It's the first time in the history of the world.
In the next 28 days, ChatGPT will be bloody alive!Just as the "Godfather" Tibo raid announced the Day 1 plan, the whole network exploded.
The default reasoning of GPT-6 Astra and GPT-6.1 Sol is 50% faster
TPS jumped from 30 to 50, covering all ChatGPT logins, as well as OpenCode, Pi, Amp, Devin, and more.
Additionally, the GPT-6 system also replaced "optimized terminology" to reduce the amount of tokens needed to complete the same task.
The AI circle didn't even get to shout "Ottoman consciousness", and then it went straight to Silicon Valley.
SemiAnalysis released a limit test, which compared the subscription value of the two giants across the Internet, and found:
In the case of Agent programming heaviness, the actual API value of Opus 5.5 from Anthropic is more than five times that of GPT-6.1!
Today, the GPT-6 system is running faster, but is it worth the money?
GPT-6 is 50 percent faster.Day 1 is up.
Just yesterday, Tibo set up a flag for himself on X.
For the next 28 days, every day, either a visible improvement was posted online or it was re-posted to everyone.
28 days are counted, starting today.
The first day of the update hit the fastest point of developer reasoning.
As of today, the default generation speeds of GPT-6 Astra and GPT-6.1 Sol have increased by 50% from the previous 30 TPS (Token per second) to 50 TPS!
What's even more disturbing is that this acceleration not only covers all of the official endpoints of ChatGPT, but also the entire ecosystem:
- Devin is a well-known AI programmer
- Personal terminals and workflow tools that exploded OpenCode, Pi, amp
Wherever you call, you don't need to change a single line of code, and as long as you go through the subscription channel, you can speed up instantly!
Officials say this is not just a synchronous adjustment in computing, but more importantly a replacement for the extremely radical underlying Tokenizer {\\displaystyle Tokenizer}).
This means that in the face of the same long-code generation, complex mathematical analysis, the total amount of tokens consumed by the model at the bottom is squeezed to an extremely low level, allowing effective throughput to achieve a qualitative leap.
However, many developers have observed that the sensation is not fast enough and that it is necessary to go through the API to reach 50.
 |  |
Claude subscribes to GPT five timesAuthoritative testing of faces
Today, the renowned semiconductor and AI research organization SemiAnalysis released a test report called "facing".
They directly purchased high-priced subscription services from OpenAI and Anthropic to test long-term programming and Agent-equivalent loading tasks until they reached the weekly usage limit and then discounted theoretical fees according to public API prices.
The result is shocking.
At $200 per month, the equivalent API value of the ChatGPT Pro correspondent has dropped from about $14,000 to about $2,084 after the benchmark jump.
Claude Max's 20x model is about $11,726, which is more than five times the gap.
- $200 file: Claude's equivalent value is about 5.6 times the value of ChatGPT
- $100 file: Claude is about 5.4 times the size of ChatGPT
- $20 file: Claude is about 5.6 times the size of ChatGPT
Chetaslua, a renowned AI company, created a visual contrast between ChatGPT Pro and Claude Max bills.
The most bizarre operation is still behind us.
Many people remember that before June of this year, OpenAI was the "price butcher" with the highest-end packages capable of running almost 140,000 dollars.
Last week, however, the Ottman countered with a knife, directly cutting the actual call limit of the $200-a-month ChatGPT Pro top package by half.
The real reason behind this is hidden in the subscription data dismantled by SemiAnalysis.
In the example of Claude Pro and Claude Max 5x, anthropic can achieve a net loss balance when the actual user consumption reaches around 20%.
OpenAI's situation is more serious: ChatGPT Plus and Pro 5x subscriptions, which start losing once the utilization rate exceeds about 11.4%.
In other words, these subscription plans are essentially aimed at dissatisfying the majority of users.
But the popularity of Agent programming is breaking this balanced, multi-step, autonomous calling tool workflow, with Token consuming up to a thousand times the level of traditional single-step conversations.
The super app is coming!The four main entrances are fully integrated
In addition to this, OpenAI is opening a bigger chess opening.
Recently, OpenAI plans to merge the four main product lines of Chat, Work, Codex and Dots into a single interface, transforming it into a single super application.
This is not only a UI-level integration, but also a shift in the entire product philosophy.
Specifically, the changes include:
The model selector is hidden, and the reasoning process is no longer visible to users.
In the future, everyone will no longer be able to choose which model to use, and we don't know how many steps AI is "thinking", you just have to tell it what to do and wait for the results.
All text generated by ChatGPT and Codex will be required to add an invisible watermark to identify AI content.
Just this morning, OpenAI announced the announcement.
Additionally, OpenAI's commercialization strategy is also being adjusted, allowing users to see inserts while waiting for image generators to process images.
ChatGPT's Chat and Work models are expected to be combined to reduce complexity, and the full capabilities of Dots will eventually be built into ChatGPT, covering 1.2 billion users.
Today, Day 1,50 TPS has been rolled out.
There are 27 days remaining until the real OpenAI test!
RoboInd editorial channelResearch, robotics companies and market intelligence for cross-border sourcing.Explore the channel ↗
Share this
End of editorialExplore more from Global Robotics News ↗