OpenAI's new model is unveiled! GPT-6.1 Terra is flying the flag?

  It's the first time in the history of the world.

Just recently, the next generation of OpenAI models came to light.

GPT-6.1 Terra, which was released by top-notch developers thanks to updates to Microsoft's and GitHub's internal code libraries.

Now, the outbreaks of the GPT-6.1 Terra have been screened on X.

This is what Tibo predicted this week.

Tibo announced yesterday: "This week is a good one.

Interestingly, various clues suggest that this mid-range model is the only "Old 2 and 3" in the family that not only jumps in price, but even outperforms its older brother GPT-6.1 Sol directly in core capabilities.

Some developers are happy and think it's a big benefit, but others are worried about OpenAI: It's about time, are you still squeezing toothpaste?

The explosion came from a code warehouse.

Just when everyone thought OpenAI's "28 Days Codex Improvement Plan" was about to close, netizens found a dead end in an open code repository.

Clue one: The mysterious submission of the Microsoft Entity Framework Core warehouse

On October 16, Microsoft's EF Core team merged a submission called PR #39217.

The eye-catching V @chetaslua instantly discovered that in this configuration, there were a total of 18 judge_models in the evaluation configuration, which was replaced by the original gpt-5.6-terra.gpt-6.1-terra！

The mission execution model is still GPT-6.1 Sol, while the new Terra serves as a referee, specializing in scoring code and mission output.

This proves that GPT-6.1 Terra's logical reasoning and discernment capabilities must be extremely strong.

Clue 2: "Switching the Pillars" by GitHub Copilot

Microsoft's testing is just a stew, and GitHub employee Nate Finch is a Commit, and it's almost a live feed.

According to V Salio, GitHub Copilot is testing GPT-6.1 Terra during a code review of the production environment.

The most explosive point is that：Terra is fully replacing GPT-6.1 Sol! in its highly rational code review, review, integration and CI review phase.

Furthermore, the insiders have explicitly described Terra as a "cheaper alternative to GPT-6.1 Sol".

A lower-priced, cheaper model replaces Big Brother in the most difficult logical review tasks.

It's just the opposite of heaven.

Why is the developer community so confused about Terra?This completely breaks the naming rules of OpenAI in the GPT-6 era.。

It should be noted that the current model of OpenAI has strict rules for naming planets/stars.

Luna (the moon)It's a great way to get started: lightweight, fast, cheap entrance fees.

It's the Earth.In the middle of the night, there is a big difference in price and balance.

It's the sun.High-end mid-range power, close to flagship performance, is the absolute driving force.

Astra (star)It's absolutely flagship and represents the current smart ceiling.

According to this logic: Luna, Terra, Sol, Astra.

GPT-6.1 Sol is currently priced at about $2 input/output of $10 per million tokens. It has a context window at 1.05M and is the flagship Astra in Agentic Coding and Computer Use.

So the problem is, as the "Old 2" Terra below Sol, if it's really like what it shows in the leak, it can outperform Sol in high-difficulty code reviews and sell cheaper...

It's also called Terra, doesn't it directly squeeze Sol?

The Big V Token Gremlin has sent out a breath of fresh air:

You're right. A lower-level model than Sol beat Sol, and it's cheaper. This directly violates OpenAI's own naming logic.

Luna, Terra, Sol, Astra used to represent the progression of abilities, and now these names look more like a price tag than a grade of ability.

If true, then the meaning of Terra will go far beyond its name.

Many people were excited about the GPT-6.1 Terra explosion.

Cheap and big bowls, this is the god of workers!

The big V LuminaBench gave a very high rating: "I think the GPT-6.1 Terra is going to be so strong that OpenAI can't release junk at this point in time. I'd like to see something like the Opus 5.5 play a low-level model straight to the big cup!"

Many people are excited to say, "I sincerely hope that the 6.1 Luna will be able to work together and will definitely be the driving force behind the coding".

Of course, there are also breaches:We want flags!

Someone said, "OpenAI is completely lost now. 6.1 Sol is rubbish when it comes to 3D modeling and spatial perception, and to be honest, Astra is nowhere to be found. I can only model with Opus 5.5 and Gemini 4 Argon, but it's too expensive".

Big V Chubby also wrote: "Luna is cheap, Sol and Astra are intelligent, and Terra is the headache-stricken old man, no one hurts, no one loves. Please, OpenAI, don't send".

Why is OpenAI pushing Terra so hard at this point?

The answer is simple: because their backyard is about to be burned by Anthropic!

Recently, there has been a tremendous migration of users in the AI world.

According to data from OpenRouter, an API aggregation platform, since the release of Claude Opus 5.5, there has been a significant increase in the number of users of the platform.The seven-day average token consumption of the Anthropic model has skyrocketed by 3.6 times!

It was a full-scale stampede, and not only did its older brother Opus 5.5 seal, but its younger brothers Sonnet 5.5 and Haiku 5.5 also make a name for themselves.

According to data from Lisan al Gaib, OpenAI is even more fucked up.GPT-5.6 and GPT-6 are used, completely pressurized by the Opus series!

Andrew Curran, the great V, told X:

For the past two weeks, I've been in the flow of information.Because Opus 5.5 announced the transfer of the Anthropic camp to an unprecedented scale.！

He specifically declared his absolute neutrality: "I didn't take any money from any lab".

And even if Opus were defeated, they still had Fable 5.5 on their hands.

Ethan Mollick, a professor at the Wharton School of Business, has even rebuilt his new book website by writing his own code using only the long text and reasoning skills of Opus 5.5.

Intelligent developers have now crafted an unrivalled set of tricks to "dry Anthropic": simple everyday tasks with ultra-cheap Haiku 5.5; hard-core logic and complex architectural reasoning with Opus 5.5.

This attempt to make the cheapest models explode calls has also allowed Anthropic to fill a niche in high-end computing consumption.

That's the fundamental reason why OpenAI has to pull Terra out of this knot!

In front of the Anthropic, "High End Opus + Low End Haiku", OpenAI's mid-low end shield river is collapsing.

A report by The Information also confirmed this: OpenAI's core goal in preparing the new Terra is to "push down prices while strengthening performance to meet the price competition from Anthropic".

Terra had to provide coding capabilities beyond the old version of Sol, at an extremely low price, to become the "Guardian of the Walled River" who blocked the vulnerability.

Even if it means "stabbing" your own Sol, OpenAI must also play this card.

Not only that, but LMSYS Chatbot Arena also has new clues.

On October 23, two mysterious new models appeared at the Arena:shooting-star-red-alphaandshooting-star-green-alpha。

The mysterious model on the Arena is usually the clearest early release signal.

What exactly are these two models with the "meteor" code?

Some speculate that this is the GPT-6.1 Terra with different parameters or different architectures (such as the more radical MoE architecture); others speculate that it may be a completely new visual model.

In short, OpenAI has a lot more to offer.

Now, OpenAI is back in the battle for "sexual equality".

GPT-6.1 Terra's leak may tell us that peak performance is important, but whoever brings strong reasoning to the table can really win in the end.