Dan
@DanDr1s
I track AI so you don't have to. Breaking news, model leaks & benchmark breakdowns.
1.7K Following 9.9K Followers
Haiku 5.5 beats GPT-6 Luna on every benchmark Anthropic compared
Computer use → 72.4% vs 48.9%
Terminal-Bench → 39.2% vs 16.4%
OpenAI's cheap model just got lapped
Haiku 5.5 is a significant step up over Haiku 4.5 across coding, computer use, and knowledge work.
OpenAI is putting Google's watermark in its own images
Nvidia and Kakao too, Apple next
Google basically won the AI watermark race
Our SynthID technology lets you easily verify if an image, video, or audio file is AI-generated, and today we’re expanding access.
Now anyone can go to to check files for a SynthID watermark — whether they were generated by us or industry partners like @OpenAI, @nvidia, Kakao, and, coming soon, @Apple.
To date, we’ve watermarked 180 billion images and videos and more than 240,000 years of audio content, regularly handling 1 million verification requests daily.
This new portal joins our built-in verification features already live in Search, @GeminiApp, and @googlechrome.
Show more
Claude mogs Codex on long agent work and it's not close
Codex → $400k in tokens, nothing
Opus → $20k, done in 2 weeks
Same project, same dev
Announcing tsc-rs (aka ts-rust), my complete rewrite of the Typescript compiler, type checker, and LSP, all in Rust.
This is a project I've had agents working on for 5 months now. I burned ~$400k of Codex tokens and got nowhere. Burned ~$20k of Opus and got there in 2 weeks.
I have not read a single line of the code.
tsc-rs is an open source drop-in replacement for tsc, and it's available now.
Show more
OpenAI just published 722 math manuscripts from an unreleased model.
They span 372 result families, after the model was tested on about 4,000 problems.
This is a insane amount of AI-generated math in one release.
Show more
We’re releasing a broad range of new mathematical results produced by an internal frontier model.
We’ve been consulting with the independent Advisory Group on Mathematics and Artificial Intelligence at the Institute for Advanced Study, and we have drawn on their advice and public recommendations to inform how we release these results.
Show more
Mistral just shipped the best open model outside China
DeepSWE:
Mistral Large 4 → 62%
GLM 5.3 → 61%
DeepSeek V4 Pro → 57%
Qwen 3.8 Max → 51%
1T params, 49B active, weights drop Oct 27
Show more
Meet Mistral Large 4, aka Le Chonk.
• 1T parameters, natively multimodal. 49B active.
It is the best open weights model from US or Europe on aggregated benchmarks.
• State-of-the-art on critical workloads, including cyber defense, manufacturing and finance and it surpasses closed frontier models on visual grounding.
• Forged in Europe end-to-end and is deployable from Europe via our own Mistral Cloud infrastructure.
• Available to all via API today. Working with cybersecurity partners privately.
Open weights release end of October.
Show more
🚨 Google just dropped Gemini 4 Argon
It's top or tied on 14 of the 19 benchmarks in Google's table, against GPT-6 Astra, Claude Opus 5.5 and Claude Fable 5.1:
- DeepSWE v1.1: 77.9% (Opus 5.5 74.2%, Astra 74.1%)
- AutomationBench: 51.3% vs 42.5%
- Long context (256K–1M): 84.2% vs 71.8%
- LVBench: 91.7% vs 87.5%
Insanely good benchmarks, but we have to see real outputs. Hopefully its not another benchmax.
Show more
🚨 Codex is now in the cloud.
OpenAI just shipped it at DevDay. You can run Codex from wherever you are:
- On your computer
- Remotely from your phone
- In the cloud from any device
Show more
🚨 GPT-6.1 Sol just DROPPED!
What's in it:
→ $2 in / $10 out, same price as GPT-6 Sol
→ 1.05M context window
→ 128K max output
→ Highest reasoning tier, up to "max" effort
→ Knowledge cutoff Apr 30, 2026
→ Built for coding, computer use and professional work
Show more
POV: you paid $200/mo for 20x
OpenAI: "everyone will get more in the end"
Also OpenAI: 20x → 10x
Something doesn't add up..
