The dataset is currently empty. Upload or create new data files. Then, you will be able to explore them in the Dataset Viewer.
Lokale LLM Tests (Mac 32GB)
Dieses Repository dokumentiert meine Tests mit lokalen LLMs (Qwen, Mistral, GPT-OSS, etc.) auf Apple Silicon (32GB RAM).
Ziel ist kein Benchmark-Rekord, sondern praxisnahe Erfahrungen. Kein LLM ist schlecht, nur vielleicht nicht das Richtige für den Einsatzzweck.
- Performance
- Einsatzzweck
- Sprache
- Erfassen von komplexen Themen
- Strukturelle Aufgaben (JSON, YAML)
Hardware
- Mac (Apple Silicon)
- 32GB RAM
- Provider: LM Studio
Struktur
Jeder Test enthält:
- Modell
- Quant
- Kontextgröße
- Settings
- Evaluation Batch Size
- Tokens/s (subjektiv)
- Eindruck
- Fazit
Wichtiger Hinweis
Alle Eindrücke sind subjektiv und abhängig von:
- Use Case (Chat / Creative / YAML / Summaries)
- Promptstil
- Template
- Kontextlänge
Keine wissenschaftlichen Benchmarks – nur reale Nutzung.
- Downloads last month
- 12