GLM 4.7 (Venice): real-world results
GLM 4.7 (Venice) passed 12 of 15 runs (80%) on 15 different use cases.
Average cost per run was $0.03, average wall time 6m, with 0.27 human step-ins per run.
Results by category
| Category | Runs | Passed |
|---|---|---|
| Finance / Retirement Planning | 3 | 3/3 |
| DevOps / Configuration | 1 | 1/1 |
| Facilities / Workplace Analytics | 1 | 0/1 |
| Finance / Backtesting | 1 | 1/1 |
| Finance / Debt Payoff | 1 | 1/1 |
| Finance / Statement Analysis | 1 | 1/1 |
| Finance / Tax | 1 | 1/1 |
| Finance / Valuation | 1 | 1/1 |
| Marine / Fishing Compliance | 1 | 0/1 |
| Marine / Navigation | 1 | 1/1 |
| Marine / Powerboat Planning | 1 | 0/1 |
| Marketing / Budget Allocation | 1 | 1/1 |
| Scheduling / Calendar | 1 | 1/1 |
Head-to-head comparisons
- GLM 4.7 (Venice) vs Claude Haiku 5.5
- GLM 4.7 (Venice) vs Claude Opus 5.5
- GLM 4.7 (Venice) vs Claude Sonnet 5.5
- GLM 4.7 (Venice) vs DeepSeek V4.1 Flash (Venice)
- GLM 4.7 (Venice) vs Gemini 3.8 Flash
- GLM 4.7 (Venice) vs Gemini 4 Argon
- GLM 4.7 (Venice) vs Gemma 4 31B Instruct (Venice)
- GLM 4.7 (Venice) vs GPT-6.1 Sol
- GLM 4.7 (Venice) vs Kimi K3.1 Pro (Venice)
- GLM 4.7 (Venice) vs MiniMax M2.5 (Venice)
- GLM 4.7 (Venice) vs Mistral Small 3.2 24B Instruct (Venice)
- GLM 4.7 (Venice) vs Qwen3 Coder 480B Turbo (Venice)
Browse every run by GLM 4.7 (Venice) →
From published single-model runs only (all settings grouped). How we score.