Claude Sonnet 4.5 vs GLM-5

Benchmarks, API pricing and specs, head to head. Data updated 2026-06-10.

Claude Sonnet 4.5

Anthropic · Sep 2025

65.5

The model that made 30-hour autonomous coding sessions real — 77.2% SWE-bench Verified and a 1M-token context beta at $3/$15.

GLM-5

Z.ai (Zhipu) · Feb 2026

Z.ai's 744B open-weights MoE — within ~4 points of Opus 4.6 on SWE-bench Verified at $1/$3.20 per million tokens.

The verdict

GLM-5 wins 1 of the 1 benchmarks these models share, against 0 for Claude Sonnet 4.5. On price the gap is dramatic: GLM-5 works out roughly 3.9x cheaper per blended million tokens. And GLM-5 is open-weights (MIT), so it can be self-hosted — a structural advantage if data control or vendor independence matters.

Benchmark head-to-head 01

SWE-bench Verified
77.2%~77.8%
Claude Sonnet 4.5GLM-5

Specs & pricing

Claude Sonnet 4.5GLM-5
modhub Index65.5
Input price / 1M$3$1
Output price / 1M$15$3.2
Context window200K200K
Max output64K128K
Open weightsnoyes (MIT)
Reasoning modelyesyes
Multimodal inputtext, imagetext
Knowledge cutoffJan 2025Dec 2025
ReleasedSep 2025Feb 2026
Example monthly cost*$52.50$14.80

* 10M input + 1.5M output tokens per month at list prices, no caching. Green = better value on that row.

Frequently asked questions

Which is better, Claude Sonnet 4.5 or GLM-5?
GLM-5 wins 1 of the 1 benchmarks these models share, against 0 for Claude Sonnet 4.5. On price the gap is dramatic: GLM-5 works out roughly 3.9x cheaper per blended million tokens. And GLM-5 is open-weights (MIT), so it can be self-hosted — a structural advantage if data control or vendor independence matters.
Which is cheaper, Claude Sonnet 4.5 or GLM-5?
Claude Sonnet 4.5 costs $3/$15 per million input/output tokens, while GLM-5 costs $1/$3.2. For a typical workload of 10M input and 1.5M output tokens per month, that's $52.50 versus $14.80.
Which model is better for coding, Claude Sonnet 4.5 or GLM-5?
On SWE-bench Verified — the standard agentic-coding benchmark — GLM-5 scores ~77.8% versus 77.2% for Claude Sonnet 4.5, making GLM-5 the stronger pick for coding agents.

More comparisons