directions. On September 30, it announced Gemini 4 Argon, its new frontier model, built for long-running software engineering, enterprise knowledge work, and cyber defense. Google says Argon can autonomously find, validate, and patch serious software vulnerabilities. But as of October 11, access is limited to vetted cyber defenders in Google’s Fairwind Program and internal teams, with no public release date. Google has said Ultra subscribers will get access before any other consumer plan, and the planned API pricing is an introductory $2 per million input tokens and $10 per million output, rising to $4 and $20 later.

The second change affects far more people. Starting October 9, the free Gemini app is limited to Flash-Lite, the smallest model in the family. Free users lose the full Flash model and Gemini 3.1 Pro, and the model picker effectively collapses to a single option with a 32,000-token context window. Here’s how the tiers now line up:

• Free: Flash-Lite only.
• Google AI Plus ($4.99/month): Flash-Lite and Flash, but no longer Gemini 3.1 Pro.
• Google AI Pro ($19.99/month) and AI Ultra (from $99.99/month): all three models, plus Deep Think.

The practical effect is that the cheapest way to reach Gemini’s Pro model is now the $19.99 plan. Students who relied on the free tier for lecture notes and PDFs, and developers who used it for code help, will feel this most. There have also been reports of Google Cloud’s new “Gemini Agent” letting users choose Anthropic’s Claude models alongside Google’s own, a sign that even rivals are becoming interchangeable parts of the same toolkit.

Who should care: anyone on Gemini’s free plan, and anyone budgeting for the AI Plus tier.

Side-by-Side: Pricing at a Glance

Prices change often, so confirm current rates on each provider’s own page, such as Anthropic’s pricing documentation and OpenAI’s release notes, before budgeting.

The Bigger Pattern: Free Tiers Are Diverging

Look at the three companies together and a pattern emerges that has nothing to do with benchmarks. OpenAI is pushing its newest generation down to free users through Luna. Anthropic is widening what free users can create, with documents, slides, and designs. Google is doing the opposite, pulling its free tier back to a single small model while saving its best work for paying subscribers and vetted partners.

That divergence matters because the free tier is where most people form their opinion of an AI assistant. For the first time in a while, which assistant feels best for free is a genuinely different answer than it was in the summer.

The other clear trend is price compression. GPT-6 Sol, Claude Sonnet 5.5, and Gemini 4 Argon’s introductory price all sit at the same $2 and $10 point, which suggests the market is settling on a common price for capable, everyday models, with premium tiers like Astra and Fable 5.1 reserved for the hardest work.

Which One Fits You Right Now?

• Free user who wants the best everyday quality: ChatGPT’s GPT-6 Luna or Claude’s free plan are the stronger bets now that Gemini’s free tier is limited to Flash-Lite.
• Someone who creates documents, decks, or designs: Claude, since Docs, Slides, and Design are now free on every plan.
• Developer watching costs: GPT-6 Sol and Claude Sonnet 5.5 share a price, Opus 5.5 offers a step up, and Haiku 5.5 and Luna cover high-volume work cheaply.
• Heavy Google Workspace user: Gemini remains the natural fit, though the useful models now sit behind paid plans.
• Anyone chasing the absolute frontier: Claude Fable 5.1 and GPT-6 Astra are the options you can use today, while Gemini 4 Argon stays gated.

For earlier context on how these models got here, see our coverage of Claude Opus 5’s launch and our breakdown of the newest AI models, and for work-focused guidance, our ChatGPT vs. Gemini guide for freelancers.

Frequently Asked Questions

Is GPT-6 available to free ChatGPT users? Yes, in the form of GPT-6 Luna, which began reaching Free and Go users on October 8. Paid users get the more capable GPT-6 Sol in regular chat.

What changed for free Gemini users on October 9? Free accounts are limited to Flash-Lite. The full Flash model and Gemini 3.1 Pro are no longer available without a paid subscription.

Are Claude Docs, Slides, and Design really free now? Yes. Since October 8 they are out of beta and available on every Claude plan, including Free, though they count against your plan’s usage limits, and the free allowance is the smallest.

Can I use Gemini 4 Argon today? Not unless you’re in Google’s Fairwind Program. It was announced on September 30 but remains restricted, and Google hasn’t published a public release date.

Which model is cheapest for developers? GPT-6 Luna and Claude Haiku 5.5 both start around $0.10 per million input tokens, making them the most economical options for high-volume work.

The Bottom Line

October 2026 shows the three big assistants moving in different directions on the same question: who gets the best model, and at what price. OpenAI is spreading GPT-6 widely, Anthropic is making its creative tools free while refining its model lineup, and Google is concentrating its strongest work behind paywalls and partner programs. For most people, the best move is to retest your favorite tool on the free tier you actually use, since the answer may have changed since the summer — and to check back soon, because at this pace, something here will change again within weeks.