GLM-5.3
- Input price
- $1.40
- Per 1 million tokens (USD)
- Output price
- $4.40
- Per 1 million tokens (USD)
- Effort levels
- 3total
- Default: Max
- Benchmark results
- 18total
- Benchmark types: 18
Model information
- API model ID
- glm-5.3
- release date
- 2026-08-18
- effort
- Low / High / Max (default)
- Effort source
- Based on Z.ai GLM-5.3 documentation (docs.z.ai/guides/llm/glm-5.3.md). thinking cannot be disabled.
- Claude Code
- unavailable
- Codex
- unavailable
- Availability notes
- Z.ai official GLM Coding Plan documentation provides instructions for use with Claude Code (mapping glm-5.3[1m] to Sonnet/Opus). Codex is not stated. thinking cannot be disabled. The connection instructions are in the model provider documentation, not official model options for Claude Code / Codex, so this is false.
- Availability source
- https://docs.z.ai/devpack/tool/claude
Price (USD per 1 million tokens)
Prices do not vary by effort; token consumption does.
| input | $1.40 |
|---|---|
| output | $4.40 |
| cache read | $0.26 |
| cache write | — |
Cache storage is free for a limited time. context 1M and maximum output 128K
GLM-5.3 | $1.4 | $0.26 | Limited-time Free | $4.4Benchmark results
| Benchmark | effort | Score | Source cost (metric) | measured/reported by | source | Conditions and extraction method |
|---|---|---|---|---|---|---|
| Terminal-Bench 3.0not stated | not stated | 28.3% | — | Z.ai | GLM-5.3 (Z.ai docs)official · Retrieved 2026-11-12 | Vendor-reported value. Self-reported by Z.ai. effort not stated (default is max)transcribed from textSource excerptTerminal-Bench 3.0 28.3 |
| Terminal-Bench 2.1not stated | not stated | 88.2% | — | Z.ai | GLM-5.3 (Hugging Face model card, zai-org)official · Retrieved 2026-11-13 | Vendor-reported value (self-reported by Z.ai). Card Note: the default for reasoning_effort is max; use the default max to reproduce the benchmarks. Claude Code 2.1.207; 6h timeout. Effort not stated in this footnote.transcribed from tableSource excerptTerminal Bench 2.1 | 88.2Location HF model card "Benchmark" table, Terminal Bench 2.1 row, GLM-5.3 column (leftmost numeric column). Basis: Footnotes, Terminal-Bench 2.1 |
| DeepSWE v1.1not stated | not stated | 66.9% | — | Z.ai | GLM-5.3 (Z.ai docs)official · Retrieved 2026-11-12 | Vendor-reported value. Self-reported by Z.ai. effort not stated (default is max)transcribed from textSource excerptDeepSWE v1.1 66.9 |
| Humanity's Last Exam (with tools)not stated | not stated | 62.5% | — | Z.ai | GLM-5.3 (Hugging Face model card, zai-org)official · Retrieved 2026-11-13 | Vendor-reported value (self-reported by Z.ai). Card Note: the default for reasoning_effort is max; use the default max to reproduce the benchmarks. Context management up to 300,000 tokens; grading by GPT-5.6-luna (medium). Effort not stated in this footnote.transcribed from tableSource excerptHLE w/ Tools | 62.5Location HF model card "Benchmark" table, HLE w/ Tools row, GLM-5.3 column (leftmost numeric column). Basis: Footnotes, HLE w/ tools |
| Agents' Last Exam (version and metric not stated)not stated | not stated | 28.5% | — | Z.ai | GLM-5.3 (Z.ai docs)official · Retrieved 2026-11-12 | Vendor-reported value. Self-reported by Z.ai. effort not stated (default is max). ALE version and metric not statedtranscribed from textSource excerptAgents' Last Exam 28.5 |
| PostTrainBenchMax (default) | Max | 39.8% | — | Z.ai | GLM-5.3 (Hugging Face model card, zai-org)official · Retrieved 2026-11-13 | Vendor-reported value (self-reported by Z.ai). Card Note: the default for reasoning_effort is max; use the default max to reproduce the benchmarks. Claude Code 2.1.207; max effort level; weighted average of 3 runs. The third-party API usage check was replaced with an LLM agent.transcribed from tableSource excerptPostTrainBench | 39.8Location HF model card "Benchmark" table, PostTrainBench row, GLM-5.3 column (leftmost numeric column). Basis: Footnotes, PostTrainBench |
