Which AI model for agents and automation?
Agents and workflows that act on your tools: sorting emails, summarizing reviews, filing documents.
DeepSeek V4.1 Flash
€0.27 / €1.07 per million tokens (input / output)
On the Pareto line, capability score of 155 at a low price: an agent rereads a lot of text at each step.
Cost for 1 million tasks of about 8,000 tokens read and 1,200 written over several steps: €3,419.
See in the ranking →Qwen3.8 27B
Free to use: only your hardware cost (a 24 GB GPU is enough with Q4 quantization).
Open weights under the Apache 2.0 license: plugged into a local n8n, no data goes to an external API.
See in the ranking →Claude Opus 5.5
€3.56 / €17.82 per million tokens (input / output)
The best capability score in the index: for agents that chain many steps, every error propagates.
Cost for 1 million tasks of about 8,000 tokens read and 1,200 written over several steps: €49,884.
See in the ranking →Why these choices
For an agent, a mistake at one step skews all the following ones: reliability matters more than for a single answer. Keep a human approval before any irreversible action (sending, deleting, paying).
Take action
Data: Epoch AI index (CC BY 4.0), developers' official prices converted at the ECB rate, updated 2026-10-03 · AI model ranking: intelligence, benchmarks and price