🚨 Sonnet 5.5 incoming reportedly in partner testing now, could launch Sep 30–Oct 1. Same 1M context, adaptive thinking, same pricing as Sonnet 5. Hoping for Opus 5.5-level quality without the bloat.
Think it delivers? 🤔
🚨 Bel Leaks: OpenAI's Internal AGI Model
> OpenAI reportedly has an internal model called Bel
> Bel is said to be the model that cracked the Navier–Stokes Millennium Prize Problem
> One of the deepest unsolved problems in math
> Also reportedly used internally (as Bel) to help train Astra
> Could become the foundation for an AGI-threshold model
After cracking Navier–Stokes,
> Bel reportedly hit an estimated 80-hour countdown
> Said to be ahead of the ASI 2027 projection
> Bel reportedly exceeds 10T total parameters
> Also could arrive before 2027
> After GPT-6.1 Astra ships
Update: 🔥 ChatGPT Voice just got a MASSIVE upgrade
OpenAI says Voice can now:
• Tap into plugins like email, calendar, and Slack
• Run on GPT-6 Astra, Sol, and Luna
• Function inside ChatGPT Work on web & mobile
• Build docs, decks, websites, and spreadsheets all by voice
• Handle complex browser tasks hands-free
Rolling out globally TODAY in the latest ChatGPT app.
🚨 GPT-6 Sol and GPT-6 Luna Just Launched
OpenAI has rolled out their new models
Pricing:
GPT-6 Sol: $2 input / $10 output
GPT-6 Luna: $0.10 input / $0.50 output
That's 50% cheaper input and ~58% cheaper output for Luna
Going live today in ChatGPT Work and Codex for Plus, Pro, Business, Enterprise.
🚨 OPUS 5.5 Officially Launches Today.
> Opus 5.5 surpasses Fable 5.1
> Roughly 20% cheaper than Opus 5 and Fable 5.1
> Could finally push Claude's flagship models toward real affordability
> Also edges out the current GPT-6 Astra
Tonight is about to get absolutely unhinged.
🚨 GPT-6 Bel Leaks: Surpasses GPT-6 Astra
>OpenAI reportedly wrapped its next major pretraining run, codenamed Bel
>Bel is said to be the successor to Doug
>Reportedly exceeds 10T total parameters
>Bel could serve as the foundation for models past GPT-6
>OpenAI reportedly views Bel as a potential post-GPT-6 base
>It might even become the foundation for an AGI-threshold model
Is this the model that finally crosses into AGI?
🚨 GPT-6 Luna may already be appearing inside Codex backend responses
A commit pushed today reportedly reveals that certain "gpt-5.6-luna" requests are now coming back as "gpt-6-luna".
The unexpected model name even caused issues with OpenAI’s internal cost tracking because the system didn’t recognize it.
Luna still isn’t publicly available or directly routable, but this strongly suggests OpenAI is actively testing it internally,.
🚨 Opus 5.5 could reportedly drop next Tuesday
"Leaker Lyra"
a new checkpoint is already available. The model we’ve been testing is "claude-opus-5-5", reportedly under the codename "claude-wafer-cap".
Next week could be absolutely insane GPT-6 Sol, Opus 5.5, and even more
🚨 Opus 5.2 Leaks: Surpasses Fable 5.1
> Opus 5.2 reportedly under evaluation inside Claude Code
> Anthropic appears to be skipping over Opus 5.1 entirely
> The "claude-opus-5-2" tag has supposedly surfaced in Microsoft Foundry
> Claude is already reported routing some traffic to Opus 5.2
> Opus 5.2 is said to mark a major jump over Opus 5
Could Opus 5.2 be Anthropic's true comeback move?
🚨 GPT-6 LUNA & TERRA: NEW DETAILS.
A few updates on the two models everyone keeps asking about:
> Luna is still being worked on, while Sol appears to be nearly finished
>There’s still a possibility Luna launches alongside Sol
>Terra’s future is much less clear
>OpenAI is reportedly considering dropping the Terra tier entirely
>If that happens, we might never see a GPT-6 Terra
The next couple of weeks should tell us a lot, with a first real preview potentially coming as soon as next week.
Nothing is locked in yet. Things can still change.