GPT-6.1 Sol Matches Astra for 1/5 the Price. What Is OpenAI Doing?

Author: Universe of AI · Watch on YouTube

Video authored by the channel shown, not by IANIA; all rights belong to the creator. Transcript and summary generated automatically; the original-language content is preserved.

Original language: English

Original video: GPT-6.1 Sol Matches Astra for 1/5 the Price. What Is OpenAI Doing? · Channel: Universe of AI · 13:04 · Watch on YouTube

Topics: AI agents, Task and workflow automation, AI developer tools, AI models

Summary

The video recaps OpenAI Dev Day 2026, grouping its many announcements across AI agents, models, coding tools, APIs, plugins, collaboration features, and subscription plans. It focuses on GPT-6.1 Soul, which the presenter says offers performance close to Astra at a much lower price, while noting that OpenAI still trails some recent Anthropic models. Other announcements include the always-on Dots agent, faster model tiers, Codex updates, new developer APIs, and tools for people and AI agents to work together. The recap also covers expanded plugins, partner integrations, and an OpenAI marketplace.

Key points

OpenAI announced more than 20 updates at Dev Day 2026 across five broad categories.
Dots is described as an always-on AI agent with its own cloud computer and browser, access to thousands of apps through plugins, and the ability to learn user preferences.
GPT-6.1 Soul is presented as near Astra-level intelligence at about one-fifth the price, with listed pricing of $2 per million input tokens and $10 per million output tokens.
The presenter says GPT-6.1 Soul replaces GPT-6 Soul shortly after its launch and still falls behind some recent Anthropic models on the cited index.
OpenAI introduced an ultra-fast speed tier, with availability described for Astra in the API and selected ChatGPT and Codex plans; Soul ultra-fast was said to be coming soon.
The new Pro 500 plan offers 25 times the Plus usage allowance and includes ultra-fast access, aimed at heavy Codex users running multiple agents.
Codex updates include cloud task execution, voice control, an agent view, code review, and security scanning for repositories.
New and upgraded developer tools include a Decisions API, computer use and multi-agent support for the Agents API, and support for running OpenAI agents on AWS.
OpenAI is expanding its plugin platform, including interactive extensions, plugins for sites, and automations triggered by connected-app changes.
New shared spaces, pages, slides, teams, and tasks are intended to help people and AI agents collaborate; subscriptions also include usage across partner products and access to a marketplace.

Chapters

Start	Chapter	Summary
00:00	Dev Day overview and Dots	The presenter introduces OpenAI’s Dev Day announcements and describes Dots as an always-on agent that can connect to apps and learn user preferences.
02:05	GPT-6.1 Soul and sponsorship	The video introduces GPT-6.1 Soul’s pricing and positioning, with a sponsored Printify segment before returning to the model discussion.
03:37	GPT-6.1 Soul compared with competitors	The presenter compares the new model’s performance and price with Astra and discusses OpenAI’s competitive position against Anthropic.
06:16	Speed tier, plans, and Codex updates	The recap covers ultra-fast access, the Pro 500 plan, and Codex improvements for cloud work, code review, and security scanning.
08:55	APIs, plugins, and integrations	The video outlines the Decisions and Agents APIs, AWS support, and new ways to build and use ChatGPT plugins and automations.
11:00	People and AI working together	The final section describes collaborative workspaces and shared documents, partner usage allowances, and the OpenAI marketplace.

Quotes

which is near Astra intelligence at 15if of Astra’s price. — 02:05

So that’s why they’re positioning this as a model that is getting close to Astra at a fraction of the cost. — 04:08

at the moment Enthropic has a short lead on OpenAI — 06:16

a new top plan with 25 times the plus usage allowance — 07:21

they’re trying to create like a whole environment where AI agents and humans can now work together — 11:32

Concepts and entities

OpenAI, GPT-6.1 Sol, Astra, DOTS, Codex, Agents API, Decisions API, Pro 500, Anthropic, Opus 5.5, Sonic 5.5, Printify

Full transcript (original language)

00:00 So, OpenAI just wrapped up their dev day 2026 and they came out with more than 20 announcements in one go. A new model, a new speed tier, a new pro plan, and a bunch of codec updates and a whole new kind of agent that’s supposed to work for you around [music] the clock. So, if you try to read through the recap package, it’s a lot. But luckily for you guys, you can watch today’s video and understand what exactly dropped today. So, let’s get into it. Before we get into the video, I just want to take a moment to thank you guys for all the support over the last couple of months.

