claude fable 5 is on par with gpt-5.5 on DeepSWE
fable 5: 70% success rate at $10.3/task
gpt-5.5: 70% success rate at $6.6/task
opus 4.8: 58% success rate at $12.6/task
@datacurve hasn't officially released the mythos 5 scores yet, but @theo somehow shared them on his youtube channel
The energy around Codex in Europe right now is incredible!
I met founders, enterprises, and developers adopting Codex in all kinds of ways.
The London meetup was one of my favorite moments: nearly 200 builders joined last minute to hang out and build. Such a special community!
Hey, @bcherny and @claudeai - I am sure not a first reporter - but still have this floating bug on the latest version (CLI Claude Code v2.1.150).
Hope it can help - efe4670d-4838-452a-9547-d894fb1b38ee
@thsottiaux@dkundel 1. Default model for subagents in config.toml not to use in every agent.toml as hardcoded
2. Auto mode where tools calling with cheap models and expensive use only for real matter - like an orchestrator