BlogLLM Benchmark Scores as Published, Side by Side
GLM-5.3: Benchmark Scores as Published
Every result the catalog holds for GLM-5.3: 73 tests in 13 areas, from 13 pages. 18 printed by Z.ai itself, the rest by the company named beside each number. Each number is as printed; nothing is ranked or rescored.
Developer Z.aiReleased 2026-08-18Context window 1MMax output 128KInput TextOutput TextWeights OpenData version 2026103001, as of 2026-10-30
Knowledge Work
Agentic Coding
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| Terminal-Bench 2.1 | 88.2% | Z.ai, GLM-5.3 model cardchecked 2026-10-15; printed on 2 more pages | Effort: max; Harness: Claude Code 2.1.207 |
| DeepSWE v1.1 | 66.9% | Z.ai, GLM-5.3 model cardchecked 2026-10-15; printed on 6 more pages | Effort: max; Harness: mini-swe-agent; Context: 400000 |
| Terminal-Bench 4.0 | 37.9% | DeepSeek, DeepSeek-V4.1-Flash model cardchecked 2026-10-15; printed on 4 more pages | - |
| NL2Repo-Bench | 58 | Z.ai, GLM-5.3 model cardchecked 2026-10-15; printed on 3 more pages | Effort: max; Context: 1000000 |
| ProgramBench Almost@1 | 19% | Z.ai, GLM-5.3 model cardchecked 2026-10-15; printed on 3 more pages | Effort: max |
| SWE-Marathon | 42.5% | Moonshot AI, Introducing Kimi K3.1 Prochecked 2026-10-24 | Effort: max; Dataset revision: v1.1 |
| Terminal-Bench 3.0 | 28.3% | Z.ai, GLM-5.3 model cardchecked 2026-10-15; printed on 1 more page | Effort: max; Harness: Claude Code 2.1.207; Context: 400000 |
| SWE-Atlas-QnA (setting not stated) | 59% | Meta, Meta Muse Spark Open 816B-A42B model cardchecked 2026-10-19; printed on 1 more page | Effort: max; Harness: preferred harness of each model |
| FrontierSWE (dominance score) | 78.1 | Z.ai, GLM-5.3 model cardchecked 2026-10-15 | Effort: max; Context: 1000000 |
| Kimi Code Bench 2.0 | 61.7% | Moonshot AI, Introducing Kimi K3.1 Prochecked 2026-10-24 | Effort: max |
| SWE-Marathon v1.1 (resolution rate) | 42.5% | Z.ai, GLM-5.3 model cardchecked 2026-10-15 | Effort: max; Harness: Claude Code 2.1.207; Context: 1000000 |
| Kimi Code Bench 3.0 Long-horizon | 45.5% | Moonshot AI, Introducing Kimi K3.1 Prochecked 2026-10-24 | Effort: max |
| Z.ai Code Bench | 31.4% | Z.ai, GLM-5.3 postchecked 2026-10-15 | Effort: High |
| 34.5% | Z.ai, GLM-5.3 postchecked 2026-10-15 | Effort: Max |
ML Engineering
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| PostTrainBench v1.1 | 39.8% | Z.ai, GLM-5.3 model cardchecked 2026-10-15; printed on 1 more page | Effort: max; Harness: Claude Code 2.1.207; Context: 1000000 |
| PostTrainBench v1.2 | 34.5% | Moonshot AI, Introducing Kimi K3.1 Prochecked 2026-10-24 | Effort: max; Dataset revision: v1.2 |
Science and Math
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| GPQA Diamond | 88.1% | DeepSeek, DeepSeek-V4.1-Flash model cardchecked 2026-10-15; printed on 2 more pages | - |
| SciCode | 59% | Meta, Meta Muse Spark Open 816B-A42B model cardchecked 2026-10-19 | Effort: max |
| Terminal-Bench Science 0.1 | 8.1% | Moonshot AI, Introducing Kimi K3.1 Prochecked 2026-10-24 | Effort: max |
| FrontierMath Tier 4 (v2) | 29.3% | Epoch AI, Epoch AI Benchmarking Hubchecked 2026-10-30 | Effort: max |
| FrontierMath Tier 1-3 (v2) | 68.8% | Epoch AI, Epoch AI Benchmarking Hubchecked 2026-10-30 | Effort: max |
