I second that. Gemini 3.5 Flash rocks the benchmark charts but is terrible as an agent. Horrible instruction adherence and makes WAY too many tool calls

▲

luckydata 3 hours ago | parent [-]

which cheap models have you found work best as agents?

	▲	verdverm an hour ago \| parent [-]
		Most of the bigger open weight models are pretty good. You can get them per-token from companies like Fireworks or OpenCode