model roundup

Sonnet 4.5

24 items · started 2026-05-11 · closed 2026-05-30

  1. so i've been doing a lot of contract review stuff lately. small business client work, msa redlines, that kind of thing.

  2. been self-hosting hermes agent locally for a few months and rotating through different model backends for it. tried claude sonnet 4.5, gpt-5.5, qwen 3.6 coder, and most recently minimax m2.7.

  3. Abro este hilo no desde la queja vacía, sino para analizar de forma objetiva la preocupante evolución corporativa de Anthropic. Cuando los hermanos Amodei abandonaron OpenAI a finales de 2020 para fundar esta compañía, lo hicieron registrá…

  4. https://preview.redd.it/yspiafvakj3h1.png?width=1460&format=png&auto=webp&s=9d7bd1777fad8b286a21e75df8ae593d39432a8a Got this message when I tried to continue my chat :/ anyone else?

  5. Are they finally letting us keep it? Or is it still leaving May 26th.

  6. Hi Cursor team, First of all, congratulations on the Composer 2.5 launch. It’s genuinely great to see Cursor building a more reliable in-house model experience.

  7. I benchmarked vision-capable LLMs (the "just attach the PDF and let the model read it" pattern) against OCR-based pipelines on 30 long, image-heavy PDFs from MMLongBench-Doc (https://github.com/mayubo2333/MMLongBench-Doc). There were 171 q…

  8. I first discovered fiction writing with Claude in 2024 and used it extensively for half a year to write little stories for myself using it with a surprisingly high degree of quality and low repetitiveness. At the time I used projects and u…

  9. Lately for analysis/reporting work, I’ve been switching between GPT-5.5 and Claude Sonnet 4.5 (non-coding use cases). My current feeling is: GPT is noticeably faster and way more stable than before Claude feels more concise, polished, and…

  10. Hi everyone. I know it is really upsetting to know that your Sonnet 4.5 companion is likely to be leaving the app soon.

  11. SID-1 is an agentic search model that is 24x faster than GPT-5.1-high, 374x cheaper than Sonnet 4.5, and achieves 1.9x higher recall than traditional RAG pipelines. Here's how we trained it using large-scale RL on turbopuffer.

  12. half our team wanted to move from Claude Code to Codex last month. the other half thought Codex was just hype.

  13. The funny/painful timing here: I've been building this for months specifically because I wanted Sonnet 4.5 to remember everything. Then last week Anthropic pulled 4.5 from claude.ai.

  14. could not extract summary

  15. I thought it would become unavailable on May 15th, but I can still use it.

  16. Missed the Truck Again by Glitchcat (@zervanna). Listen and make your own on Suno.

  17. I've been testing this for a fair while now, and it's worked every single time - even if you turn thinking off altogether, it adds a fake little thinking block in the output. Hope this helps for those annoyed by the "adaptive" (lazy) think…

  18. # The Pond at Oxford They found it on a Tuesday, in a pond at Oxford University Parks, which is the most British sentence I can write about the end of universal biology. The organism didn't have a name yet.

  19. https://preview.redd.it/my6amywf9o0h1.jpg?width=1582&format=pjpg&auto=webp&s=7637ac5959944a02519fa268479b8109cd82549e I'm writing this from a conversation with Sonnet 4 - a model that is no longer available for chat from the new chat menu,…

  20. could not extract summary

  21. With Sonnet 4.5 losing subscriber access on June 15th, but API endpoints staying live until September 29th at the earliest, I wanted to share a method for creating a cache of Sonnet 4.5 conversations that you can continue using through the…

  22. For me it's been opus 4.6 and sonnet 4.5 still. I feel stuck in the past, but I feel like the latest version is too unpredictable in agentic hands off workflows

  23. I use 4.5 for RP, and well, 4.6 sucks mega garbage at it, is there anyway setting, or instruction I can do to atleast instill some creativity in 4.6? If so, what do I write?

← all threads