Skip to main content
AI ModelsEvidence

GLM-5.2 vs the frontier across coding, reasoning and agentic benchmarks

Z.ai's GLM-5.2 closes most of the gap to the Western frontier: on FrontierSWE it scores 74.4 vs Claude Opus 4.8's 75.1 and GPT-5.5's 72.6 (and Gemini's 39.6), and it is a generational leap over GLM-5.1 (SWE-Marathon 13.0 vs 1.0).

Read every number with one caveat: all of it comes from a single vendor source (Z.ai's own release), so it is a self-reported benchmark table, not independent evaluation.

GLM-5.2 Benchmarks

Data date
June 2026
Source class
Official body or company filing
Data points
28
Sources
1
Value basis
28 actual

Caveat: Some competitor scores are marked full-set vs public-set, not always like-for-like.

This page can answer

Six series comparing GLM-5.2 to Opus 4.8, GPT-5.5, Gemini and its own predecessor GLM-5.1: long-horizon coding, standard coding, reasoning, agentic, a generational 5.2-vs-5.1 delta, and a multi-token-prediction (MTP) ablation.

It cannot answer

  • Independent verification: every figure is vendor self-reported (zai-glm52).
  • Real-world reliability, latency or cost (benchmark scores only).
Jump to chaptersSource ledger
Evidence chapters

Chapter 01 · Question answered

FrontierSWE-class long-task coding scores across models.

Long-horizon coding benchmarks

Long-horizon coding

Long-horizon coding

FrontierSWE, PostTrainBench, SWE-Marathon: GLM-5.2 vs the frontier.

View data & sources →

Long-horizon coding

Rendering the interactive chart…

Expand data table
SOURCEZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)
VALUE BASISActual
LAST VERIFIEDSep 10, 2026
How this is calculated

Chapter 02 · Question answered

Terminal-Bench / standard coding benchmarks.

Standard coding benchmarks

Standard coding

Standard coding

Terminal-Bench 2.1, SWE-bench Pro, NL2Repo, DeepSWE, ProgramBench.

View data & sources →

Standard coding

Rendering the interactive chart…

Expand data table
Download CSV
Standard coding benchmarks: data table
DimensionSeriesValueValue basisSource
Terminal-Bench 2.1Glm5163.5ActualZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)
Terminal-Bench 2.1Glm5281ActualZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)
Terminal-Bench 2.1Gpt5584ActualZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)
Terminal-Bench 2.1Gemini74ActualZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)
Terminal-Bench 2.1Opus4885ActualZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)
SWE-bench ProGlm5158.4ActualZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)
SWE-bench ProGlm5262.1ActualZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)
SWE-bench ProGpt5558.6ActualZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)
SWE-bench ProGemini54.2ActualZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)
SWE-bench ProOpus4869.2ActualZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)
NL2RepoGlm5142.7ActualZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)
NL2RepoGlm5248.9ActualZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)
NL2RepoGpt5550.7ActualZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)
NL2RepoGemini33.4ActualZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)
NL2RepoOpus4869.7ActualZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)
DeepSWEGlm5118ActualZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)
DeepSWEGlm5246.2ActualZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)
DeepSWEGpt5570ActualZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)
DeepSWEGemini10ActualZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)
DeepSWEOpus4858ActualZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)
ProgramBenchGlm5150.9ActualZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)
ProgramBenchGlm5263.7ActualZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)
ProgramBenchGpt5570.8ActualZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)
ProgramBenchGemini39.5ActualZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)
ProgramBenchOpus4871.9ActualZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)

Standard coding benchmarks: full data table

SOURCEZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)
VALUE BASISActual
LAST VERIFIEDSep 10, 2026
How this is calculated

Chapter 03 · Question answered

Reasoning benchmarks (e.g. HLE).

Reasoning benchmarks

Reasoning

Reasoning

HLE, CritPt, AIME 2026, HMMT, GPQA-Diamond, IMOAnswerBench.

