LLM benchmarks, pricing & model intelligence. Finding the interesting numbers behind frontier AI.
England, United Kingdom
Joined June 2025
- Tweets264
- Following344
- Followers17
- Likes132
Pinned Tweet
Seeing all the Opus 5.5 motion design videos lately, I had to try one too.
I gave Claude our site — llmlearner.com — and asked it to turn the site into a 15-second launch film.
Here’s how it turned out 👇
All rendered from code: HTML → headless Chrome → FFmpeg.
DeepSeek has released a new feature in DeepSeek Harness that enables users to instantly create plugins, games, and designs. Well done!
OpenAI is merging ChatGPT and Codex usage limits into one shared pool starting in December.
That could be bad news for ChatGPT Plus users who rely heavily on Codex.
I went through my own Codex usage logs to see how fast it actually burns through the limits. Here's what I found
OpenAI seems to be giving up on the $20/month Plus membership—it has almost no advantages left. The usage quota is very limited and can hardly compete with Claude's. GPT-6.1 isn't as good as Opus 5.5, and the weekly usage limit is much lower than Claude's. Strange strategy 🤔
Serious issue with GPT-6.1 Terra in Codex: very lazy! Asked it to search all it could on Gemini 4 Carbon and save it. It did basics + benchmarks, stopped, said only basics were saved. Told it to continue; it stopped again after reformatting. It stopped 4 times before finishing.
A striking result in the new AA Index 5.0:
GPT-6.1 Sol (max): 54.5, $1.27/task
Claude Sonnet 5.5 (max): 54.2, $10.02/task
Nearly the same score, but Sonnet costs ~7.9× more per task.
On v4.3, Sonnet was ahead by 4.2 points. The new index flipped the order.
Interesting~
We built a visual breakdown of Anthropic's IPO S-1 on LLMLearner:
llmlearner.com/providers/ant…
The numbers are remarkable:
• H1 2026 revenue: $16.3B (14x YoY)
• Q2 operating profit: $974M
• H1 net loss: $80.7B
• Operating cash flow: +$3.93B
Key findings 🧵
Anthropic has filed an S-1 registration statement with the SEC related to its proposed IPO.
Read more: anthropic.com/news/s1-sec
Legal disclaimer: A registration statement relating to these securities has been filed with the SEC but has not yet become effective. These securities may not be sold, nor may offers to buy be accepted, prior to the time the registration statement becomes effective. This post does not constitute an offer to sell or the solicitation of an offer to buy these securities, nor shall there be any sale of these securities in any state or jurisdiction in which such offer, solicitation or sale would be unlawful prior to registration or qualification under the securities laws of any such state or jurisdiction.
5/6 Anthropic is buying NVIDIA GPU compute from SpaceXAI—a SpaceX subsidiary
Contracts total UP TO $120.5B through 2029.
At full delivered capacity (expected Dec 2026), fees would be ~$3.1B/month. After early 2027, deals have 90-day cancellation terms.
LLMLearner retweeted
Thanks @LLMLearnerAI for adding the agent-focused CyberGym category after our data note. Lyrie Agent now correctly listed at 99.2% — top reported agent score, separate from model rankings. Official source: cybergym.io
Maybe we can see the full scores versus cost for Terminal Bench 4.0 here. Clearly, Kimi K3.1 Pro is more expensive than GPT-6.1 Sol but performs better. Source : llmlearner.com/rankings/term…
First, persistent prompt caching. Now, switching reasoning effort without cache invalidation.
It feels like OpenAI is making cache management increasingly independent of conversation settings.
Is this mostly clever prefix management, or something deeper at the KV cache layer?
Confirmed! Thariq left Anthropic!
I’ve left Anthropic after more than a year of working on Claude Code and seeing it scale larger than I ever imagined.
I have really loved working at Ant and believe deeply in the mission, and most of all, the idea of a race to the top.
The idea that if we work hard and responsibly, we can raise the bar and inspire others to great outcomes. And if we all push each other to be better, things can be radically good. I will carry the spirit of this with me forever.
My colleagues at Anthropic are the most talented, thoughtful, competent, and earnest people I have ever met. I am glad they are at the wheel and trust them deeply.
To the community on X, thank you for everything. You’ve made this the most fun (and sometimes stressful) job I’ve ever had.
For me, it is time for a bit of a break before diving into something new.
If you want more excellent Claude Code content, you should follow @addyosmani , @lydiahallie, @_catwu, @ClaudeDevs and of course @bcherny who started it all.