The State of AI in 2026: Chatbots, Image Generators, and Video Tools, Compared
Free tiers, paid tiers, and why Sora just dropped out of the race entirely.
Three chatbots, four image generators, four video generators — and one very notable dropout.
Key takeaways
- Among the major chatbots, Claude has stayed positioned around professional and coding use, GPT has built out the broadest bundled ecosystem (chat, image, video, and agents together), and Gemini has leaned on subscription bundling with perks like YouTube access to differentiate.
- Claude is the notable exception in not building image or video generation into its product at all — it reads and analyzes images, but doesn't generate them, which is a deliberate positioning choice rather than a missing feature.
- AI image generation has consolidated around a handful of leaders — GPT Image, Google's Nano Banana/Imagen family, Midjourney, and Grok Imagine — each with meaningfully different free-tier generosity and pricing structures.
- The video generation category saw the biggest shake-up: OpenAI discontinued Sora entirely rather than releasing a successor, folding its video ambitions into a different internal project, while Google's Veo and Runway's Gen-4 have continued iterating.
- General-purpose image and video generators solve a different problem than template-based consumer apps — one is built for open-ended creation from a prompt, the other for taking your own photo and applying a specific preset result.
The AI landscape has moved fast enough in 2026 that a comparison from even six months ago is already out of date in places — free tiers have shrunk, at least one major product line has been discontinued outright, and pricing has shifted across almost every platform in this space. Here's where things stand across the three categories that matter most for anyone actually using these tools day to day: chatbots, image generators, and video generators.
The chatbot "three-way race": Claude, GPT, and Gemini
The positioning differences between the three biggest general-purpose AI chatbots have become more distinct over the past year, not less:
- Claude has stayed focused on professional and coding-oriented use, without expanding into image or video generation at all — more on that below.
- GPT has built out the broadest bundled ecosystem of the three, tying chat together with image generation, video generation, and agent-style features under one subscription and one account.
- Gemini has leaned into subscription bundling as its differentiator, pairing its AI features with other Google services — including perks like YouTube access — as part of the value pitch rather than competing purely on model capability.
On free-tier limits specifically, GPT's free tier has reportedly become fairly restrictive — around 10 messages within a 5-hour window before falling back to a lighter model — and Gemini's free tier has reportedly been pushed away from the top-tier "Pro" model entirely as of a 2026 update, leaving free users on lighter models across both platforms.
Claude's deliberate exception
Worth calling out on its own: Claude has not added image or video generation, even as both of its biggest competitors have made generation a core, bundled feature. It can read and analyze an image you share with it, but it doesn't create new images or video. Of the three major chatbot platforms, it's the only one that hasn't tried to compete on generation at all — a positioning choice that stands out more the further the other two lean into it.
AI image generation, compared
| Tool | Free tier | Paid highlights |
|---|---|---|
| GPT Image (OpenAI) | Tied to a ChatGPT account; Plus users reportedly get roughly 50 images every 3 hours on the current model | API pricing reportedly ranges roughly $0.005–$0.25 per image depending on resolution and quality tier |
| Nano Banana / Imagen (Google) | Up to 500 images per day (1024×1024) through Google's AI Studio, on the current reporting | The higher-end "Pro" tier has reportedly been pulled from free access entirely; Imagen's faster tier starts around $0.02 per image via API |
| Midjourney | No permanent free tier as of 2026 — fully subscription-based | Reported tiers roughly $10 / $30 / $60 / $120 per month (with an annual discount), with unlimited "relax mode" generation starting at the mid tier and up |
| Grok Imagine (xAI) | Tied to an X account | Bundled into X Premium (from roughly $8/month) or a higher "Heavy" tier around $300/month; API reportedly around $0.04–$0.08 per image |
AI video generation, compared
| Tool | Free tier status | Paid access |
|---|---|---|
| Sora (OpenAI) | Discontinued — free access ended first, then the consumer app, then the API | N/A — the product has been shut down rather than continued; see below |
| Veo (Google) | Not included in Gemini's free tier | Unlocked starting around $20/month, with a credit-based system for generating clips; API pricing reportedly ranges roughly $0.03–$0.75 per second depending on quality tier |
| Runway Gen-4 | A one-time allotment of starter credits on signup, not an ongoing free tier | Reported tiers roughly $12–15/month and $76/month depending on credit volume and rollover |
| Grok Imagine Video (xAI) | Tied to an X account | API-based pricing reportedly around $0.08/second for output |
Why Sora dropping out is the real story here
Of everything in this comparison, the most notable development isn't a new model — it's a product being discontinued entirely. According to current reporting, OpenAI didn't release a "Sora 3." Free access was cut off first, then the consumer app and web version were taken offline, and the API followed after that. Rather than continuing the Sora line, OpenAI has reportedly redirected its video generation ambitions into a different internal effort, and previous Sora subscribers were shifted onto its general ChatGPT Pro plan instead.