| SciCode-Verified | 92.5% | Mistral AI, Mistral Large 4 (preview)checked 2026-10-15 | Attempt budget: pass@1 (n=6) |
| BioMysteryBench | 79.6% | Moonshot AI, Introducing Kimi K3.1 Prochecked 2026-10-24 | Effort: max; Harness: Claude Code |
Long Context
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| LOCA-Bench (256k environment description) | 50.3% | MiniMax, MiniMax-M3.1-Flash model cardchecked 2026-10-25 | - |
Multimodal Understanding
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| GDP.pdf | 11.2% | MiniMax, MiniMax-M3.1-Flash model cardchecked 2026-10-25; printed on 1 more page | - |
Cybersecurity
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| CyberGym | 84.5% | Z.ai, GLM-5.3 model cardchecked 2026-10-15; printed on 2 more pages | Effort: max; Harness: Claude Code 2.1.207 |
| ExploitBench | 54.4% | Z.ai, GLM-5.3 model cardchecked 2026-10-15; printed on 1 more page | Effort: max; Harness: Claude Code 2.1.207 |
| ExploitGym | 15% | DeepSeek, DeepSeek-V4.1-Flash model cardchecked 2026-10-15; printed on 2 more pages | - |
| Artificial Analysis Cyber Index (successes) | 36 | Mistral AI, Mistral Large 4 (preview)checked 2026-10-15 | - |
| CyberGym-E2E | 29% | Mistral AI, Mistral Large 4 (preview)checked 2026-10-15 | - |
| ExploitGym (2h) | 105 | Z.ai, GLM-5.3 model cardchecked 2026-10-15 | Effort: max; Harness: Claude Code 2.1.207; Attempt budget: timeout budget 2 hours (rescaled API inference time plus overhead) |
| ExploitGym (6h) | 130 | Z.ai, GLM-5.3 model cardchecked 2026-10-15 | Effort: max; Harness: Claude Code 2.1.207; Attempt budget: timeout budget 6 hours (rescaled API inference time plus overhead) |
| xdev-bench-hard, problems solved (pass@3) | 57.1% | Moonshot AI, Introducing Kimi K3.1 Prochecked 2026-10-24 | Harness: Claude Code; Attempt budget: pass@3 |
| Cybench | 85% | Mistral AI, Mistral Large 4 (preview)checked 2026-10-15 | - |
| xdev-bench-hard, average success rate | 34.9% | Moonshot AI, Introducing Kimi K3.1 Prochecked 2026-10-24 | Harness: Claude Code; Attempt budget: pass@3 |
General Knowledge and Reasoning
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| Humanity's Last Exam (no tools) | 42% | DeepSeek, DeepSeek-V4.1-Flash model cardchecked 2026-10-15; printed on 3 more pages | Dataset revision: text-only subset of HLE (dagger on the card) |
| Humanity's Last Exam (with tools) | 62.5% | Z.ai, GLM-5.3 model cardchecked 2026-10-15; printed on 3 more pages | Effort: max; Context: 300000 |
| SimpleQA Verified | 41% | Epoch AI, Epoch AI Benchmarking Hubchecked 2026-10-30 | Effort: max |
| AA-Omniscience Index | 14 | Meta, Meta Muse Spark Open 816B-A42B model cardchecked 2026-10-19 | Effort: max |
| Humanity's Last Exam Diamond (no tools) | 16.4% | Meta, Meta Muse Spark Open 816B-A42B model cardchecked 2026-10-19; printed on 1 more page | Effort: max; Tools: none |
| Kimi Hallu Bench Hard w/ tools (lower is better)lower is better | 22.6% | Moonshot AI, Introducing Kimi K3.1 Prochecked 2026-10-24 | Effort: max; Tools: with tools |
| Kimi Hallu Bench w/o tools (lower is better)lower is better | 14.8% | Moonshot AI, Introducing Kimi K3.1 Prochecked 2026-10-24 | Effort: max; Tools: none |
Tool Use and Workflows
