Benchmark

MCP Atlas

Public and full splits are both reported under one name; the tool server set is part of the measurement.

Released
2025-11-18
Categories
tool_use
Openness
unknown
Source
model_reports
Reported scores
8

Reported scores

model_reports

ModelOrganizationReported valueReportedEvidence
DeepSeek-V4-FlashDeepSeek69.02026-06-25evidence
DeepSeek-V4-ProDeepSeek73.62026-04-22evidence
Gemini 3.1 ProGoogle69.22026-02-19evidence
GLM-5Z.ai67.82026-02-11evidence
GLM-5.1Z.ai71.82026-04-03evidence
GLM-5.2Z.ai76.82026-06-16evidence
Hy4 previewTencent83.72026-08-28evidence
Kimi K3Moonshot AI84.22026-06-13evidence

Scores are partitioned by the source that reported them and are never merged into a single cross-source ranking, because the sources measure different things and say so.