Best API models for agents & automation
Start with a cheaper model and move up when it fails.
Terra is the low-price visual and computer-use option among mid-tier models. Haiku 5.5 is the budget browser-use option. Complex agents may need stronger models, supervision and actions you can undo.
Low-cost screen work, visual reasoning and drafting.
Benchmark $0.96/task · $1.00 in / $5.00 out per 1M
Classification, extraction, routing, summaries, subagents and browser use at very low cost.
Benchmark $0.41/task · $0.10 in / $0.50 out per 1M
Everyday coding, general work and tool-using agents.
Benchmark $1.01/task · $2.00 in / $10.00 out per 1M
Everyday coding, multimodal work and agents.
Benchmark $3.16/task · $1.50 in / $7.50 out per 1M
Difficult software, computer-use and document workflows.
Benchmark $4.00/task · $8.00 in / $40.00 out per 1M