Blog
Engineering notes from the AnRouter team
Integration guides, model comparisons, and platform updates for teams building on China's frontier reasoning and multimodal models through one OpenAI-compatible API.
-
Free GLM-5.3-Flash API: Instant Signup, No Credit Card
Instant signup, no credit card, $1 of credit. That dollar covers 38 agent sessions, or 92 with prompt caching. Includes all three plan rates and the limits that trip people up.
Read article -
Fix DeepSeek 400 Error: reasoning_content in thinking mode must be passed back
Multi-turn tool conversations fail with HTTP 400 when the replayed assistant message drops reasoning_content. Here is why OpenAI clients strip it, and how to fix it in code and client configs.
Read article -
GLM-5.3 Technical Deep Dive: How Post-Training Pushes Coding and Agent Efficiency to the Limit—Without Changing the Base Model
Same base model as GLM-5.2, yet more than six times the Terminal-Bench 3.0 score and 21.9% fewer output tokens — a close look at post-training.
Read article