Blueprint Blog https://llmblueprint.ai/blog Technical deep-dives on LLM optimization from Inspire AI Lab. Fri, 26 Jun 2026 00:00:00 GMT https://validator.w3.org/feed/docs/rss2.html https://github.com/jpmonette/feed en © 2026 Inspire AI Lab LLC <![CDATA[VRAM math for any open LLM, from 7B to 405B]]> https://llmblueprint.ai/blog/vram-math https://llmblueprint.ai/blog/vram-math Fri, 26 Jun 2026 00:00:00 GMT Inspire AI Lab self-host <![CDATA[Self-hosted vs API: the actual math at 1M / 10M / 100M tokens/month]]> https://llmblueprint.ai/blog/selfhost-vs-api-math https://llmblueprint.ai/blog/selfhost-vs-api-math Fri, 26 Jun 2026 00:00:00 GMT Inspire AI Lab cost <![CDATA[RAG vs fine-tuning vs both: a decision tree]]> https://llmblueprint.ai/blog/rag-vs-fine-tuning https://llmblueprint.ai/blog/rag-vs-fine-tuning Fri, 26 Jun 2026 00:00:00 GMT Inspire AI Lab architecture <![CDATA[QLoRA on a consumer GPU: fitting a 13B fine-tune in 24GB]]> https://llmblueprint.ai/blog/qlora-on-consumer-gpu https://llmblueprint.ai/blog/qlora-on-consumer-gpu Fri, 26 Jun 2026 00:00:00 GMT Inspire AI Lab adaptation <![CDATA[Prompt cache in production: what hits, what misses, and how to design around it]]> https://llmblueprint.ai/blog/prompt-cache-production https://llmblueprint.ai/blog/prompt-cache-production Fri, 26 Jun 2026 00:00:00 GMT Inspire AI Lab runtime <![CDATA[Picking the right open model for your workload]]> https://llmblueprint.ai/blog/picking-the-right-model https://llmblueprint.ai/blog/picking-the-right-model Fri, 26 Jun 2026 00:00:00 GMT Inspire AI Lab self-host <![CDATA[LoRA from first principles: rank, alpha, and why low-rank works]]> https://llmblueprint.ai/blog/lora-from-first-principles https://llmblueprint.ai/blog/lora-from-first-principles Fri, 26 Jun 2026 00:00:00 GMT Inspire AI Lab adaptation <![CDATA[When llama.cpp beats vLLM (and vice versa)]]> https://llmblueprint.ai/blog/llamacpp-vs-vllm https://llmblueprint.ai/blog/llamacpp-vs-vllm Fri, 26 Jun 2026 00:00:00 GMT Inspire AI Lab runtime <![CDATA[What imatrix actually does (and why off-the-shelf Q4 leaves accuracy on the table)]]> https://llmblueprint.ai/blog/imatrix-explained https://llmblueprint.ai/blog/imatrix-explained Fri, 26 Jun 2026 00:00:00 GMT Inspire AI Lab calibration <![CDATA[Building an evaluation harness that survives prod]]> https://llmblueprint.ai/blog/evaluation-harness https://llmblueprint.ai/blog/evaluation-harness Fri, 26 Jun 2026 00:00:00 GMT Inspire AI Lab architecture <![CDATA[Building a calibration corpus from your client's prompts]]> https://llmblueprint.ai/blog/calibration-corpus https://llmblueprint.ai/blog/calibration-corpus Fri, 26 Jun 2026 00:00:00 GMT Inspire AI Lab calibration <![CDATA[Break-even analysis: when on-prem GPUs beat reserved cloud]]> https://llmblueprint.ai/blog/break-even-on-prem-vs-cloud https://llmblueprint.ai/blog/break-even-on-prem-vs-cloud Fri, 26 Jun 2026 00:00:00 GMT Inspire AI Lab cost