GLM 4
#113 in Frontier-Spraakmodellezhipu · v4 · siet 2024-01-16 · 5× · tolest 30. Juni 2026
11
Momentum
GLM-4 Plus is the commercial flagship language model of the fourth generation from Zhipu AI (bigmodel.cn), available via API since January 2024. It was fully developed in-house and trained on approximately ten trillion tokens primarily in Chinese and English (plus 24 additional languages). Zhipu claims a 60% performance improvement over GLM-3 and an IFEval instruction-following score of 90%. In the 128K needle-in-a-haystack test, the vendor reports 100% accurate recall.
Momentum-Verloop
04.04.03.07.
Features
| Benchmark Performance (Coding/Reasoning Score) | MMLU: 96.3% of GPT-4 (0314) level; GSM8K, MATH, BBH, GPQA, HumanEval: comparable to GPT-4 Turbo and Claude 3 Opus per arXiv tech report; IFEval (Instruction Following): 90% |
| Context Window (Tokens) | 128,000 tokens (128K); 100% recall in needle-in-a-haystack test per vendor |
| Multilingualism / Multimodality | Primarily Chinese and English; pretraining on corpus in 24 languages; text-only (no native vision model). Separate multimodal variant GLM-4V available. |
| Price Tier/Monthly Plan | $0.20 / 1M input tokens, $0.80 / 1M output tokens (API, pay-per-use); reduced by up to 90% following industry-wide price cuts in China in late 2024/early 2025 |