Newshot

GPT-6.2 Astra and Sol Drop as a Double Blockbuster! 3D Is in Trouble Again

AI summary

Sol comes close to flagship performance.

Translation(model translation, not the source's own text)

GPT-6.2 Astra and Sol Launch as a Double Blockbuster! 3D Is in Trouble Again

Mengyao, reporting from Waifeisi

Quantum位 | WeChat Official Account QbitAI

OpenAI, you really know how to play the double-blockbuster card.

GPT-6.2 Astra and GPT-6.2 Sol, this time they've arrived together!!! (Cheers.jpg)

Gotta say, Astra's performance across the board has been pushed to the limit this time—

  • Complex Agent capabilities upgraded: Astra's professional task benchmark improved by 5.2 percentage points over the previous generation, with task cost reduced to less than one-third.
  • Scientific research capability sets a new record: related benchmarks improved by 9 percentage points, and OpenAI directly calls it the world's strongest scientist model.
  • Coding capability upgraded again: scored 44.2% on the FrontierCode 2.0 benchmark, a 1.4 percentage point improvement over the previous generation.
  • 3D generation tops the leaderboard: took first place on Artificial Analysis's AA-3D benchmark with 1318 Elo, surpassing Fable 5.2's 1271 points.

Just how capable is the model? Let's go straight to the talent show~

Take a look at this—Astra transforms into a web designer, handling everything from layout to imagery, and can even craft tailored visual styles for different websites:

Now look at this—using only photos and a manual, Astra modeled a remote-controlled car down to the engine, and even put together an assembly animation. Pretty impressive, huh:

On the AA-3D benchmark, it topped the leaderboard with 1318 Elo, cranking out detailed 3D models on demand and even generating immersive interactive scenes directly, be like:

△x@fal

The lightweight Sol is equally impressive, with coding capability approaching Astra at only one-fifth of its task cost.

On pricing, 6.2 Astra's API price dropped directly by 20%—

$8 per million input tokens, $40 for output, and cached input pricing dropped by 60%.

6.2 Sol maintains its original pricing, with API input and output at $2 and $10 per million tokens respectively.

Both models have now landed on ChatGPT Work, Codex, and the API, available to Plus, Pro, and other subscription users.

Without further ado, let's dive straight into the specific capabilities of the new models~

Astra Sets New Records on Professional Tasks, Sol Approaches Flagship Performance

First, let's look at just how much GPT-6.2 has upgraded this time.

In its official announcement, OpenAI positions 6.2 Astra as a flagship model for high-difficulty projects.

6.2 Sol handles high-frequency, everyday professional tasks.

I've also put together a quick summary of the capability highlights of the two models for everyone—let's take a first look!!!

△AI-generated

From the published results, Astra's most obvious improvements are concentrated in complex task execution, scientific research, coding, and visual design—

Complex professional tasks set new records, with completion cost reduced to less than one-third of the previous generation.

On the Agents' Last Exam benchmark, GPT-6.2 Astra improved by 5.2 percentage points over GPT-6 Astra.

Compared with Claude Fable 5.2, it also leads by 5.5 percentage points, while per-task cost is only 5% of the latter's. It really has something going for it...

Scientific research capability also continues to improve, with OpenAI even directly giving it the positioning of "world's strongest scientist model."

On Terminal-Bench-Science 0.1, Astra improved by 9 percentage points over the previous generation, with task cost likewise reduced to less than one-third.

The Vals Science Index also set a new record, requiring models to understand real biological data to design experiments and infer the operating rules of simulated systems through experiments.

In other words, compared to simply answering scientific knowledge, it's now closer to the actual workflow of researchers~

Of course, coding, 3D design, and computer operation capabilities have also collectively gone next level—

On the comprehensive intelligence index of third-party evaluator Artificial Analysis, Astra scored 62, a 5-point improvement over the previous generation.

Sol matched the previous generation's flagship with 57 points, but at a lower cost.

3D generation is even more impressive: Astra topped the AA-3D leaderboard with 1318 Elo, surpassing Claude Fable 5.2's 1271 points, and Sol also took 1259 points.

OpenAI also demonstrated a rather interesting demo—

GPT-6.2 Astra went hands-on with mechanical design directly in CAD software.

