GPT-5.6 The first batch of actual measurements are here! Precise sniper on Mythos

Just now, Anthropic released its “big killer” – a product that has been in development for two months.Claude Fable 5andMythos 5It's like dropping a bomb.

Apply direct pressure on OpenAI now.

At the same time, GPT-5.6 was also leaked.

OpenAI has been testing its internal code since last week.keplerandkindleTwo new checkpoints. Kindle-alpha has been selected as a candidate for release.

The internal test version of GPT-5.6 has started to gain widespread attention among overseas developers and in the leak community, with extensive testing being conducted. The code, candidate versions, and related technical details have all been made available for review.

Whether it's competing for an IPO or having their flagship models collide, the two companies say, "I'll also submit my application; if you release a new model, I'll release one as well."

Purity means being in a state of intense conflict or struggle.

But the question is, can GPT-5.6 really defeat Mythos?

The text "GPT-5.6" has been translated into English as follows: "GPT-5.6" has come to light.

So far, OpenAI has confronted GPT-5.6 without any official announcements from OpenAI, and it has not been officially released yet.

However, many overseas internet users have not yet made it public that the testing of the "internal checkpoint" probes has been completed.

A checkpoint is a snapshot of the model's parameters that is saved at a certain point in time during the training process.

There will be multiple versions within OpenAI, which will be compared side by side. Then, one version that is deemed “good enough to be released” will be selected and called the Release Candidate (RC).

Since last week, OpenAI has been testing two new checkpoints internally, codenamed Kindle and Kepler. Among themkindle-alphaThe candidate version has been selected for release.

Judging from the information that has been leaked, the upgrade that was mentioned the most repeatedly in this instance is the one referred to by GPT-5.6.Generate front-end/UI。

According to netizen Pankaj Kumar, the Kindle has greatly enhanced the front-end generation capabilities of Alpha.It is possible to produce a more powerful user interface output directly, without the need for complex prompts or additional skills.。

In addition, its visual capabilities are also very good; it performs well in tasks related to image understanding and image retrieval. There have been significant improvements in areas such as reasoning, encoding, and UI generation.

This is the result measured by netizen Chris using the medium setting for the Kindle:

This is the result measured by another netizen on the non-reasoning version of Joule before:

The former one is much more refined.

But netizen Leo used the same prompt, Kepler, and Kindle to make measurements at the xhigh setting.

Compared to Kepler, it has been found that the Kindle is still regressing in terms of performance or features.

Well... it's really hard to evaluate this effect.

He even believes that OpenAI is very likely to continue to refine and improve its capabilities.It is not ruled out that Kindle may eventually be dropped as a candidate option.。

The latest news is that the Kindle has been removed from the Arena, and a new model has been released.Levi。

Some netizens have speculated that "Levi" could also be the code name for an internal version of "GPT-5.6", and have compared its front-end capabilities with those of "GPT-5.5":

It can be seen that Levi's front-end is also quite impressive; it has a fresh and minimalist style with a sense of sophistication, and the details are handled very well.

However, some netizens have discovered that Levi might actually come from Meta, rather than GPT-5.6.

So, can GPT-5.6 actually defeat Mythos?

Netizen mark_k claims that GPT-5.6 "multiple agents have surpassed the Mythoss coding benchmarks."

However, what is more convincing at present is the actual test result provided by the netizen Leo, which was shown earlier. He believes that the situation with GPT-5.6 is not optimistic:

Compared to Kepler, the Kindle is a step backward. In its current form,It can be easily defeated by Mythos.。

In June, the three prominent families staged a "Fast and Furious" style event.

June has arrived, and summer has come, and the big model community has also become more active.

The release times of the three leading overseas AI models all coincided: Fable 5, Gemini 3.5 Pro, and GPT-5.6, creating a real “race against time” in the AI industry.

And they are playing with the same set of capabilities: reasoning, agents, coding, and front-end generation.

Interestingly, although all three companies focused on that aspect in June,So far, only Company A has actually submitted their reports.。

Gemini 3.5 Pro was unveiled at Google's I/O conference on May 19, focusing on processing contexts with up to 2 million tokens and utilizing Deep Think reasoning capabilities.

But it hasn't been officially launched yet; the official release is scheduled for June.

GPT-5.6The news was announced that it will be released later this month.

This also adds another layer of tension to the situation with OpenAI: the competitor has already released their scores, and they are probably still debating within the company which RC version to submit.

But besides benchmarking,PricingIt is also an important factor.

Fable5.Mythos5 Unified PricingToken10Dollar,Token50Dollars.

It is approximately twice the size of the existing Opus.

If GPT-5.6 has similar capabilities to Mythos, or even a slight decrease in performance, but at a much lower price, it could still potentially increase its practical usage rate in a city~

Currently, OpenAI has not made any official announcements. The actual duel will have to wait until the official release of GPT-5.6 and the moment when Fable receives direct performance evaluations.

It's very likely that we'll get the answer this month. Stay tuned!

Reference Links:

[1]https://x.com/mark_k/status/2063922897341567488?s=20

[2]https://x.com/AiBattle_/status/2064078302394917157?s=20

[3]https://x.com/pankajkumar_dev/status/2063272015214354908?s=20

[4]https://x.com/synthwavedd/status/2063245096951160865?s=20

[5]https://x.com/ChrissGPT/status/2063135842906808579?s=20

[6]https://x.com/koltregaskes/status/2062806155139912164?s=20

The author comes from "Quantum Bit" and goes by the name "Ting Yu".