AIEvals

Models

GLM 5.3

Z.ai’s GLM 5.3 release, with published coding, reasoning and knowledge-work evaluations.

Z.aiReleased 18 August 2026Context 1M tokensOutput 128k tokensOpen weights$1.40 in / $4.40 out per Mtok

Family GLM 5.363 benchmarks with a published resultNo first place on the independent columns among the 46 models this site listsRead 17 October 2026

Where it stands

Rank compares tracked models with a score. Published shows the source’s rank and field size, when available.

BenchmarkScoreRankPublishedLeaderBehind
Vals IndexVals AI53.51%16of 3316of 41Gemini 4 Argon15.39 pts
Artificial Analysis Intelligence IndexArtificial Analysis4514of 42no list readClaude Opus 5.513 pts
Terminal-Bench 4.0Artificial Analysis41.92%10of 40no list readClaude Sonnet 5.521.72 pts
SWE-bench (Verified)Vals AI95.40%5of 235of 88Claude Opus 51.60 pts
LiveCodeBenchVals AI80.54%20of 2168of 143Claude Fable 5.19.98 pts
GPQA DiamondVals AI88.13%20of 2335of 138Gemini 3.1 Pro7.32 pts
MMLU ProVals AI86.77%18of 2438of 138Claude Fable 5.15.61 pts
Vibe Code Bench v1.1Vals AI78.13%23of 3824of 106Claude Sonnet 5.514.26 pts
Terminal-Bench Science 0.1Artificial Analysis9.52%14of 24no list readGPT-6 Astra53.81 pts
ProgramBenchVals AI1.50%9of 299of 54Claude Opus 5.517.00 pts
τ³-BankingArtificial Analysis50.31%3of 30no list readQwen3.8 Max1.03 pts
ITBench-AAArtificial Analysis46.08%11of 18no list readGPT-5.6 Sol10.13 pts
MLCR-AAArtificial Analysis48.33%8of 37no list readClaude Sonnet 5.526.67 pts
SWE-Bench Pro V2 FullScale AI / SEAL95.60%5of 10no list readClaude Opus 53.80 pts
SWE-Bench Pro V2 HARDScale AI / SEAL84.30%7of 10no list readClaude Opus 513.70 pts
Vals RSI IndexVals AI22.91%11of 2011of 23Claude Opus 5.514.40 pts
Vibe Code Bench 1–100Vals AI19.99%8of 208of 20Claude Opus 5.510.37 pts
EnterpriseOps-Gym-AAArtificial Analysis36.44%19of 22no list readClaude Fable 514.68 pts
AutomationBench-AAArtificial Analysis62.20%10of 40no list readGemini 4 Argon15.31 pts
SciCodeArtificial Analysis59.03%8of 40no list readClaude Opus 5.57.87 pts
GDP.pdfArtificial Analysis11.20%37of 40no list readGPT-6.1 Sol19.80 pts
CritPtArtificial Analysis19.14%24of 42no list readGPT-5.6 Sol13.15 pts
AA-OmniscienceArtificial Analysis14.325of 42no list readClaude Opus 5.532.12
AA-LCR v1.1Artificial Analysis79.67%30of 42no list readKimi K39.00 pts
Terminal-Bench 2.1Vals AI71.54%23of 3526of 75Claude Opus 5.516.10 pts
ProofBench v1.1Vals AI49.00%25of 3429of 46Claude Sonnet 5.551.00 pts
LegalBenchVals AI84.84%17of 2727of 149Claude Fable 53.72 pts
Harvey's Legal Agent BenchmarkVals AI8.33%16of 3720of 73Muse Spark 1.217.09 pts
SkillsBenchVals AI47.51%16of 2025of 35DeepSeek V4.1 Flash22.29 pts
MysteryMechanismVals AI22.97%15of 2115of 21GPT-6 Astra30.18 pts
Code MigrationVals AI44.22%21of 3721of 71Claude Sonnet 5.525.61 pts
CyberBench v1.1Vals AI72.08%12of 3114of 43GPT-6 Sol5.90 pts
EMBVals AI56.34%30of 3741of 69Claude Fable 5.120.33 pts
Finance Agent (v2)Vals AI55.84%14of 3718of 73Gemini 4 Argon9.56 pts
IOIVals AI68.44%12of 3112of 38Gemini 4 Argon31.56 pts
Legal Research BenchVals AI49.04%7of 377of 72Muse Spark 1.3 Max6.25 pts
MedCodeVals AI42.86%28of 3547of 104Claude Opus 520.71 pts
MedScribeVals AI88.81%8of 369of 106Claude Opus 5.52.62 pts
Public Benefits Bench v1.1Vals AI68.54%8of 298of 45Claude Opus 58.39 pts
Tax Agent BenchVals AI73.09%5of 375of 64Claude Fable 5.14.55 pts
DeepSWE v1.1Datacurve68.96%7of 207of 28GPT-6 Astra5.16 pts
GDPval-AA v2.1Artificial Analysis1,643.629of 42no list readClaude Opus 5.5202.55
AA-Briefcase v1.1Artificial Analysis1,511.4115of 41no list readClaude Opus 5.5310.44
Humanity's Last ExamArtificial Analysis42.26%31of 42no list readClaude Opus 5.519.09 pts

