GLM-Z1-32B
32B reasoning model matching R1-class performance per Zhipu
About
GLM-Z1-32B is Zhipu AI's reasoning model built on GLM-4-32B-0414, trained with cold-start data plus extended reinforcement learning focused on math, code, and logic. Zhipu claims the 32B model matches DeepSeek-R1's 671B on tasks like AIME and LiveCodeBench — a vendor claim worth treating with the usual caution. A 9B sibling targets resource-constrained deployment. The weights are Apache 2.0, which is genuinely permissive. English-language evaluation coverage and community tooling are thinner than for the Qwen or DeepSeek ecosystems, and independent benchmark reproductions are harder to find.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
Zhipu-reported: matches DeepSeek-R1 (671B) on AIME 24/25, LiveCodeBench, GPQA in some tasks
Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.
Use cases
Trust score — 6/10
Scored 6/10 (metrics tier, metrics-v1). Evidence: Zhipu/Z.ai announcement April 2025: GLM-Z1-32B-0414 built on GLM-4-32B-0414, cold-start + extended RL; 9B sibling exists.; Apache 2.0 weights per third-party model table.; Vendor claim (matches R1-671B on some tasks) flagged as vendor-reported; exact day not pinned (releasedAt null).; Limitation: thin English evaluation coverage and community tooling; metrics tier (press-release sources)..