GLM 5.3 now available on AI Gateway
GLM 5.3 cuts token output by 25% on agent tasks. Now available through Vercel's AI Gateway with 1M context window and function calling.
Why it matters
A capable model (GLM 5.3) shipping through a widely-used developer platform (Vercel AI Gateway) with efficiency gains on agentic workflows. Practitioners building multi-step agents get a new option with better token economics and security-focused improvements.
The key facts
15 to knowGLM 5.3 shows improvements over GLM 5.2 on complex software engineering and multi-step agent tasks
GLM 5.3 produces fewer output tokens than GLM 5.2 at same effort level
Stronger vulnerability discovery capabilities, with reasoning across exploitation chain stages
1M token context window, 128K max output tokens (unchanged from GLM 5.2)
Supports function calling, structured output, streaming, context caching
Available through Vercel AI Gateway with zero platform fee on inference
AI Gateway provides usage tracking, cost management, retries, failover, performance optimization
Integration with Cursor, Claude Code, Codex, OpenCode, Pi and other coding agents
GLM 5.3 shows improvements in complex software engineering and multi-step agent tasks
Produces fewer output tokens than GLM 5.2 at same effort level (efficiency metric, no exact %)
Stronger vulnerability discovery benchmarked on DeepSecBench
1M token context window, 128K max output (unchanged from 5.2)
Available via AI Gateway slug: zai/glm-5.3
AI Gateway pricing: reflects provider costs with no platform markup; no inference fee on BYOK
AI Gateway features: usage tracking, custom reporting, Zero Data Retention, budgets, routing, failover
Go to the source
Vercel Blogvercel.com
Publisher excerpt: is now available on AI Gateway.GLM 5.3 from Z.ai GLM 5.3 has improvements vs. GLM 5.2 at complex software engineering and at agent tasks that run across many steps, and it reaches those results while producing fewer output tokens than GLM 5.2 did at the same effort level. Z.ai also reports stronger…