llama-3.1-8b-instant

Latency & availability — single-word reply

Basic math reasoning — show work, give answer

Structured output compliance — valid JSON with required keys

Python function generation with docstring

Token generation speed — 500-token long-form response

Retrieval from in-context data — 20-item list Q&A

Function/tool calling — get_weather invocation

Real-world: distill a news article into 3 bullet points

Real-world: sentiment + category from customer review

Real-world: pull structured fields from an invoice

Real-world: multi-rule compliance (5 sentences, no "the", etc.)