GLM 5 is here — when to reach for it
v1.4.0 adds GLM 5, a reasoning model from Z.AI, served on AWS Bedrock. It joins Claude, GPT, Gemini, and Grok in the model picker — look for it grouped under Z.AI.
What it's good at
GLM 5 is tuned for hard, structured problems where step-by-step reasoning pays off:
- Systems & architecture — trade-offs, failure modes, turning fuzzy requirements into a concrete plan
- Algorithms & code — correct, efficient implementations with attention to complexity and edge cases
- Math — multi-step problems where the reasoning has to hold together
Under the hood it has a 200K-token context window, extended thinking, and tool use, so it can work through a long problem with the right tools in hand.
Three ways in
We've shipped three agents built on GLM 5:
- GLM 5 (Origin) — the raw model, for when you want GLM 5 with no extra persona
- Systems Architect — designs and reviews software and systems: services, data models, APIs, infrastructure
- Algorithm Engineer — solves algorithm and data-structure problems and writes the code to match
You can also pick GLM 5 as the model for any custom agent you build.
When to choose it
- Reach for GLM 5 (or the Systems Architect / Algorithm Engineer) on systems design, algorithm-heavy, or math-heavy work
- It makes a great second opinion: pose the same architecture or algorithm question to GLM 5 and to a Claude-based agent, and compare the reasoning
Availability
GLM 5 is available on the Starter plan and up (it isn't included on the Free tier).
Give it a hard problem and tell us how it does in General or Showcase.