I’ll explain the new Pro 200 plan differently, before I start live tweeting from DevDay on things that are going out!
Today we are going to ship a number of things that increase what you can do across the Plus and Pro plans. A lot of compute is online for this increase. As we increase the floor, we are changing the relative difference between plans to be
Plus = 1X
Pro 100 = 5X
Pro 200 = 10X
and we are reopening subscriptions for Pro 200 (we had paused it). If you have an existing plan you will keep the 20X multiplier for a bit and also receive a lot of additional credits because we know changes are hard even if it means that everyone will get more in the end.
Show more
Sonnet 5.5 benchmarks are out and they're wild..
Terminal-Bench (coding):
Sonnet 5 → 10.3%
Sonnet 5.5 → 70.6%
that's 7x better, and it even beats Opus 5.5 (66.4%)
computer use:
Sonnet 5 → 57%
Sonnet 5.5 → 80.1%
On almost every other test it's within a few points of Opus 5.5, and it beats GPT-6 Sol wherever they were both tested
Show more
Sonnet 5.5 is out 🚨
• 1M token context window
• 128K max output
• knowledge cutoff Jan 2026
price: $2 input / $10 output per million tokens
1M context at $2 is actually insane
Show more
Elon, I'm going to be honest.
I don't think it's possible for Grok to catch up with OpenAI or Anthropic at this point.
They almost certainly have some impressive level of recursive self improvement going on internally. Their models are writing code, running experiments and speeding up research for the next model.
Every month they get faster. You're not just 3 months behind. You're 3 months behind a rocket that's still accelerating.
I genuinely hope you prove me wrong though. I'm rooting for Grok.
Show more
@notjazii I am cautiously optimistic that SpaceX will have a Fable/GPT-6 level model in 2 to 3 months
Just cancelled my Codex subscription. Going back to Claude.
Opus 5.5 was enough to convince me to switch back.
It mogs GPT-6 Astra. Easy decision.
Opus 5.5 absolutely mogs GPT-6 Astra.
Astra needs MAX just to lose to Opus on HIGH at roughly twice the cost per task.
🚨 GPT-6 Sol is HALF the price of GPT-5.6 Sol.
Standard API pricing per 1M tokens (input / output):
Sol: $4 / $20 → $2 / $10
Luna: $0.20 / $1.20 → $0.10 / $0.50
That’s 50% cheaper input and nearly 60% cheaper output for Luna.
Show more
🚨 Opus 5.5 JUST HIT 58 ON THE ARTIFICIAL ANALYSIS INTELLIGENCE INDEX.
The next highest score in this chart is 53.
Fable 5.1: 53.
GPT-6 Astra: 53.
What did Anthropic feed this thing 😭
Show more
🚨 Claude Sonnet 5.5 and Haiku 5.5 are coming in the next few weeks.
Both will get many of Opus 5.5’s performance, efficiency and safety improvements.
Anthropic is also raising Pro, Max and Team usage limits, plus giving users a banked reset.
Show more
🚨 Claude Opus 5.5 is OUT.
Beats Fable 5.1 on EVERY benchmark in this chart.
Even beats GPT-6 Astra on Terminal-Bench: 66.4% vs 57.9%.
And scientific research nearly doubles over Opus 5: 29% → 58.7%.
Okay Anthropic 👀
Show more
Opus 5.5 on medium beats GPT-6 Astra at its highest setting on GDPval-AA.
Estimated cost per task:
• Opus 5.5: ~$0.85
• GPT-6 Astra: ~$4.50
Better score. Roughly 80% cheaper.
Show more
🚨 Claude Opus 5.5 is OUT.
Beats Fable 5.1 on EVERY benchmark in this chart.
Even beats GPT-6 Astra on Terminal-Bench: 66.4% vs 57.9%.
And scientific research nearly doubles over Opus 5: 29% → 58.7%.
Okay Anthropic 👀
Show more
GPT-6 Sol today?
An OpenAI employee just posted “I’m sol excited.”
That’s a pretty direct hint 👀
I’m sol excited
