What every Ollama timing field means and the arithmetic that turns it into a claim - tokens per second, time-to-first-token, why total_duration is not latency, why a second run is faster, what changes with context length, quantization, GPU offload and parallel requests - plus the reporting rules that make a benchmark a measurement instead of an anecdote.
Install
npx skills add vecnode/vn-code --skill ollama-inference-metricsnpx skills add vecnode/vn-code --skill ollama-inference-metrics -a claude-code -gnpx skills add vecnode/vn-code --skill ollama-inference-metrics -a codex -ggemini skills install https://github.com/vecnode/vn-code.git --path packages/dsh-ollama/skills/ollama-inference-metricsgh skill install vecnode/vn-code ollama-inference-metricsgit clone --depth 1 https://github.com/vecnode/vn-code /tmp/vecnode__vn-code
mkdir -p ~/.claude/skills ~/.agents/skills
cp -R /tmp/vecnode__vn-code/packages/dsh-ollama/skills/ollama-inference-metrics ~/.claude/skills/ollama-inference-metrics # Claude Code
cp -R /tmp/vecnode__vn-code/packages/dsh-ollama/skills/ollama-inference-metrics ~/.agents/skills/ollama-inference-metrics # Codex, Cursor, Gemini CLI, Copilot, OpenCode