Z.ai GLM-5
Strong Chinese-origin model — 72.3% on DualEntry, aggressive price point.
Accounting overall
72.3%
Input / Output
$0.60 / $1.80 per MTok
Context
200K
Speed
~110 tok/s
Released
2026-02
Cutoff
2025-10
Eight accounting-task categories borrowed from DualEntry's 101-task benchmark. Measured where published, synthesized from adjacent benchmarks otherwise.
GLM-5 is the top-performing Chinese-origin model on DualEntry's accounting benchmark at 72.3%, and it's open-weights — meaningful for buyers who want to self-host agentic workflows for data-residency or cost reasons. The pricing is aggressive (~$0.60 / $1.80 per MTok on hosted endpoints), and the quality is within striking distance of GPT-5.4-Mini.
Caveats for Western buyers: the knowledge cutoff is slightly older (October 2025), which shows up in questions about recent accounting standards updates. The 200K context window is meaningfully smaller than flagship Western models' 1M. And for enterprise buyers with China data-sovereignty concerns, hosted GLM-5 requires additional diligence — though the open-weights release sidesteps that.
Citations
- DualEntry benchmark (GLM-5 72.3%)dualentry.com/blog/claude-opus-4-7-accounting-ai-benchmark-results