That's a genuinely unusual move in a category where every other major player — Google's Veo, Runway's Gen-4, xAI's Grok Imagine Video — has kept iterating and releasing new versions rather than stepping back. Whatever the internal reasoning, it leaves the current "top tier" of general-purpose AI video generation as a three-way conversation rather than four, with the company that arguably popularized consumer AI video no longer in it.
The biggest news in AI video this year wasn't a new model. It was the absence of one, from the company that started the category.
A quick look at each vendor's own model lineup
OpenAI
Current reporting has OpenAI running GPT-5.3 on the free tier and GPT-5.4 Pro for paid subscribers, with free users limited to roughly 10 messages within a 5-hour window before dropping to a lighter "mini" model. Its image generation line has moved to GPT Image 2.5 as of early September 2026, following GPT Image 2.0, 1.5, and 1 (and DALL·E 3 before that), reportedly supporting 4K output, up to 16 reference images, and new higher-quality output tiers, split into a faster "Flare" variant and a more detailed "Sunburst" variant via API.
Google's chatbot line is currently on Gemini 3.1 Pro, though free-tier access to the "Pro" model tier was reportedly removed as of an April 2026 update. On the image side, Google runs two parallel product lines rather than one: the Nano Banana family (built on its Flash Image models, positioned around conversational, back-and-forth photo editing — "make the sky a bit more orange" style requests) and the Imagen family (positioned around one-shot, high-quality text-to-image generation), with six models reportedly on offer across both lines as of this writing. Its video line is currently on Veo 3.1, reportedly supporting 4K output, clip extension, first/last-frame generation, and native synchronized audio, split into Standard, Fast, and Lite tiers.
Anthropic (Claude)
Anthropic hasn't built a dedicated image or video generation model at all — Claude's positioning has stayed centered on text and code, with image capability limited to understanding and analyzing images you provide rather than creating new ones. Among the three major chatbot platforms, it's the one deliberately sitting out the image/video generation race entirely.
Midjourney
Midjourney's current default version is reportedly V8.2, focused on aesthetic quality and personalization, following V8.1 (a significant speed improvement) and V7 before that. It also shipped its first motion model, V8 Video, which can extend a static image into a short clip (reportedly up to 21 seconds), currently limited to its higher-tier Pro and Mega subscribers.
If you're not trying to build the next viral AI video — a quick note
Everything above is aimed at general-purpose creation: describing something in a prompt and getting an entirely new image or video back. That's a genuinely different problem from what a lot of people actually want day to day, which is closer to "take this one photo of me and turn it into something specific" — a new outfit, a different art style, a short video from a template. That narrower, more predictable task is exactly the gap that template-based apps like Movcl are built for, and it's worth understanding the difference before assuming you need a $60-a-month general-purpose subscription for something a purpose-built app already handles in about a minute. We covered how that underlying technology actually works, separately from any of the general-purpose tools above, in our piece on how AI face swap works.
Frequently asked questions
Is Sora still available in 2026?
As of the most recent reporting, OpenAI discontinued Sora rather than releasing a successor. Free access ended first, the consumer app and web version were later taken offline, and the API was shut down after that, with OpenAI folding video capability into a different internal effort instead of a "Sora 3."
Does Claude generate images or videos?
No. Unlike GPT and Gemini, Claude has stayed positioned as a text-and-code assistant. It can read and analyze images, but image and video generation aren't part of what it does — a deliberate difference from its two biggest competitors, which have both bundled generation features into their subscriptions.
Which AI image generator has the best free tier?
Based on current reporting, Google's Nano Banana/Imagen line offers one of the more generous free allowances, available through Google's AI Studio. Free tiers change frequently across this category, though, so it's worth checking current terms directly before assuming a specific number still holds.
Do I need a general-purpose AI image generator to make fun photo edits of myself?
Not necessarily. General-purpose tools like Midjourney or GPT Image are built for open-ended creation from a text prompt, while template-based apps are built specifically around taking one of your own photos and applying a preset style, outfit, or scene — a narrower, often simpler task suited to a different kind of tool.