Nearby on Vals Index

The two models above and below on the Vals Index (Vals AI), on every column where this model has a published number.

ModelVals Index%Artificial Analysis Intelligence IndexindexTerminal-Bench 4.0%SWE-bench%LiveCodeBench%GPQA Diamond%MMLU Pro%Vibe Code Bench v1.1%Terminal-Bench Science 0.1%ProgramBench%
Grok 4.7xAI54.95%4625.76%––––86.17%–0.50%
Gemini 3.8 FlashGoogle54.83%4119.70%80.00%89.48%94.44%90.22%78.65%10.00%1.00%
GLM 5.3Z.ai53.51%4541.92%95.40%80.54%88.13%86.77%78.13%9.52%1.50%
MiMo-V2.6-FlashXiaomi53.23%3822.73%––––78.96%–0.50%
GPT-5.6 TerraOpenAI53.09%4235.35%95.40%85.93%90.91%86.66%74.59%13.81%0.50%
Bold marks a figure better than this model’s on the same column, whichever model it belongs to. Ranks cover the models shown from the stated source.

What the hosted runtime reports

Ollama Cloud serves its own copy of this model, so its context cap and its prices are the ones a caller actually gets. Read 25 September 2026.

Parameters
753B parameters
Hosted context
1M tokens
Hosted price / Mtok
$1.40in · $0.260 cached · $4.40 out
Tool callingReasoning modeServed by the hostInput: Text

Z.ai's own harness numbers in the model card come from Claude Code at maximum reasoning effort, which is the setting its Terminal-Bench 3.0 and DeepSWE rows belong to.

ollama.com/library/glm-5.3

Results by benchmark

Independent and vendor results retain their own harnesses and effort settings.

