BlogLLM Benchmark Scores as Published, Side by Side
Qwen3.8-Max vs GLM-5.3: Benchmark Scores Side by Side
Every test both Qwen3.8-Max by Alibaba and GLM-5.3 by Z.ai have a published number for: 24 tests in 6 areas. Each number is as printed, with the page it comes from and the setup it was run under; nothing is ranked or rescored, and two numbers under different setups are two results, not one comparison.
Models Qwen3.8-Max, GLM-5.3Data version 2026103001, as of 2026-10-30
Model
What each developer prints on its own model page; a dash means it prints none.
Knowledge Work
| Test | Qwen3.8-Max | GLM-5.3 |
|---|---|---|
| AutomationBench | 27.3%Alibaba, Qwen3.8-Max comparison, checked 2026-10-14; printed on 1 more page | |
| GDPval-AA v2 | 1739Z.ai, GLM-5.3 model card, checked 2026-10-15 | |
| JobBench | 53.4Alibaba, Qwen3.8-Max comparison, checked 2026-10-14 | |
| Vals Finance Agent v2 | 50.6%Mistral AI, Mistral Large 4 (preview), checked 2026-10-15 | 55.8%Mistral AI, Mistral Large 4 (preview), checked 2026-10-15 |
| Harvey's Legal Agent Benchmark | 10.4%Mistral AI, Mistral Large 4 (preview), checked 2026-10-15 | 8.3%Mistral AI, Mistral Large 4 (preview), checked 2026-10-15 |
Agentic Coding
| Test | Qwen3.8-Max | GLM-5.3 |
|---|---|---|
| Terminal-Bench 2.1 | 86.6%Alibaba, Qwen3.8-Max comparison, checked 2026-10-14; printed on 1 more page | 88.2%Z.ai, GLM-5.3 model card, checked 2026-10-15; printed on 2 more pagesEffort: max; Harness: Claude Code 2.1.207 |
| DeepSWE v1.1 | 56.6%Alibaba, Qwen3.8-Max comparison, checked 2026-10-14; printed on 2 more pages | 66.9%Z.ai, GLM-5.3 model card, checked 2026-10-15; printed on 6 more pagesEffort: max; Harness: mini-swe-agent; Context: 400000 |
| Terminal-Bench 4.0 | 17%Mistral AI, Mistral Large 4 (preview), checked 2026-10-15 | 37.9%DeepSeek, DeepSeek-V4.1-Flash model card, checked 2026-10-15; printed on 4 more pages |
| NL2Repo-Bench | 55.9Alibaba, Qwen3.8-Max comparison, checked 2026-10-14; printed on 1 more page | 58Z.ai, GLM-5.3 model card, checked 2026-10-15; printed on 3 more pagesEffort: max; Context: 1000000 |
| ProgramBench Almost@1 | 10.5%Z.ai, GLM-5.3 model card, checked 2026-10-15 | |
| SWE-Atlas-QnA (setting not stated) | 62%Mistral AI, Mistral Large 4 (preview), checked 2026-10-15 | 59%Meta, Meta Muse Spark Open 816B-A42B model card, checked 2026-10-19; printed on 1 more pageEffort: max; Harness: preferred harness of each model |
| FrontierSWE (dominance score) | 73.5Alibaba, Qwen3.8-Max comparison, checked 2026-10-14 |
Science and Math
| Test | Qwen3.8-Max | GLM-5.3 |
|---|---|---|
| GPQA Diamond | 92.6%Alibaba, Qwen3.8-Max comparison, checked 2026-10-14 | 88.1%DeepSeek, DeepSeek-V4.1-Flash model card, checked 2026-10-15; printed on 2 more pages |
| FrontierMath Tier 4 (v2) | ||
| FrontierMath Tier 1-3 (v2) |
Cybersecurity
| Test | Qwen3.8-Max | GLM-5.3 |
|---|---|---|
| CyberGym | 78.5%Z.ai, GLM-5.3 model card, checked 2026-10-15 | 84.5%Z.ai, GLM-5.3 model card, checked 2026-10-15; printed on 2 more pagesEffort: max; Harness: Claude Code 2.1.207 |