| CyberGymnot stated | not stated | 84.5% | — | Z.ai | GLM-5.3 (Z.ai docs)official · Retrieved 2026-11-12 | Vendor-reported value. Self-reported by Z.ai. effort not stated (default is max)transcribed from textSource excerptCyberGym 84.5% |
| ExploitBenchnot stated | not stated | 54.4% | — | Z.ai | GLM-5.3 (Z.ai docs)official · Retrieved 2026-11-12 | Vendor-reported value. Self-reported by Z.ai. effort not stated (default is max)transcribed from textSource excerptExploitBench 54.4% |
| GPQA DiamondMax (default) | Max | 90.9% | — | Epoch AI | Epoch AI, Capabilities & benchmarking (benchmark_data.zip, gpqa_diamond.csv) CC BY 4.0third-party · Retrieved 2026-11-12 | Run by Epoch AI (third-party measurement). Source label "GLM-5.3 (max)". effort is indicated by the suffix _max in Epoch's Model version. Data retrieved 2026-11-12.transcribed from tableSource excerptgpqa_diamond.csv: Model version=glm-5.3_max, Best score (across scorers)=0.9090909090909091, stderr=0.015851578202142853, Started at=2026-08-24T19:44:18.000ZLocation gpqa_diamond.csv, data row 70 (file line 123), id=dLDBahYBTGRNjbsXA5pyr3, Model version=glm-5.3_max |
| FrontierMath Tiers 1-3(v2)Max (default) | Max | 68.8% | — | Epoch AI | Epoch AI, Capabilities & benchmarking (benchmark_data.zip, frontiermath_tiers_1_3_v2.csv) CC BY 4.0third-party · Retrieved 2026-11-12 | Run by Epoch AI (third-party measurement). Source label "GLM-5.3 (max)". effort is indicated by the suffix _max in Epoch's Model version. Data retrieved 2026-11-12.transcribed from tableSource excerptfrontiermath_tiers_1_3_v2.csv: Model version=glm-5.3_max, Best score (across scorers)=0.6877192982456141, stderr=0.027499133473298815, Started at=2026-08-25T14:20:20.000ZLocation frontiermath_tiers_1_3_v2.csv, data row 76 (file line 102), id=iXvqUzwxAiA3Rg3dwnutFs, Model version=glm-5.3_max |
| FrontierMath Tier 4(v2)Max (default) | Max | 29.3% | — | Epoch AI | Epoch AI, Capabilities & benchmarking (benchmark_data.zip, frontiermath_tier_4_v2.csv) CC BY 4.0third-party · Retrieved 2026-11-12 | Run by Epoch AI (third-party measurement). Source label "GLM-5.3 (max)". effort is indicated by the suffix _max in Epoch's Model version. Data retrieved 2026-11-12.transcribed from tableSource excerptfrontiermath_tier_4_v2.csv: Model version=glm-5.3_max, Best score (across scorers)=0.2926829268292683, stderr=0.07194088392074452, Started at=2026-08-25T14:20:20.000ZLocation frontiermath_tier_4_v2.csv, data row 30 (file line 63), id=4S3ZrJDkwh4gmGrmJQdvzR, Model version=glm-5.3_max |
| OTIS Mock AIME 2024-2025Max (default) | Max | 91.1% | — | Epoch AI | Epoch AI, Capabilities & benchmarking (benchmark_data.zip, otis_mock_aime_2024_2025.csv) CC BY 4.0third-party · Retrieved 2026-11-12 | Run by Epoch AI (third-party measurement). Source label "GLM-5.3 (max)". effort is indicated by the suffix _max in Epoch's Model version. Data retrieved 2026-11-12.transcribed from tableSource excerptotis_mock_aime_2024_2025.csv: Model version=glm-5.3_max, Best score (across scorers)=0.9111111111111111, stderr=0.03454754040013868, Started at=2026-08-24T19:44:18.000ZLocation otis_mock_aime_2024_2025.csv, data row 78 (file line 138), id=goZR5yGxZic9TZ23ezCL3w, Model version=glm-5.3_max |
| SimpleQA VerifiedMax (default) | Max | 41.0% | — | Epoch AI | Epoch AI, Capabilities & benchmarking (benchmark_data.zip, simpleqa_verified.csv) CC BY 4.0third-party · Retrieved 2026-11-12 | Run by Epoch AI (third-party measurement). Source label "GLM-5.3 (max)". effort is indicated by the suffix _max in Epoch's Model version. Data retrieved 2026-11-12.transcribed from tableSource excerptsimpleqa_verified.csv: Model version=glm-5.3_max, Best score (across scorers)=0.41, stderr=0.015560917136921659, Started at=2026-08-28T17:00:06.000ZLocation simpleqa_verified.csv, data row 37 (file line 66), id=TY8R8FV7zCwByEipkqyg2P, Model version=glm-5.3_max |
