Agentic benchmarks
MCP-Atlas and Tool-Decathlon.
About this data
On agentic tool-use benchmarks GLM-5.2 is near the frontier on MCP-Atlas (76.8 vs Opus 4.8’s 77.8) but trails more on the harder Tool-Decathlon, where Opus 4.8 and GPT-5.5 lead. It still improves clearly over GLM-5.1 on both.
Score (higher = better). Tool use and multi-step agent tasks.
View data & sources →Data table
| glm51 | glm52 | gpt55 | gemini | opus48 | series | benchmark | source_ref | value_basis |
|---|---|---|---|---|---|---|---|---|
| 71.8 | 76.8 | 75.3 | 69.2 | 77.8 | agentic | MCP-Atlas | zai-glm52 | Full Benchmark Table — MCP-Atlas (public set) |
| 40.7 | 48.2 | 55.6 | 48.8 | 59.9 | agentic | Tool-Decathlon | zai-glm52 | Full Benchmark Table — Tool-Decathlon |
Methodology & sources
Last updated: Jul 17, 2026Methodology
Source-backed values for all six charts come from the Z.ai GLM-5.2 launch post (2026-06-16): long-horizon coding (FrontierSWE, PostTrainBench, SWE-Marathon), standard coding (Terminal-Bench 2.1, SWE-bench Pro, NL2Repo, DeepSWE, ProgramBench), reasoning (HLE, CritPt, AIME, HMMT, GPQA-Diamond, IMOAnswerBench), agentic (MCP-Atlas, Tool-Decathlon), the GLM-5.2 vs GLM-5.1 generational leap, and the MTP acceptance-length ablation. Every numeric point carries a sources[].ref and a value_basis naming the table row. Scores are reproduced verbatim from the Full Benchmark Table; cells the vendor published as "-" (not reported) are omitted, never estimated. CAVEAT: these are the model developer’s self-reported figures under their stated harnesses and prompts (see the post’s footnotes); some competitor HLE/CritPt cells are full-set scores. They are comparison-as-published, not an independent eval. Architecture context (not charted): 1M-token context (up from 200K), IndexShare cuts per-token indexer FLOPs ~2.9× at 1M, MIT-licensed open weights. Re-verified 2026-06-22.
Comparisons are informative, not definitive. See each source for definitions and limits.