Google spent an estimated $72,000,000,000 getting Gemini 3 out the door.
A lab in Beijing spent less than that company spends on snacks, shipped Kimi K3, beat it on cost per task, and then did the one thing Mountain View will not do: they put the whole model on a public link.
4. The AI agent swarm is where things get really interesting.
Multiple agents can work together using models like Opus 4.8, Fable 5, and GLM 5 Sol to build full-stack SaaS products including an entire CRM.
Four models drew their own self-portraits.
GLM-5.3-Flash, GLM-5.3, Kimi K3 and Claude Opus 5. GLM-5.3-Flash somehow came out looking the most handsome.
AI benchmarks are getting WILD.
🤯 GPT-5.6 Sol: 91.9% on Terminal-Bench Claude Opus 5: still at the frontier GLM-5.3-Flash: insanely cheap + punching WAY above its weight MiniMax M3: 114 tokens/sec ⚡ So instead of posting another boring benchmark table... I animated the
You have reached the end of the archive
All of Claude Opus 5