BenchmarkIndependent runAs reported by a publisher
τ³-BankingTool use50.31%Artificial Analysisnot published
ITBench-AAIT operations46.08%Artificial Analysisnot published
MLCR-AAMedical reasoning48.33%Artificial Analysisnot published
SWE-Bench Pro V2 FullRepository coding95.60%Scale AI / SEALnot published
SWE-Bench Pro V2 HARDRepository coding84.30%Scale AI / SEAL
84.3%SWE Bench Pro v2-HardReflection AI · 5 October 2026Rank 2 of 4 in this publisher table
Vals RSI IndexAutonomous research22.91%Vals AInot published
Vibe Code Bench 1–100Application development19.99%Vals AInot published
EnterpriseOps-Gym-AAEnterprise workflows36.44%Artificial Analysisnot published
Humanity's Last Exam (text-only, without tools)Multidisciplinary reasoningno published result
42.0%Text-only HLE (w/o tools)DeepSeek · 17 October 2026Rank 2 of 3 in this publisher table
Apex Bench v1.1Reasoningno published result
56.6%Apex Bench v1.1DeepSeek · 17 October 2026Rank 4 of 6 in this publisher table
SWE Atlas Codebase QnACodebase understandingno published result
61.0%SWE Atlas Codebase QnAReflection AI · 5 October 2026Rank 2 of 3 in this publisher table
AutomationBench (public split)Business workflowsno published result
48.2%AutomationBench publicReflection AI · 5 October 2026Rank 2 of 6 in this publisher table
MCP AtlasAgentic toolingno published result
84.2%MCP AtlasReflection AI · 5 October 2026Rank 2 of 7 in this publisher table
AA-LCR (version not specified)Long-context reasoningno published result
79.7%AA-LCRReflection AI · 5 October 2026Rank 4 of 8 in this publisher table
AA-Omniscience (public split)Factual knowledgeno published result
22.6AA Omniscience (index) - public split only (our results)Reflection AI · 5 October 2026Rank 1 of 5 in this publisher table
AutomationBench-AAWorkflow automation62.20%Artificial Analysisnot published
SciCodeScientific coding59.03%Artificial Analysis
59.0%SciCodeReflection AI · 5 October 2026Rank 1 of 7 in this publisher table
GDP.pdfDocument reasoning11.20%Artificial Analysisnot published
CritPtScientific reasoning19.14%Artificial Analysis
19.1%CriPT AAReflection AI · 5 October 2026Rank 4 of 8 in this publisher table
AA-OmniscienceFactual knowledge14.3Artificial Analysisnot published
AA-LCR v1.1Long-context reasoning79.67%Artificial Analysisnot published
Terminal-Bench 4.0Agentic coding41.92%Artificial Analysis
38.89%Vals AI
37.9%Terminal-Bench 4.0DeepSeek · 17 October 2026Rank 4 of 6 in this publisher table
37.9%Terminal-Bench 4.0 (Pass@1)DeepSeek · 10 September 2026Rank 3 of 7 in this publisher table
Terminal-Bench 2.1Agentic coding71.54%Vals AI
83.90%Artificial Analysis
88.2%Terminal Bench v2.1Reflection AI · 5 October 2026Rank 3 of 8 in this publisher table
88.2%Terminal Bench 2.1Z.ai · 14 August 2026Rank 3 of 8 in this publisher table
88.2%Terminal-Bench 2.1 (Pass@1)DeepSeek · 10 September 2026Rank 5 of 7 in this publisher table
SWE-bench (Verified)Software engineering95.40%Vals AInot published
LiveCodeBenchCode generation80.54%Vals AInot published
GPQA DiamondScience reasoning88.13%Vals AI
91.72%Artificial Analysis
91.7%GPQA DiamondReflection AI · 5 October 2026Rank 3 of 8 in this publisher table
88.1%GPQA Diamond (Pass@1)DeepSeek · 10 September 2026Rank 7 of 7 in this publisher table
MMLU ProKnowledge86.77%Vals AInot published
Vibe Code Bench v1.1Product coding78.13%Vals AInot published
Terminal-Bench Science 0.1Agentic science9.52%Artificial Analysis
4.29%Vals AI
not published
ProgramBenchAgentic coding1.50%Vals AI
19.0%ProgramBenchDeepSeek · 17 October 2026Rank 5 of 6 in this publisher table
19.0%ProgramBench (Almost Solved)Z.ai · 14 August 2026Rank 3 of 7 in this publisher table
19.0%ProgramBench (Almost@1)DeepSeek · 10 September 2026Rank 4 of 6 in this publisher table
ProofBench v1.1Mathematics49.00%Vals AInot published
LegalBenchProfessional knowledge84.84%Vals AInot published
Harvey's Legal Agent BenchmarkProfessional knowledge8.33%Vals AInot published
SkillsBenchAgentic tooling47.51%Vals AInot published
MysteryMechanismScientific reasoning22.97%Vals AInot published
Vals IndexComposite53.51%Vals AInot published
Code MigrationAgentic coding44.22%Vals AInot published
