BlogLLM Benchmark Scores as Published, Side by Side
GLM-5.3 vs Mistral Large 4: Benchmark Scores Side by Side
Every test both GLM-5.3 by Z.ai and Mistral Large 4 by Mistral AI have a published number for: 10 tests in 7 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 GLM-5.3, Mistral Large 4Data version 2026111201, as of 2026-11-12
Model
What each developer prints on its own model page; a dash means it prints none. Runs on is the smallest device that holds the model at 4-bit: the developer's words where it prints them, otherwise our estimate from the size.
| Fact | GLM-5.3 | Mistral Large 4 |
|---|---|---|
| Released | 2026-08-18 | 2026-10-29 |
| Context window | 1M | 1M |
| Max output | 128K | - |
| Input | Text | Text, image |
| Output | Text | Text |
| Weights | Open | Open (Mistral Ecosystem Open-Weight License) |
| Size | 753.3B | 1T (52B active) |
| Runs on | 8 × H100 (about 499 GB at 4-bit, estimate) | 8 × B300 (about 662 GB at 4-bit, estimate) |
Knowledge Work
| Test | GLM-5.3 | Mistral Large 4 |
|---|---|---|
| FinWorkBench (Finch) | 65.1%Mistral AI, Mistral Large 4 (preview), checked 2026-10-15; printed on 1 more page | 73.3%Mistral AI, Mistral Large 4 (Le Chonk): Built for Coders, Trusted by Defenders, checked 2026-11-05 |
Agentic Coding
| Test | GLM-5.3 | Mistral Large 4 |
|---|---|---|
| Terminal-Bench 4.0 | 37.9%DeepSeek, DeepSeek-V4.1-Flash model card, checked 2026-10-15; printed on 5 more pages | 41%Mistral AI, Mistral Large 4 (Le Chonk): Built for Coders, Trusted by Defenders, checked 2026-11-05 |
| Vibe Code Bench | ||
| FrontierCode 2.0 | 17.8%Mistral AI, Mistral Large 4 (Le Chonk): Built for Coders, Trusted by Defenders, checked 2026-11-05 | 19.7%Mistral AI, Mistral Large 4 (Le Chonk): Built for Coders, Trusted by Defenders, checked 2026-11-05 |
Science and Math
| Test | GLM-5.3 | Mistral Large 4 |
|---|---|---|
| GPQA Diamond | 88.1%DeepSeek, DeepSeek-V4.1-Flash model card, checked 2026-10-15; printed on 2 more pages |
Cybersecurity
| Test | GLM-5.3 | Mistral Large 4 |
|---|---|---|
| Artificial Analysis Cyber Index (successes) | 36Mistral AI, Mistral Large 4 (preview), checked 2026-10-15; printed on 1 more page | 51Mistral AI, Mistral Large 4 (Le Chonk): Built for Coders, Trusted by Defenders, checked 2026-11-05 |
General Knowledge and Reasoning
| Test | GLM-5.3 | Mistral Large 4 |
|---|---|---|
| SimpleQA Verified |
Web Search and Research
| Test | GLM-5.3 | Mistral Large 4 |
|---|---|---|
| BrowseComp | 82.5%Mistral AI, Mistral Large 4 (Le Chonk): Built for Coders, Trusted by Defenders, checked 2026-11-05 | 84.5%Mistral AI, Mistral Large 4 (Le Chonk): Built for Coders, Trusted by Defenders, checked 2026-11-05 |
Safety and Alignment
| Test | GLM-5.3 | Mistral Large 4 |
|---|---|---|
| B3 AI Security Benchmark (attack resistance) | 93.3%Mistral AI, Mistral Large 4 (preview), checked 2026-10-15; printed on 1 more page | 97.1%Mistral AI, Mistral Large 4 (Le Chonk): Built for Coders, Trusted by Defenders, checked 2026-11-05 |
| Cyber refusal rate (average of JailbreakBench, StrongREJECT, AgentHarm) | 87%Mistral AI, Mistral Large 4 (preview), checked 2026-10-15; printed on 1 more page | 96.6%Mistral AI, Mistral Large 4 (Le Chonk): Built for Coders, Trusted by Defenders, checked 2026-11-05 |
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 | GLM-5.3 | Mistral Large 4 |
|---|---|---|
| Input | $1.4Z.ai, Z.AI API pricing, checked 2026-10-15 | $0.7Mistral AI, Mistral API pricing, checked 2026-11-05 |
| Cached input | $0.26Z.ai, Z.AI API pricing, checked 2026-10-15 | $0.07Mistral AI, Mistral API pricing, checked 2026-11-05 |
| Output | $4.4Z.ai, Z.AI API pricing, checked 2026-10-15 | $2.1Mistral AI, Mistral API pricing, checked 2026-11-05 |
| Cost per taskVals AI, USD, lower is better |
GLM-5.3 has 67 more and Mistral Large 4 has 2 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 2026111201. 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.