GLM 5.3 Flash
Z.ai·Released Aug 20, 2026Open weights
Updated Oct 8, 11:55 PM ET
The score is on the board's scale (reference average 50, about 15 points per standard deviation), not a percentage.
By domain
Each domain is scored on the same scale as the overall score. Coverage is the share of the domain's kinds of evaluation with a result; it says how much evidence there is, not how good the model is.
Every result behind the score
The result as the evaluator publishes it, and the reading it gives on the board's scale. A hard evaluation can give a high reading for a modest result; results near 95% or 5% only say "at least" or "at most".
Coding
9 results · 55.7- Terminal-Bench 4.0 (AA run)Artificial Analysis32.8%Reading 69.8
- CyberBench PatchVals AI · max effort85.7%Reading 68.5
- ProgramBenchVals AI · max effortNear the floor0.0%Reading at most 65.7
- LiveBench CodingLiveBench67.9%Reading 61.5
- APEX-SWEMercor · max effort46.1%Reading 61.3
- Vibe Code Bench 1–100Vals AI · max effort16.0%Reading 56.8
- SciCodeArtificial Analysis51.6%Reading 54.0
- Code MigrationVals AI · max effort20.5%Reading 42.2
- Vibe Code BenchVals AI · max effort30.8%Reading 36.9
Research and reasoning
8 results · 37.6- Humanity's Last Exam (AA run)Artificial Analysis39.9%Reading 56.5
- Terminal-Bench ScienceVals AI · max effortNear the floor2.9%Reading at most 49.0
- FrontierMath Tiers 1–3Epoch AI · max effort55.8%Reading 46.1
- ProofBenchVals AI · max effort21.0%Reading 41.2
- FrontierMath Tier 4Epoch AI · max effort17.1%Reading 40.4
- LiveBench ReasoningLiveBench79.4%Reading 36.6
- Chess PuzzlesEpoch AI · max effort14.0%Reading 17.6
- Mystery Game PuzzlesEpoch AI · max effortCapped8.0%Reading 11.7
Professional work
8 results · 60.5- τ-Bench Banking (AA run)Artificial Analysis47.2%Reading 72.0
- Finance AgentVals AI · max effort57.9%Reading 69.1
- Legal Research BenchVals AI · max effort45.2%Reading 67.8
- Tax Agent BenchVals AI · max effort31.6%Reading 65.3
- LiveBench Data AnalysisLiveBench76.4%Reading 62.1
- APEX-AgentsMercor · max effort52.8%Reading 60.3
- EMBVals AI · max effort55.9%Reading 52.3
- Harvey Legal Agent BenchmarkVals AI · max effort6.7%Reading 40.4
Knowledge and accuracy
3 results · 39.3- LiveBench LanguageLiveBench77.3%Reading 50.6
- AA-LCRArtificial Analysis80.0%Reading 49.1
- LiveBench Instruction FollowingLiveBenchCapped52.8%Reading -1.4
Human preference
1 result · 48.3- Arena TextLMArena1,473.0Reading 47.3
Shown, not scored
Composite indices, saturated or older evaluations, evaluations still being checked, and vision, writing and multilingual results. They do not move the score.
- IOIVals AI · CodingReference only52.5%
- LiveCodeBenchVals AI · CodingReference only80.5%
- SkillsBenchVals AI · CodingWatching40.2%
- SRE BenchVals AI · CodingWatching0.8%
- SWE-bench VerifiedVals AI · CodingReference only92.0%
- OTIS Mock AIMEEpoch AI · Research and reasoningReference only93.9%
- APEX-AccountingMercor · Professional workWatching1.7%
- LegalBenchVals AI · Professional workReference only83.9%
- MedScribeVals AI · Professional workWatching88.9%
- TaxEvalVals AI · Professional workReference only75.6%
- GPQA DiamondEpoch AI · Knowledge and accuracyReference only90.2%
- GPQA DiamondVals AI · Knowledge and accuracyReference only86.4%
- MMLU-ProVals AI · Knowledge and accuracyReference only86.1%
- AA Intelligence IndexArtificial Analysis · Composite indicesReference only41.8
- LiveBench averageLiveBench · Composite indicesReference only71.6%
- Arena VisionLMArena · VisionReference only1,278.4
- MMMU ProVals AI · VisionReference only86.0%
- Arena WebDevLMArena · Writing and designReference only1,610.0
No result yet
Scored evaluations this model has not taken. A missing result neither adds nor subtracts.
- Terminal-Bench 4.0 · Vals AI
- MysteryMechanism · Vals AI
- MedCode · Vals AI
- τ²-Bench Telecom (AA run) · Artificial Analysis
- SimpleQA Verified · Epoch AI
- BullshitBench · BullshitBench
- IFBench (AA run) · Artificial Analysis