CyberBench v1.1Security72.08%Vals AInot published
EMBFinancial modelling56.34%Vals AInot published
Finance Agent (v2)Financial analysis55.84%Vals AInot published
IOICompetitive programming68.44%Vals AInot published
Legal Research BenchProfessional knowledge49.04%Vals AInot published
MedCodeHealthcare administration42.86%Vals AInot published
MedScribeHealthcare administration88.81%Vals AInot published
Public Benefits Bench v1.1Public services68.54%Vals AInot published
Tax Agent BenchProfessional knowledge73.09%Vals AInot published
AutomationBenchBusiness workflowsno published result
48.8%AutomationBench (v1.0.6)DeepSeek · 17 October 2026Rank 4 of 6 in this publisher table
48.2%AutomationBench (v1.0.6)Z.ai · 14 August 2026Rank 1 of 8 in this publisher table
48.8%AutomationBench (Pass@1)DeepSeek · 10 September 2026Rank 3 of 7 in this publisher table
DeepSWE v1.1Software engineering68.96%Datacurve
66.9%DeepSWE v1.1DeepSeek · 17 October 2026Rank 6 of 6 in this publisher table
61.0%DeepSWE v1.1Reflection AI · 5 October 2026Rank 3 of 6 in this publisher table
66.9%DeepSWE (v1.1)Z.ai · 14 August 2026Rank 4 of 8 in this publisher table
66.9%DeepSWE v1.1 (Resolved)DeepSeek · 10 September 2026Rank 5 of 7 in this publisher table
GDPval-AA v2.1Knowledge work1,643.62Artificial Analysisnot published
AA-Briefcase v1.1Knowledge work1,511.41Artificial Analysisnot published
Humanity's Last ExamMultidisciplinary reasoning42.26%Artificial Analysis
42.3%HLE no toolsReflection AI · 5 October 2026Rank 3 of 8 in this publisher table
62.5%HLE w/ ToolsZ.ai · 14 August 2026Rank 3 of 8 in this publisher table
42.0%HLE (Pass@1)DeepSeek · 10 September 2026Rank 5 of 7 in this publisher table
62.5%HLE w/ tools (Pass@1)DeepSeek · 10 September 2026Rank 3 of 6 in this publisher table
Agents' Last ExamComputer useno published result
28.5%Agents' Last ExamDeepSeek · 17 October 2026Rank 4 of 6 in this publisher table
28.5%Agents' Last Exam (ALE-CLI)Z.ai · 14 August 2026Rank 2 of 8 in this publisher table
28.5%Agents' Last Exam (Pass@1)DeepSeek · 10 September 2026Rank 3 of 7 in this publisher table
Artificial Analysis Intelligence IndexComposite45Artificial Analysisnot published
Terminal-Bench 3.0Agentic codingno published result
28.3%Terminal Bench 3.0Z.ai · 14 August 2026Rank 3 of 6 in this publisher table
28.3%Terminal-Bench 3.0 (Pass@1)DeepSeek · 10 September 2026Rank 4 of 7 in this publisher table
NL2Repo-BenchAgentic codingno published result
58.0%NL2Repo-BenchDeepSeek · 17 October 2026Rank 4 of 6 in this publisher table
58.0%NL2RepoZ.ai · 14 August 2026Rank 3 of 6 in this publisher table
58.0%NL2Repo-Bench (Score)DeepSeek · 10 September 2026Rank 4 of 7 in this publisher table
CyberGymSecurityno published result
84.5%CyberGymZ.ai · 14 August 2026Rank 1 of 8 in this publisher table
84.5%CyberGym (Pass@1)DeepSeek · 10 September 2026Rank 2 of 6 in this publisher table
ExploitGymSecurityno published result
105ExploitGym (2h)Z.ai · 14 August 2026Rank 3 of 7 in this publisher table
130ExploitGym (6h)Z.ai · 14 August 2026Rank 3 of 7 in this publisher table
15.0%ExploitGym (Pass@1)DeepSeek · 10 September 2026Rank 4 of 6 in this publisher table
FrontierSWEAgentic researchno published result
78.1%FrontierSWEZ.ai · 14 August 2026Rank 2 of 4 in this publisher table
SWE-MarathonAgentic codingno published result
42.5%SWE-Marathon (v1.1)Z.ai · 14 August 2026Rank 3 of 6 in this publisher table
PostTrainBenchAI researchno published result
39.8%PostTrainBenchZ.ai · 14 August 2026Rank 2 of 6 in this publisher table
ExploitBenchSecurityno published result
54.4%ExploitBenchZ.ai · 14 August 2026Rank 3 of 7 in this publisher table
Toolathlon VerifiedAgentic toolingno published result
73.0%Toolathlon VerifiedZ.ai · 14 August 2026Rank 6 of 8 in this publisher table
GDPval-AA v2Knowledge workno published result
1,769GDPval-AA v2Z.ai · 14 August 2026Rank 1 of 8 in this publisher table
Independent column: the suite named under each figure, one harness per suite. Publisher column: each lab’s own table, reproduced on the benchmark’s page. Publisher ranks compare reported scores in the same table row only; ties share a rank and unreported cells are excluded.

Sources

Read on 17 October 2026.