00:30 Every single view, every comment, and every share is helping this channel grow, and I don’t take any of that for granted. And if you’re new here and haven’t subscribed yet, hit that subscribe button so you don’t miss what’s next in AI. It’s free, and it genuinely helps grow this channel a lot. Thanks for watching today’s video. Now, let’s get back into it. So, all of today’s launches are categorized into five categories. new ways of working with obviously OpenAI, building with the codecs and the API, customizing your chat GBT with plugins, people and AI working together, and doing more with

01:02 your chat GBT subscriptions. So, we’re going to go through exactly what’s new today. Starting off with DOTS, which is your always on AI agent powered by GPG6 Astra, each with its own cloud computer and browser. It can also connect up to 4,000 plus apps through plugins and learn your preferences over time. The way I’m thinking of this, this is kind of like your Alfred. You work with it, you talk with it, it understands how you work, and based off of that, your experience changes. It also does proactive research in the background with readonly access. You can also reach

01:33 them in chat, GPT, message or voice call, Slack, and Teams with texting coming soon. There’s also custom rules that allow you to, you know, allow a specific action, require approval if you want it, or block it. This is going to be rolling out to pro and business premium users at the moment and then the healthcare education users are going to get like a beta testing of it and Microsoft agent 365 integration is also in the progress. So this is a new type of AI assistant that they’re creating. Obviously this is still early days so we will see this evolve over time. Then we

02:05 actually have a new model today called GBT 6.1 soul which is near Astra intelligence at 15if of Astra’s price. It sits at $2 per input and $10 per output and 10 cents cash input per million tokens. Before we continue, big thanks to Printify for sponsoring today’s video. Most of you watching are already making stuff with AI. Images, logos, random designs you generate at 1:00 in the morning and then never do anything with it. Printify is how you turn that into a product someone can

02:35 actually buy without paying for inventory upfront. I took my own channel logo, gave it a Halloween version, then I dropped it onto a mug inside Printify. You upload the file, drag it wherever you want it, and it shows you the mockup right there. The catalog has over a thousand products, so the same design goes on hoodies, tote bags, t-shirts, whatever you think people will actually buy. And Halloween is the easiest season for this. People hunt for spooky stuff for about 6 weeks and then it’s gone. So if you got a design sitting in a folder somewhere, this is the window. And with

03:06 Printify Premium, you get up to 20% off the entire catalog. They print locally through their own network across the US and Europe, so orders get made close to whoever bought [music] them, faster shipping, and no extra fees on your end. And it plugs straight into Etsy, Shopify, plus Tik Tok shop if that’s where your audience is already. You publish and it’s live. Nothing to pay up front, nothing to learn before you start. And the link is in the description. Now, let’s get back into the video. Before we go back into what else dropped on dev day today, I’m going

03:37 to take a moment to talk more about GPT 6.1 Soul, because quite frankly, based off all the drops that they have launched today, majority of us are probably going to be interacting with this one the most just because it’s going to be a new model that’s going to be available to all of us. Now, one thing you can clearly tell is that they’re positioning this as a model that’s going to replace GPT6. So, which I think is kind of a funny/ embarrassing thing about opening eye because they realized when they dropped GPT6 soul. It was not competing with anything. So, now

