papersail

6 ms·

1. 1.▲
   by papersail 3mo ago

   rank score age size name 1 62.0 8 - Claude Fable 5 (Adaptive Reasoning, Max Effort, Opus 4.8 Fallback) 2 59.1 55 - GPT-5.5 (xhigh) 3 58.5 55 - GPT-5.5 (high) 4 57.2 104 -
2. 2.▲
   by papersail 3mo ago

   score age size name 62.0 8 - Claude Fable 5 (Adaptive Reasoning, Max Effort, Opus 4.8 Fallback) 59.1 55 - GPT-5.5 (xhigh) 58.5 55 - GPT-5.5 (high) 57.2 104 - GPT-5.4 (xhigh) 56.7 20
3. 3.▲
   by papersail 3mo ago

   I had similar doubts. I think expectations differ because the workload differs. For small scripts, glue code, or simple CRUD changes, smaller models such as Qwen3.6-27B can work wonders than they do on a larger, messier code base.
4. 4.▲
   by papersail 3mo ago

   I'm not sure I would put too much weight on DeepSWE as a benchmark, given that GPT-5.4-mini ended up close to Opus 4.6 there.
5. 5.▲
   by papersail 3mo ago

   My usual workflow is GPT-5.5 for planning, DeepSeek V4 Flash for milestones implementation, then GPT-5.5 again for review. It has worked pretty well so far.
6. 6.▲
   32GB of DDR5 now costs $375 – AI shortage continues to squeeze PC building(tomshardware.com)

   436 points by papersail 4mo ago|393 comments
7. 7.▲
   by papersail 4mo ago

   I used to collect old hardware and try to install NetBSD/OpenBSD on them. For me, a lot of the fun was the "challenge accepted" aspect. Getting them to boot and work on some weird and obsolete machines felt like winning a sma