Zhipu AI
GLM-5
Zhipu AI's most capable model with the best tool-call reliability we've tested. Excellent for complex agentic tasks requiring precise tool use.
$1/M input
$3.2/M output
Context Window
200K tokens
Best For
Best tool-call reliability
SWE-Bench
64.1%
HumanEval
91.5%
Capabilities
Yes — Tool Calls
Yes — Streaming
Yes — Extended Thinking
Yes — Vision
Benchmark Scores
SWE-Bench64.1
HumanEval91.5
Tool Calling96
Speed75
Cost Efficiency80
Getting Started
- 1Get an API key from Zhipu AI's platform
- 2Install the Ava extension from the VS Code Marketplace or use the CLI
- 3Select GLM-5 as your model in the settings and paste your API key
- 4Start coding! Ava handles the rest.
Research Papers
We celebrate the labs whose models power Ava. Read the science behind GLM-5.