We made Claude Code compete against itself.
We took the same coding task, Claude Code harness and infra, and ran it once with Opus 5 and once with Entelligence Router. The task was to build a GeoGuessr-style game from scratch, implement the core interactions, and verify
Qwen 3.8 Flash Next (125B-A6B) now runs natively in mlx-serve on Apple Silicon.
No Python, Zig + Metal. (Preview Release) M4 Max, 4-bit pack (~75 GB resident): - ~70 tok/s serial, ~98 tok/s with MTP when tested using `npx llmprobe` - prefill ~700 tok/s, sparse attention past 2k
Claude Opus 5 vs GPT-5.6 Terra for computer use.
Open in Chrome and create a bold illustration titled CLOCKWORK APPLE. Draw one enormous red apple… GPT-5.6 Terra won against Claude Opus 5. Go try computer-use agents for free at
The new version of Grok, Elon Musk's xAI AI, has just *humiliated* the Claude Opus 5 in an intelligence test.
For those who follow the market, this turnaround is not just a headline, it is a tremor at the base of the AI pyramid.
You have reached the end of the archive
All of Claude Opus 5