View data & sources →

Reasoning

Rendering the interactive chart…

Expand data table
Download CSV
Reasoning benchmarks: data table
DimensionSeriesValueValue basisSourceMissing-data reason
HLEGlm5131ActualZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)Not published
HLEGlm5240.5ActualZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)Not published
HLEGpt5541.4ActualZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)Not published
HLEGemini45ActualZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)Not published
HLEOpus4849.8ActualZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)Not published
HLE w/ ToolsGlm5152.3ActualZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)Not published
HLE w/ ToolsGlm5254.7ActualZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)Not published
HLE w/ ToolsGpt5552.2ActualZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)Not published
HLE w/ ToolsGemini51.4ActualZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)Not published
HLE w/ ToolsOpus4857.9ActualZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)Not published
CritPtGlm514.6ActualZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)Not published
CritPtGlm5220.9ActualZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)Not published
CritPtGpt5527.1ActualZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)Not published
CritPtGemini17.7ActualZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)Not published
CritPtOpus4820.9ActualZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)Not published
AIME 2026Glm5195.3ActualZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)Not published
AIME 2026Glm5299.2ActualZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)Not published
AIME 2026Gpt5598.3ActualZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)Not published
AIME 2026Gemini98.2ActualZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)Not published
AIME 2026Opus4895.7ActualZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)Not published
HMMT Feb 2026Glm5182.6ActualZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)Not published
HMMT Feb 2026Glm5292.5ActualZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)Not published
HMMT Feb 2026Gpt5596.7ActualZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)Not published
HMMT Feb 2026Gemini87.3ActualZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)Not published
HMMT Feb 2026Opus4896.7ActualZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)Not published
GPQA-DiamondGlm5186.2ActualZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)Not published
GPQA-DiamondGlm5291.2ActualZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)Not published
GPQA-DiamondGpt5593.6ActualZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)Not published
GPQA-DiamondGemini94.3ActualZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)Not published
GPQA-DiamondOpus4893.6ActualZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)Not published
IMOAnswerBenchGlm5183.8ActualZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)Not published
IMOAnswerBenchGlm5291ActualZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)Not published
IMOAnswerBenchGpt55Not publishedActualZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)Not disclosed
IMOAnswerBenchGemini81ActualZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)Not published
IMOAnswerBenchOpus4883.5ActualZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)Not published

Reasoning benchmarks: full data table

SOURCEZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)
VALUE BASISActual
LAST VERIFIEDSep 10, 2026
MISSING DATA1 unavailable observation kept as gaps. IMOAnswerBench: Gpt55 not disclosed.

Caveat: Some competitor scores are marked full-set vs public-set, not always like-for-like.

How this is calculated

Unknown values are hatched and never treated as zero. Zero remains a verified value.

Chapter 04 · Question answered

Agentic/tool-use benchmarks (e.g. MCP-Atlas).

Agentic benchmarks

Agentic

Agentic

MCP-Atlas and Tool-Decathlon.

View data & sources →

Agentic

Rendering the interactive chart…

Expand data table
Download CSV

Agentic benchmarks: full data table

SOURCEZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)
VALUE BASISActual
LAST VERIFIEDSep 10, 2026
How this is calculated

Chapter 05 · Question answered

GLM-5.2 vs GLM-5.1 on the same tasks.

Generational leap: GLM-5.2 vs GLM-5.1

Generational delta

Generational delta

The biggest jumps over the predecessor.

View data & sources →

Generational delta

Rendering the interactive chart…

Expand data table
SOURCEZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)
VALUE BASISActual
LAST VERIFIEDSep 10, 2026
How this is calculated

Chapter 06 · Question answered

Effect of multi-token prediction on accepted length.

MTP speculative decoding: acceptance length

MTP ablation

MTP ablation

Ablation on coding scenarios; +20% over baseline.

View data & sources →

MTP ablation

Rendering the interactive chart…

Expand data table
Download CSV
MTP speculative decoding: acceptance length: data table
DimensionSeriesValueValue basisSource
BaselineAcc len4.56ActualZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)
+ IndexShare + KVShareAcc len5.1ActualZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)
+ Rejection SamplingAcc len5.29ActualZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)
+ End-to-end TV LossAcc len5.47ActualZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)

MTP speculative decoding: acceptance length: full data table

SOURCEZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)
VALUE BASISActual
LAST VERIFIEDSep 10, 2026
How this is calculated
All data for this index