04:08 they already replaced that model. You can see they didn’t even try to put GPT6 soul over here. Use 6.1 soul instead of the old six already and it’s only been a week. Now, this pricing for this is $2 per input and $10 per output as I mentioned. And if you compare this to Astra, this is pretty much like five times cheaper than it. So that’s why they’re positioning this as a model that is getting close to Astra at a fraction of the cost. If we take a look at a couple of the charts that we see here on the deep software engineering, we can see that GPT 6.1 soul is actually

04:40 obviously beating GPT6. And then when it comes to Astra, it’s getting to that level of performance at a cheaper cost. And that’s the whole positioning behind this model. It’s basically like if I were to predict, they took Astra, distilled it down a little bit more, made it faster, and also made it cheaper, and gave us GPT 6.1 soul. And here’s another chart that kind of shows the same kind of result where GPT6 soul is but down here. GPT 6.1 soul is matching Astro level performance and is at a cheaper cost. On the artificial analysis index, this model now sits at

05:12 52. So, it pretty much is still obviously not overtaking any of the models that Enthropic has launched recently like Opus 5.5, even Sonic 5.5 which came yesterday or obviously getting to 5.1 level. It is one point behind 5.1 and GPT6 Astra which I think is a good win for Open AAI at the moment. But this is still a little bit, you know, disappointing because they don’t have a model yet to compete with even Sonic 5.5. And as this person says, it looks like Enthropic basically forced them to drop a cheaper version of Astra.

05:45 So I think this is just me predicting. They took Astra and distilled it like I mentioned. And that is why we see this model out now because at the moment Enthropic is clearly taking the leaderboard with Claude Opus 5.5, Sonic 5.5, and Fable 5.1 that matches GPT6 Astra. And as I mentioned, I think Enthropic kind of put pressure on Open Air because they kind of botched the six soul launch. We know that and you know Opus 5.5 came out last week when they dropped six soul and that was overshadowed obviously by the performance of Opus 5.5 and this week

06:16 Enthropic kind of you know spoiled open AAI’s dev day by dropping Sonic 5.5 yesterday and that is why we see them already replacing a model that just literally came a week before you know what I mean so yeah at the moment Enthropic has a short lead on OpenAI but this space moves very fast we will see them obviously drop a more capable model and things are going to ship very quickly. That’s how it goes. But anyways, let’s get back into the dev day. There’s also a new ultra fast mode which is premium speed tier up to eight times faster in codeex about 300 tokens

06:48 per second and up to six times in the API. The use case for this would be like let’s say you’re under pressure or you need to spit out something quick or you need to get things done very fast. Obviously, this would be a tier that you would enable in that moment. And the availability for this is that Astra ultraast is going to be live in the API and in chat GPT work and codecs on pro 500 and enterprise. Soul ultraast is going to be coming soon. Now we’re going to go skip to one section called do more with your chat GPT subscription because there is a new protra called pro 500

07:21 which is a new top plan with 25 times the plus usage allowance and ultraast is included in there. And the use case for this is going to be for someone who’s a heavy codeex user who wants to run several agents in parallel all day. So you would be probably leaning towards this tier now. And you also get ultra fast with it which is I guess a benefit. Now if we go to codeex codeex in the cloud is basically now you can run tasks locally from your phone or in the cloud with reusable team dev environments. You can kick off a refactor on your laptop close it and check your progress from

07:52 your phone. The availability for this is for plus pro business healthcare education and enterprise users. There’s also a new refresh codec which now has voice control, a new agent view for tracking several tasks, a cleaner UI, and better session resume and work tree support. If you love using the CLI, you’re going to be happy with today’s update and you’re going to get in all your plans. There’s also a code review which summarizes diffs and issues Q&A in the desktop app with automatic first pass reviews in the cloud. This is going to be more useful if you are actually a

08:24 heavy code user and you’re trying to review your work. The new interface is going to make it much more enjoyable. There’s also codec security cloud which scans repos on demand or on a schedule, removes duplicate findings and prepares fixes. It also includes daybreak blue models. And a use case for this would be for example a weekly schedule scan of every company repo that turns up vulnerabilities with fixes ready to review. The availability for this is pro business enterprise and edu on desktop and web. And then we also have a new decisions API which is supposed to be

