GLM 5.3 may now be the best open source model

Before my download of Qwen 3.8 Max has even completed, we now have a new contender for best open source model.

GLM 5.3 is faster and smaller than Kimi K3 and Qwen 3.8 Max, and it's showing Fable level results on some benchmarks.

They achieved the gains in performance in much the same way as Deepseek improved their V4 Flash and Pro models: extensively improved post-training.

Rates are:

Lite Tier: $18/month (drops to $12.60/month on annual billing); includes 10,000 credits per week. Pro Tier: $80/month (drops to $56/month on annual billing); provides roughly 6x the usage volume of Lite. Max Tier: $168/month (drops to $117.60/month on annual billing); provides roughly 14x the usage volume of Lite Off-Peak Discount: Calls made outside weekday peak hours consume 50% fewer points. Peak Hours are 14:00 to 18:00 UTC+8, Monday through Friday