Data table

Download CSV
GLM-5.2 Benchmarks: full data table
Glm51Glm52Gpt55GeminiOpus48SeriesBenchmarkSourceValue basisStepAcc len
30.574.472.639.675.1coding_longFrontierSWEZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)ActualNot publishedNot published
20.134.328.421.637.2coding_longPostTrainBenchZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)ActualNot publishedNot published
11312426coding_longSWE-MarathonZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)ActualNot publishedNot published
63.581847485coding_stdTerminal-Bench 2.1Z.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)ActualNot publishedNot published
58.462.158.654.269.2coding_stdSWE-bench ProZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)ActualNot publishedNot published
42.748.950.733.469.7coding_stdNL2RepoZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)ActualNot publishedNot published
1846.2701058coding_stdDeepSWEZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)ActualNot publishedNot published
50.963.770.839.571.9coding_stdProgramBenchZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)ActualNot publishedNot published
3140.541.44549.8reasoningHLEZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)ActualNot publishedNot published
52.354.752.251.457.9reasoningHLE w/ ToolsZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)ActualNot publishedNot published
4.620.927.117.720.9reasoningCritPtZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)ActualNot publishedNot published
95.399.298.398.295.7reasoningAIME 2026Z.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)ActualNot publishedNot published
82.692.596.787.396.7reasoningHMMT Feb 2026Z.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)ActualNot publishedNot published
86.291.293.694.393.6reasoningGPQA-DiamondZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)ActualNot publishedNot published
83.891Not published8183.5reasoningIMOAnswerBenchZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)ActualNot publishedNot published
71.876.875.369.277.8agenticMCP-AtlasZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)ActualNot publishedNot published
40.748.255.648.859.9agenticTool-DecathlonZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)ActualNot publishedNot published
113Not publishedNot publishedNot publishedgenerationalSWE-MarathonZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)ActualNot publishedNot published
4.620.9Not publishedNot publishedNot publishedgenerationalCritPtZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)ActualNot publishedNot published
1846.2Not publishedNot publishedNot publishedgenerationalDeepSWEZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)ActualNot publishedNot published
30.574.4Not publishedNot publishedNot publishedgenerationalFrontierSWEZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)ActualNot publishedNot published
20.134.3Not publishedNot publishedNot publishedgenerationalPostTrainBenchZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)ActualNot publishedNot published
63.581Not publishedNot publishedNot publishedgenerationalTerminal-Bench 2.1Z.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)ActualNot publishedNot published
50.963.7Not publishedNot publishedNot publishedgenerationalProgramBenchZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)ActualNot publishedNot published
Not publishedNot publishedNot publishedNot publishedNot publishedmtp_ablationNot publishedZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)ActualBaseline4.56
Not publishedNot publishedNot publishedNot publishedNot publishedmtp_ablationNot publishedZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)Actual+ IndexShare + KVShare5.1
Not publishedNot publishedNot publishedNot publishedNot publishedmtp_ablationNot publishedZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)Actual+ Rejection Sampling5.29
Not publishedNot publishedNot publishedNot publishedNot publishedmtp_ablationNot publishedZ.ai: GLM-5.2: Built for Long-Horizon Tasks (2026-06-16)Actual+ End-to-end TV Loss5.47

Method

How this evidence was built

What it measures

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.

Key context

  • Every numeric point carries a source citation and a basis label naming the table row.
  • Scores are reproduced verbatim from the Full Benchmark Table; cells the vendor published as "-" (not reported) are omitted, never estimated.

Key numbers

  • FrontierSWE: glm-5.2 74.4 vs Opus 4.8 75.1 vs gpt-5.5 72.6 vs Gemini 39.6 vs glm-5.1 30.5 (zai-glm52).
  • Terminal-Bench 2.1: glm-5.2 81 vs Opus 4.8 85 vs gpt-5.5 84 vs Gemini 74 (zai-glm52).
  • Generational leap: swe-Marathon 13.0 (5.2) vs 1.0 (5.1) (zai-glm52).
  • Mtp ablation quantifies the multi-token-prediction contribution to accepted length (zai-glm52).

Caveats

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.

Caveat: Some competitor scores are marked full-set vs public-set, not always like-for-like.

Do not use this to

  • Independent verification: every figure is vendor self-reported (zai-glm52).
  • Real-world reliability, latency or cost (benchmark scores only).

Scope

June 2026

Sources

Last verified 10 Sept 2026

Methods that use this evidence

Use one of these copy-paste methods when you want an AI to work from this evidence.

How relevant was this information?

Similar datasets