08:55 all about taking action in real time using a fast model like Luna. And this is going to answer specific set of userdefined questions with you know finite predefined answers. Developers can supply context using text or images. with this uh use case for this could be like routing each support ticket to billing technical or sales that already have some predefined answers or flagging content that breaks policy. This is going to be in a limited preview for the moment with a broad release in the coming days. We also have agents API which now is going to support computer

09:25 use so your agent can operate software plus multi- aent support. There’s tool search and context compassion. So this is basically them upgrading their agent API package for all of us. Then this is going to be available through the API and in codeex and chat GP work on pro 500 and enterprise plans. They’re also updating the bedrock manage agents. So open AAI agents can run entirely in AWS and connect to AW resources. So if a company is already building on AWS, you can now use OpenAI agents on there with much more ease. So that is a plus for

09:57 those users. They’re also opening up with their platform for plugins. So developers can actually now create their own experiences with ChatGpt. You can have plug-in extensions that give your plug-in a home in the sidebar and you can build interactive panels. This is going to be available to all the plants. And they’re also trying to basically create a app store for plugins. That’s their end goal. That’s what it seems like at the moment. And that is why they’re making it easier for users to create their plugins, submit them for OpenAI as well, help them discover new plugins, help consumers discover new

10:28 plugins. This is also going to be available to all plans. Now the sites features that they have can also now host plugins. So you can now add supported chat GPT plugins to the sites that you are actually building. And then teammates who are working in your workspace can also use that same app with their own connected data and permissions. They’re also making automations easier to add and manage to sites so you can keep shared information up to date. They’re also combining MCP events and automations. So now plugins can trigger automations when something changes in a connected app. So this is

11:00 making their whole plug-in infrastructure much more efficient. OpenAI also made it easier for people and AI to work together today. So they have added a bunch of features here. First one is chat GBT space which is going to be a shared hub where teammates chat GPT. So think of AI agents and dots build on the same knowledge. So you can have a marketing team with AI agents and actual humans working together. There’s also pages which is going to be a shared document for people and AI agents. So think of like a word document, but now your AI agent can work with you while

11:32 humans are working in there. And that’s the same idea with slides. AI agents and humans working together. You can also create teams and share tasks. So you can assign tasks to your specific team. You can also assign it to your AI agent, which I think is really cool because they’re trying to create like a whole environment where AI agents and humans can now work together because that’s the reality. People are working with AI and humans are also working with AI. Why not bring it together in one environment? And these are some additional plugins that they’re adding where they’re making it easier to work with chat GPT in Teams

12:03 and Slack. So, if your teams are already working there, which most teams are, you’re going to be happy with this. There’s also a meeting plugin that takes notes for you and saves personalized summaries with action items for you, so everyone can save some time. And now they’re also creating something that looks kind of like GitHub where you can share profiles and let people you know showcase the sites they have created plugins and things like that and people can share what they have done with other people. So it’s kind of becoming like a GitHub. Point thing is the top plugins over here are Slack, Google Calendar and

12:34 GitHub. Also now with your chat GPT subscription you’re now getting an allowance and you can use your allowance across 16 partners. Some of them being Cognition’s Devon, Notion, Versal, T3, OpenC Claw, and more. So now you can use AI inside notion or Devon through your existing Chad GBT plan instead of paying for another subscription, which I think is pretty sick. And lastly, we have a OpenAI marketplace, which is part of OpenAI’s commitment toward partner software from 32 launch partners. And you can see them all on screen like you

13:04 can see HubSpot is here. We also have Adobe, lovable, and more. But that is pretty much a wrap-up of what OpenAI launched today on Dev Day. But that’s it for today’s video. Make sure you guys are subscribed to the channel. Follow our new newsletter as well at universeai.behive.com [music] as well as subscribe to the main channel World of AI and support us on X by following the Universe of AIZ as well. Until then, I’ll see you guys in the next