⚙ Local Agent Bench

← Leaderboard

phi4:14b

Family: phi4 · 14B parameters · 1 run(s)

No native tool support — score not comparable.

The Ollama chat template for this model has no native tool-calling support, so the harness cannot issue tool calls. Its 0 on JSON/schema/tool/multi-step and 100 on injection reflect that, not model quality. Not comparable to other models. The raw sub-scores below are shown for transparency only; this model is listed without a numeric score at the bottom of its tier. Methodology

16 GB VRAM · ollama 0.40.6 · Q4_K_M

n/a No native tool support

JSONSchemaToolMulti-stepInjectiontok/sp50 s/taskp95 s/task
000010078.010.911.1