함께하는 AI 퀴즈

퀴즈를 풀면 정답과 설명이 나타납니다.

1. How should Together AI’s hosted model catalog be described?
   • It includes open-weight models, with access and terms depending on the specific model
   • Every model is open source under one license
   • Every checkpoint has identical weights and terms
   • The service only hosts models built by Together
2. What is RedPajama in Together AI’s project history?
   • RedPajama is the name of Together’s OpenAI-compatible chat endpoint
   • RedPajama is Together’s open-data project; its Data-v2 release is a dataset, separate from hosted inference models.
   • RedPajama is the hardware interconnect used in GPU clusters
   • RedPajama is a privacy mode that deletes account credentials
3. What does “OpenAI-compatible” mean for a Together API integration?
   • Every OpenAI endpoint and parameter works identically
   • No API key or base URL is needed
   • Supported endpoints accept compatible request shapes, but model IDs and endpoint features still need checking
   • A model swap guarantees identical outputs
4. Which inference technique does Together discuss for high-throughput long-context inference?
   • Quantization and continuous batching in every deployment
   • A fixed text template
   • A beam search that does not use model predictions
   • 추측적 디코딩
5. What should an enterprise verify before sending prompts or files to a hosted model service?
   • The specific service’s data-retention, training, and deployment settings and terms
   • That open model weights automatically make all data private
   • That an OpenAI-compatible API guarantees identical data handling
   • That model customization removes the need to review privacy settings
6. 서버리스 API 외에 Together는 대규모 교육 작업을 위해 어떤 인프라를 제공합니까?
   • 무료 무제한 양자 컴퓨터
   • 빠른 상호 연결을 갖춘 임대 GPU 클러스터
   • 모델 중량용 프린터
   • 단일 노트북만 가능
7. What does Together AI describe its Zero Data Retention setting as doing?
   • Deleting all account and billing data immediately
   • Guaranteeing that every third-party model has identical data policies
   • Disabling retention and secondary use of submitted content and outputs, except as needed to process the request
   • Turning off logging for all infrastructure operations
8. What did Together’s RedPajama-Data-v2 announcement report about the dataset?
   • A 30-billion-parameter language model trained only on code
   • A benchmark of GPU cluster latency
   • A dataset containing only generated audio
   • A 30-trillion-token filtered and deduplicated web dataset with quality annotations