📚 Resources

• Title: NVIDIA: LLM Inference Benchmarking — cost methodology: URL: https://developer.nvidia.com/blog/llm-inference-benchmarking-how-much-does-your-llm-inference-cost/
• Title: 2025–2026 GPU cloud pricing benchmarks: URL: https://introl.com/blog/cost-per-token-llm-inference-optimization
• Title: LLM Total Cost of Ownership 2025: Build vs Buy Math: URL: https://www.ptolemay.com/post/llm-total-cost-of-ownership
• Title: vLLM benchmark CLI: URL: https://docs.vllm.ai/en/latest/benchmarking/cli/
• Title: Beyond Per-Token Pricing: Concurrency-Aware LLM Cost Estimation (arXiv): URL: https://arxiv.org/pdf/2606.11690
• Title: How to Reduce LLM Inference Costs Without Sacrificing SLA — Paralleliq: URL: https://www.paralleliq.ai/blog/gpu-ops-reduce-inference-costs

Week 14 · Fractional GPUs, MIG, Quantization, and KV-Cache Savings

Notion ID: 392c08fd-0278-8185-ba1b-e5b4f5c4cc35 Notion URL: https://app.notion.com/p/392c08fd02788185ba1be5b4f5c4cc35