| ExploitBench | 28.8%Z.ai, GLM-5.3 model card, checked 2026-10-15 | 54.4%Z.ai, GLM-5.3 model card, checked 2026-10-15; printed on 1 more pageEffort: max; Harness: Claude Code 2.1.207 |
| ExploitGym (2h) | 14Z.ai, GLM-5.3 model card, checked 2026-10-15Effort: max; Harness: Claude Code 2.1.207; Attempt budget: timeout budget 2 hours (rescaled API inference time plus overhead) | 105Z.ai, GLM-5.3 model card, checked 2026-10-15Effort: max; Harness: Claude Code 2.1.207; Attempt budget: timeout budget 2 hours (rescaled API inference time plus overhead) |
| ExploitGym (6h) | 26Z.ai, GLM-5.3 model card, checked 2026-10-15Effort: max; Harness: Claude Code 2.1.207; Attempt budget: timeout budget 6 hours (rescaled API inference time plus overhead) | 130Z.ai, GLM-5.3 model card, checked 2026-10-15Effort: max; Harness: Claude Code 2.1.207; Attempt budget: timeout budget 6 hours (rescaled API inference time plus overhead) |
General Knowledge and Reasoning
| Test | Qwen3.8-Max | GLM-5.3 |
|---|---|---|
| Humanity's Last Exam (no tools) | 43.6%Alibaba, Qwen3.8-Max comparison, checked 2026-10-14 | 42%DeepSeek, DeepSeek-V4.1-Flash model card, checked 2026-10-15; printed on 3 more pagesDataset revision: text-only subset of HLE (dagger on the card) |
| Humanity's Last Exam (with tools) | 56.2%Alibaba, Qwen3.8-Max comparison, checked 2026-10-14; printed on 1 more pageTools: enabled; exact tools in source | 62.5%Z.ai, GLM-5.3 model card, checked 2026-10-15; printed on 3 more pagesEffort: max; Context: 300000 |
| SimpleQA Verified |
Tool Use and Workflows
| Test | Qwen3.8-Max | GLM-5.3 |
|---|---|---|
| Agents' Last Exam (pass rate) | 27%Alibaba, Qwen3.8-Max comparison, checked 2026-10-14; printed on 1 more page | 28.5%Z.ai, GLM-5.3 model card, checked 2026-10-15; printed on 3 more pagesEffort: max; Harness: Claude Code; Context: 1000000 |
| Toolathlon Verified | 72.5%Alibaba, Qwen3.8-Max comparison, checked 2026-10-14; printed on 1 more page |
Price
API list prices in US dollars per one million tokens, as printed on the developer's pricing page on the day checked; a dash means the developer lists that price in another unit or not at all. Cost per task is measured by Vals AI.
| Price | Qwen3.8-Max | GLM-5.3 |
|---|---|---|
| Input | $1.65Alibaba, Alibaba Model Studio pricing, checked 2026-10-15 | $1.4Z.ai, Z.AI API pricing, checked 2026-10-15 |
| Cached input | - | $0.26Z.ai, Z.AI API pricing, checked 2026-10-15 |
| Output | $4.951Alibaba, Alibaba Model Studio pricing, checked 2026-10-15 | $4.4Z.ai, Z.AI API pricing, checked 2026-10-15 |
| Cost per taskVals AI, USD, lower is better |
Qwen3.8-Max has 21 more and GLM-5.3 has 49 more: tests only one of them has a number for, listed on its model page.
More side by side
Every number here is read from the same catalog file as the benchmark table, data version 2026103001. A pair has a page when both models are their developers' flagships and both have a number for at least 10 of the same tests. Where one page printed a 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.