GLM 5.1 AI Evaluation | LayerLens - Benchmark Results & Performance

Model

GLM 5.1 by z-ai

text->text

Free credits included. No credit card required.

GLM-5.1 delivers a major leap in coding capability, with particularly significant gains in handling long-horizon tasks. Unlike previous models built around minute-level interactions, GLM-5.1 can work independently and continuously on complex engineering problems. The model excels at agentic workflows and autonomous problem-solving with enhanced reasoning capabilities.

Evaluations

across 7 benchmarks

Median latency

22.5s

Context length

203k

tokens

Cost

$1 · $3.20

input · output per 1M tokens

Key takeaways

GLM-5.1 is a text-to-text MoE model designed for advanced coding, autonomous problem-solving, and agentic workflows, excelling in complex engineering problems.
It demonstrates exceptional mathematical problem-solving (94.74% accuracy on AIME 2026) and performs well in rigorous scientific calculation across various physics and chemistry domains.
A key limitation is high latency (all prompts >10 seconds), making it unsuitable for time-sensitive applications. It also struggles with complex combinatorial problems and nuanced biological interpretations.
Its strengths include algebra, geometry, number theory, astrophysics, electromagnetism, and organic chemistry synthesis. Weaknesses include molecular biology, genetics, and complex inorganic reactions.
The model achieved 94.67% accuracy on a mathematical problem-solving benchmark, showing robust performance, but with issues in strict output formatting and specific complex spatial geometry problems.

Technical specifications

Context length: 203k tokens
Max output: —
Architecture: MoE
Modalities: text->text
Parameters: ~754B
Open weights: Yes
Licence: MIT
Regions: China
Provider: z-ai
Provider model key: z-ai/glm-5.1