OpenAI had been very strict about blocking reverse engineering/Ghidra/IDA_Pro-MCP tasks. I even got a warning email. I was having much more success convincing Claude Code for those tasks without warnings. Seems like they've tightened things up.
Excited to use 1 prompt and have my whole 5-hour window at 100%. They can keep releasing new ones but if they don't solve their whole token shrinkage and gaslighting it is not gonna be interesting to se.
For anyone who was wondering about Mythos release plans:
> What we learn from the real-world deployment of these safeguards will help us work towards our eventual goal of a broad release of Mythos-class models.
Too late, personally after how bad 4.6 was the past week I was pushed to codex, which seems to mostly work at the same level from day to day. Just last night I was trying to get 4.6 to lookup how to do some simple tensor parallel work, and the agent used 0 web fetches and just hallucinated 17K very wrong tokens. Then the main agent decided to pretend to implement tp, and just copied the entire model to each node...
This coming right after a noticeable downgrade just makes me think Opus 4.7 is going to be the same Opus i was experiencing a few months ago rather than actual performance boost.
Anthropic need to build back some trust and communicate throtelling/reasoning caps more clearly.
> Opus 4.7 uses an updated tokenizer that improves how the model processes text. The tradeoff is that the same input can map to more tokens—roughly 1.0–1.35× depending on the content type.
caveman[0] is becoming more relevant by the day. I already enjoy reading its output more than vanilla so suits me well.
I just subscribed this month again because I wanted to have some fun with my projects.
Tried out opus 4.6 a bit and it is really really bad. Why do people say it's so good? It cannot come up with any half-decent vhdl. No matter the prompt. I'm very disappointed. I was told it's a good model
> where previous models interpreted instructions loosely or skipped parts entirely, Opus 4.7 takes the instructions literally. Users should re-tune their prompts and harnesses accordingly.
These stuck out as promising things to try. It looks like xhigh on 4.7 scores significantly higher on the internal coding benchmark (71% vs 54%, though unclear what that is exactly)
> More effort control: Opus 4.7 introduces a new xhigh (“extra high”) effort level between high and max, giving users finer control over the tradeoff between reasoning and latency on hard problems. In Claude Code, we’ve raised the default effort level to xhigh for all plans. When testing Opus 4.7 for coding and agentic use cases, we recommend starting with high or xhigh effort.
The new /ultrareview command looks like something I've been trying to invoke myself with looping, happy that it's free to test out.
> The new /ultrareview slash command produces a dedicated review session that reads through changes and flags bugs and design issues that a careful reviewer would catch. We’re giving Pro and Max Claude Code users three free ultrareviews to try it out.
Honestly I've been doing a lot of image-related work recently and the biggest thing here for me is the 3x higher resolution images which can be submitted. This is huge for anyone working with graphs, scientific photographs, etc. The accuracy on a simple automated photograph processing pipeline I recently implemented with Opus 4.6 was about 40% which I was surprised at (simple OCR and recognition of basic features). It'll be interesting to see if 4.7 does much better.
I wonder if general purpose multimodal LLMs are beginning to eat the lunch of specific computer vision models - they are certainly easier to use.
Sigh here we go again, model release day is always the worst day of the quarter for me. I always get a lovely anxiety attack and have to avoid all parts of the internet for a few days :/
331 comments
[ 3.8 ms ] story [ 135 ms ] threadThis decision is potentially fatal. You need symmetric capability to research and prevent attacks in the first place.
The opposite approach is 'merely' fraught.
They're in a bit of a bind here.
> What we learn from the real-world deployment of these safeguards will help us work towards our eventual goal of a broad release of Mythos-class models.
The surprise: agentic search is significantly weaker somehow hmm...
You are in for a treat this time: It is the same price as the last one [0] (if you are using the API.)
But it is slightly less capable than the other slot machine named 'Mythos' the one which everyone wants to play around with. [1]
[0] https://claude.com/pricing#api
[1] https://www.anthropic.com/news/claude-opus-4-7
The surprise: agentic search is significantly weaker somehow hmm...
This coming right after a noticeable downgrade just makes me think Opus 4.7 is going to be the same Opus i was experiencing a few months ago rather than actual performance boost.
Anthropic need to build back some trust and communicate throtelling/reasoning caps more clearly.
caveman[0] is becoming more relevant by the day. I already enjoy reading its output more than vanilla so suits me well.
[0] https://github.com/JuliusBrussee/caveman/tree/main
https://github.com/rtk-ai/rtk
Tried out opus 4.6 a bit and it is really really bad. Why do people say it's so good? It cannot come up with any half-decent vhdl. No matter the prompt. I'm very disappointed. I was told it's a good model
interesting
Or `/model claude-opus-4-7` from an existing session
edit: `/model claude-opus-4-7[1m]` to select the 1m context window version
> More effort control: Opus 4.7 introduces a new xhigh (“extra high”) effort level between high and max, giving users finer control over the tradeoff between reasoning and latency on hard problems. In Claude Code, we’ve raised the default effort level to xhigh for all plans. When testing Opus 4.7 for coding and agentic use cases, we recommend starting with high or xhigh effort.
The new /ultrareview command looks like something I've been trying to invoke myself with looping, happy that it's free to test out.
> The new /ultrareview slash command produces a dedicated review session that reads through changes and flags bugs and design issues that a careful reviewer would catch. We’re giving Pro and Max Claude Code users three free ultrareviews to try it out.
I wonder if general purpose multimodal LLMs are beginning to eat the lunch of specific computer vision models - they are certainly easier to use.