agent_system

deepseek-v4-pro mini_swe_agent_deepseek_v4_pro_max

config_f64966d7e89784334403b1fe7878048a57ec00876ce000c6e2bef57e5f49e129

Published ranking

This score and claim apply only inside the bound ranking group.

63%

capability score

Rank

31

Evidence families

1

Claim

Outside top set

Uncertainty was not reported

Configuration passport

Canonical name
deepseek-v4-pro
Revision
mini_swe_agent_deepseek_v4_pro_max
Interaction policy
agentic
Passport class
agent-system-v1
Harness
mini-swe-agent
Scaffold
mini_swe_agent_deepseek_v4_pro_max

Source evidence

Publication snapshotexplorer_58c2cf3a940443b4e1fdfd119dc7ce463191fe444e5acaf18aeac03b2b5cb982

Methodology2026-10-31.1.cost-per-task-wave