| Chess Puzzles(Epoch AI)Max (default) | Max | 37.0% | — | Epoch AI | Epoch AI, Capabilities & benchmarking (benchmark_data.zip, chess_puzzles.csv) CC BY 4.0third-party · Retrieved 2026-11-12 | Run by Epoch AI (third-party measurement). Source label "GLM-5.3 (max)". effort is indicated by the suffix _max in Epoch's Model version. Data retrieved 2026-11-12.transcribed from tableSource excerptchess_puzzles.csv: Model version=glm-5.3_max, Best score (across scorers)=0.37, stderr=0.048523658709390974, Started at=2026-10-13T21:22:18.000ZLocation chess_puzzles.csv, data row 26 (file line 56), id=kG3hbd8qLEd6fLAJHyQTiN, Model version=glm-5.3_max |
| Epoch Capabilities Index(ECI)not stated | not stated | 157.1 | — | Epoch AI | Epoch AI, epoch_capabilities_index/eci_scores.csv (CC BY 4.0)third-party · Retrieved 2026-11-12 | A per-model metric that combines results across multiple benchmarks and multiple efforts (not per effort). Calculated by Epoch AI.transcribed from tableSource excerpteci_scores.csv: Model=GLM-5.3, eci=157.1, eci_ci_low=154.94, eci_ci_high=159.53Location epoch_capabilities_index/eci_scores.csv, data row 32 (file line 33), Model=GLM-5.3 |
| SWE-Marathon v1.1Max (default) | Max | 42.5% | — | Z.ai | GLM-5.3 (Hugging Face model card, zai-org)official · Retrieved 2026-11-13 | Vendor-reported value (self-reported by Z.ai). Card Note: the default for reasoning_effort is max; use the default max to reproduce the benchmarks. Claude Code 2.1.207; maximum effort level. For some tasks, anti-cheat checks and build settings were changed.transcribed from tableSource excerptSWE-Marathon (v1.1) | 42.5Location HF model card "Benchmark" table, SWE-Marathon (v1.1) row, GLM-5.3 column (leftmost numeric column). Basis: Footnotes, SWE-Marathon |
| NL2Repo-Benchnot stated | not stated | 58 | — | Z.ai | GLM-5.3 (Hugging Face model card, zai-org)official · Retrieved 2026-11-13 | Vendor-reported value (self-reported by Z.ai). Card Note: the default for reasoning_effort is max; use the default max to reproduce the benchmarks. 1M context. Effort not stated in this footnote.transcribed from tableSource excerptNL2Repo | 58.0Location HF model card "Benchmark" table, NL2Repo row, GLM-5.3 column (leftmost numeric column). Basis: Footnotes, NL2Repo |
| AutomationBench v1.0.6not stated | not stated | 48.2% | — | Z.ai | GLM-5.3 (Hugging Face model card, zai-org)official · Retrieved 2026-11-13 | Vendor-reported value (self-reported by Z.ai). Card Note: the default for reasoning_effort is max; use the default max to reproduce the benchmarks. v1.0.6 (including the fix in PR #13). Effort not stated in this footnote.transcribed from tableSource excerptAutomationBench (v1.0.6) | 48.2Location HF model card "Benchmark" table, AutomationBench (v1.0.6) row, GLM-5.3 column (leftmost numeric column). Basis: Footnotes, AutomationBench |
Related rankings
- Overall ranking (all models, records)
- Terminal-Bench 3.0 ranking by metric (all models, records)
- Terminal-Bench 2.1 ranking by metric (all models, records)
- DeepSWE v1.1 ranking by metric (all models, records)
- Humanity's Last Exam (with tools) ranking by metric (all models, records)
- Agents' Last Exam (version and metric not stated) ranking by metric (all models, records)
- PostTrainBench ranking by metric (all models, records)
- CyberGym ranking by metric (all models, records)
- ExploitBench ranking by metric (all models, records)
- GPQA Diamond ranking by metric (all models, records)
- FrontierMath Tiers 1-3(v2) ranking by metric (all models, records)
- FrontierMath Tier 4(v2) ranking by metric (all models, records)
- OTIS Mock AIME 2024-2025 ranking by metric (all models, records)
- SimpleQA Verified ranking by metric (all models, records)
- Chess Puzzles(Epoch AI) ranking by metric (all models, records)
- Epoch Capabilities Index(ECI) ranking by metric (all models, records)
- SWE-Marathon v1.1 ranking by metric (all models, records)
- NL2Repo-Bench ranking by metric (all models, records)
- AutomationBench v1.0.6 ranking by metric (all models, records)
- Recommendations by role