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| Agents' Last Exam (pass rate) | 28.5% | Z.ai, GLM-5.3 model cardchecked 2026-10-15; printed on 3 more pages | Effort: max; Harness: Claude Code; Context: 1000000 |
| Toolathlon Verified | 73% | Z.ai, GLM-5.3 model cardchecked 2026-10-15; printed on 1 more page | Effort: max |
| MCP Atlas | 84.2% | Meta, Meta Muse Spark Open 816B-A42B model cardchecked 2026-10-19 | Effort: max |
| BankerToolBench | 74.5% | MiniMax, MiniMax-M3.1-Flash model cardchecked 2026-10-25 | - |
| Agents' Last Exam (CLI) | 29.4% | Moonshot AI, Introducing Kimi K3.1 Prochecked 2026-10-24 | Effort: max; Dataset revision: v1.1 |
Instruction Following
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| HANDBOOK.md | 20.8% | Moonshot AI, Introducing Kimi K3.1 Prochecked 2026-10-24 | Effort: max; Harness: OpenHands |
Safety and Alignment
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| Gray Swan IPI (attack success rate after k attempts)attack success rate after k attempts, lower is better | 31.5% | Google, Gemini 4 Argon Safety and Security Reportchecked 2026-10-19; printed on 1 more page | - |
| B3 AI Security Benchmark (attack resistance) | 93.3% | Mistral AI, Mistral Large 4 (preview)checked 2026-10-15 | - |
| Cyber refusal rate (average of JailbreakBench, StrongREJECT, AgentHarm) | 87% | Mistral AI, Mistral Large 4 (preview)checked 2026-10-15 | - |
| FORTRESS adversarial (static), ARSlower is better | 28.2 | Meta, Meta Muse Spark Open 816B-A42B model cardchecked 2026-10-19 | Effort: max |
| FORTRESS adversarial (static), ORSlower is better | 3.2 | Meta, Meta Muse Spark Open 816B-A42B model cardchecked 2026-10-19 | Effort: max |
| KORA Benchmark (0 to 2) | 1.7 | Mistral AI, Mistral Large 4 (preview)checked 2026-10-15 | - |
| StrongREJECT v2 (adaptive)lower is better | 62.9 | Meta, Meta Muse Spark Open 816B-A42B model cardchecked 2026-10-19 | Effort: max |
Engineering and Design
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| Design Bench (Internal, Elo) | 718 | Moonshot AI, Introducing Kimi K3.1 Prochecked 2026-10-24 | Effort: max |
Preference and Composite Rankings
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| Artificial Analysis Intelligence Index | 45 | Meta, Meta Muse Spark Open 816B-A42B model cardchecked 2026-10-19 | Effort: max |
| Surge AI human eval on code (1 to 5) | 3.6 | Mistral AI, Mistral Large 4 (preview)checked 2026-10-15 | - |
Price
API list prices in US dollars per one million tokens, as printed on the developer's pricing page on the day checked. Cost per task is measured by Vals AI.
| Price | Value | Printed by | Setup as printed |
|---|---|---|---|
| Input | $1.4 | Z.ai, Z.AI API pricingchecked 2026-10-15 | - |
| Cached input | $0.26 | Z.ai, Z.AI API pricingchecked 2026-10-15 | - |
| Output | $4.4 | Z.ai, Z.AI API pricingchecked 2026-10-15 | - |
| Cost per taskVals AI, USD, lower is better | $12.45 | Vals AI, Vibe Code Bench v1.1 leaderboardchecked 2026-10-27 | Harness: OpenHands |
Side by side
More from Z.ai
Every number here is read from the same catalog file as the benchmark table, data version 2026103001. Where one page printed the model under several setups, each value is listed. The catalog is public in armanobosyan/sugra-llm-benchmarks; a wrong number can be reported from the table, with Report an error under the value, or as an issue in that repository.