It can process mechanical parts one by one, check the connection relationships between components, and even complete engineering tasks requiring such fine operations on its own:

Let's look at another demo below.

6.2 Astra directly simulated a car's driving dynamics on a mountain road.

It first creates a 3D model, then writes a Python program to simulate the motion relationship between the tires and suspension system, and finally uses Blender to generate the animation.

The car's driving visuals, suspension motion, and tire force data are presented in sync, making it easy for engineers to quickly test design solutions—

Let's look at another example in visual reasoning.

Astra directly whipped up an interactive 3D backyard garden.

With vegetable planting areas, fences, and greenery all in place, you can freely switch viewpoints and adjust flowerbed layouts.

It even simulates sunlight at different times of day, letting you intuitively see the lighting conditions for plants:

Cybersecurity capabilities have also seen notable improvement.

GPT-6.2 Astra and Sol have been fully upgraded in vulnerability discovery, exploitation, and code remediation.

In fact, both have reached OpenAI's highest risk level for cybersecurity capabilities.

On the DeepBinary vulnerability mining benchmark, the estimated total bounty for vulnerabilities discovered by Astra exceeded $1 million, far surpassing the previous generation's roughly $440,000:

After the flagship model showed off its skills, 6.2 Sol has its own highlights—

Especially in coding, it has begun to approach Astra.

On Cognition's FrontierCode 2.0 benchmark, GPT-6.2 Sol is only 0.5 points behind Astra, while task cost is only about one-fifth of the latter's.

In other words, Sol can already approach flagship model performance at a significantly lower cost in some practical development tasks.

On the daily task benchmark LifeVal, Sol even surpasses the previous generation GPT-6 Astra, at about one-seventh of the cost.

OpenAI also revealed that many employees internally have already made Sol their default model for both technical and non-technical work, which also explains the division of labor between the two models.

PPT, Excel, and Documents All Upgraded—GPT-6.2 Is Stronger at Office Work

Beyond scientific research and coding, GPT-6.2 has another upgrade this time that workers will probably feel strongly about.

Making PPTs, processing Excel, and writing professional documents—all strengthened!!!

On the GDPval v2 benchmark, which measures real professional work capability, Astra improved by 6.2 percentage points over the previous generation, and Sol also improved by 5.3 percentage points.

Moreover, the output tokens Astra needs to complete tasks decreased by 22%, so efficiency improved as well.

On the Excel side, OpenAI also had the old and new generations of Astra complete the same task: based on financial data, create revenue forecasts and valuation tables for a fictional company:

Both generations of models could complete the task, but the tables generated by GPT-6.2 Astra are clearly clearer and easier to use.

It puts the calculation process from revenue structure and enterprise valuation to per-share value, as well as valuation changes under different assumptions, all on the same page.

On the PPT side, OpenAI also had the two generations of Astra design animal-themed merchandise for four teams.

Compared with the previous generation, which designed only one character per team, GPT-6.2 Astra delivered 8 designs in one go, covering badges, T-shirts, backpack embroidery, and other formats:

In another case, OpenAI had the two generations of Astra create a one-page PPT analyzing the market for mid-to-high-end women's fashion brands.

GPT-6.2 Astra's output was clearly closer to a ready-to-present level—

It not only categorized brands by price and style, but also added brand logos, price ranges, and images and prices of 6 representative dresses:

For workers who deal with spreadsheets and slides every day, this upgrade is indeed quite practical.

Currently, GPT-6.2 Astra and Sol have both launched on ChatGPT Work, Codex, and the API, and are available to subscribers including Plus and Pro.

Those interested can give them a try~

[1]https://openai.com/index/introducing-gpt-6-2-astra-sol

  • 奥特曼 empathizes with 黄仁勋: “heroes also have flaws,” but supports him returning as 英伟达 CEO… 2026-11-17
  • 黄仁勋 prepares to return to 英伟达 2026-11-16
  • Another dark horse emerges among domestic large models, drawing Silicon Valley’s attention! 2026-11-13
  • First batch of

The translation is machine-generated and individual terms may be off. Nothing is rewritten or cut; the original is authoritative.

Published by 量子位
This site only summarises and aggregates; copyright stays with the original author. Headlines and summaries follow the source — where they differ, the original wins.
Read the original ↗