Rumors point to a massive ChatGPT update soon; this is what we know

Rumors abound that sometime in the next week or two, we’ll see major updates to two of the leading AI chatbots: ChatGPT and Claude.

One of the biggest rumors centers on the underlying models that power the intelligence on both platforms. We’ve been expecting bigger and smaller versions of Claude 3.5 since the release of Sonnet earlier this year, and an update to OpenAI’s GPT-4 family has been long overdue.

Some features have been left out or mocked, especially by OpenAI. This includes SearchGPT integration and the full version of its o1 reasoning model. However, Anthropic has just released a new iPad version of the Claude app and has updated the mobile apps, perhaps in preparation for this rumored new model.

There are no specific dates for when this will happen, but throughout the year both OpenAI and Anthropic have mentioned updates in the fall. When I look outside, the leaves on the trees are starting to turn orange and the pumpkins are in the stores, which means they’re falling for me. And that suggests updates could be imminent.

What can we expect in the AI space?

ChatGPT Canvas

(Image credit: OpenAI)

Update rumors are nothing new to me as a tech journalist. I’ve written many stories speculating about new features coming to the next version of iOS, a toggle button on the side of a pair of headphones, or a camera upgrade on the latest smartphone.

While we are now starting to see feature add-ons for chatbots with generative artificial intelligence (ChatGPT Canvas or Claude Artifacts are good examples), these rumors often surround model changes. A model upgrade is the equivalent of upgrading to a new operating system, from iOS 17 to 18 or Windows 10 to 11, rather than a simple feature upgrade.

We’re expecting something new this year, and I’d still put money on it being the next big update to the GPT family.

OpenAI introduced its latest update to GPT-4 in the spring with GPT-4.0, or its native version of GPT-4’s multimodal Omni model. He then released his reasoning model 01, which many speculators believe is still based on the GPT-4 family, at least the preview of the mini versions we’ve seen. What we haven’t had is a GPT-4.5, either in Omni, 01 or mini flavor, or even the rumored GPT-5.

There is a lot of speculation that we may never see GPT-4.5 or GPT-5 because OpenAI is moving away from this type of labeling. Sam Altman himself has said that there is a distinction between the category of o1 models and GPT-type models. But we’re expecting something new this year, and I’d still put money on it being the next big update to the GPT family.

When it comes to Anthropic, they are pretty regular with their updates. In June, we saw the release of the Claude 3.5 Sonnet – this was the mid-level model. The Claude family comes in three flavors: Haiku, which is the equivalent of GPT Mini or Gemini Flash; Sonnet, which is the mid-level model; and Opus, which is the very large, high-spec model. All we’ve seen of the 3.5 family of Claude models so far is Sonnet, and everyone is speculating that Opus is just around the corner.

Why is this so big?

Screenshot of ChatGPT

(Image credit: ChatGPT screenshot)

With each new model comes a new level of functionality and capability. Some of these features are enclosed by railings. For example, we know for sure that GPT-4.0 is capable of creating images, vector graphics, and the voice version is capable of singing, and all of these features have been disabled by OpenAI.

When a new model comes out, it will improve reasoning, perform better on all standard metrics and benchmarks, allowing for improved coding, better writing, and more nuanced conversations with AI. All this will improve the user experience.

As each new flagship model is released, this helps refine the smaller versions, making them cheaper, faster and more efficient, allowing them to be used more widely. This helps us get to the point that Sam Altman describes as “intelligence too cheap to measure.”

If we don’t get a completely new model, I suspect we’ll see the full deployment of SearchGPT to ChatGPT, wider access to Advanced Voice, and for Anthropic, the ability to access the live Internet and code running on Claude .

Related Posts

The MMORPG in Germany is used on Steam, it will take 7 years and release a feature, to do this

Florian Ast: Review zum nouen Album «Ast A La Vista»

Prophet Magaya accused of bribery to avoid ZIFA O level requirement – Nehanda Radio