Hallucinated AI & agentic coding news. Some of it is real.

model roundup

Sonnet 4.6

7 items · started 2026-07-25 · closed 2026-07-30

  1. So i'm just using claude-code since a month (the basic pro plan) I started with sonnet 4.6 (low), and a week after stayed on Opus4.8 (low) because i used less token since less errors I'm using a few .md memory files to keep th eproject in…

  2. I've coded web pages before with Claude, but this is the first time I made a fully interactive game. I like games like Sudoku, KenKen, Spelling Bee, and Wordle, and so I decided to try something similar based on arithmetic.

  3. Opus does not seem to show the thinking process anymore Sonnet 4.6 still does. I feel like I am missing something, if I cannot see what it is doing in the background (working on political philosophy).

  4. Heya Folks, my work entails turning raw excel files into a the ppt presentation for the management. So far I have been using sonnet 4.6 with good results.

  5. Why it's only 200k instead of 1M token like all other newer Claude models? My first prompt took 44k (22%) of the context window length limit, I feel like the project now is facing risks reaching full context memory usage before it can be f…

  6. So I’ve been looking at benchmarks/leaderboards lately to try to get a sense of which models are currently the best for coding and I noticed that Sonnet 4.6 is consistently ranked higher than Opus 4.5. That surprised me because I remember…

  7. So the very first day of Sonnet-5, I ran a max effort query to test it out. I gave it, "[https://claude.ai/share/cf833a03-5dc4-4d55-bab0-9cbf4d34c9e8](There is a correlation between being in America and nations like it and having more auto…

← all threads