Side-by-side benchmark comparison across agentic, coding, multimodal, knowledge, reasoning, and math workflows.
DeepSeek V3.2 (Thinking)
68
Winner · 1/8 categoriesGLM-5V-Turbo
~58
0/8 categoriesDeepSeek V3.2 (Thinking)· GLM-5V-Turbo
Pick DeepSeek V3.2 (Thinking) if you want the stronger benchmark profile. GLM-5V-Turbo only becomes the better choice if you need the larger 200K context window or you would rather avoid the extra latency and token burn of a reasoning model.
DeepSeek V3.2 (Thinking) is clearly ahead on the aggregate, 68 to 58. The gap is large enough that you do not need to squint at the spreadsheet to see the difference.
DeepSeek V3.2 (Thinking)'s sharpest advantage is in agentic, where it averages 69.4 against 58. The single biggest benchmark swing on the page is BrowseComp, 70% to 51.9%.
GLM-5V-Turbo is also the more expensive model on tokens at $1.20 input / $4.00 output per 1M tokens, versus $0.00 input / $0.00 output per 1M tokens for DeepSeek V3.2 (Thinking). That is roughly Infinityx on output cost alone. DeepSeek V3.2 (Thinking) is the reasoning model in the pair, while GLM-5V-Turbo is not. That usually helps on harder chain-of-thought-heavy tests, but it can also mean more latency and more token spend in real use. GLM-5V-Turbo gives you the larger context window at 200K, compared with 128K for DeepSeek V3.2 (Thinking).
BenchLM keeps the benchmark table and the operator tradeoffs on the same page so a better score does not hide a materially slower, pricier, or smaller-context model.
Runtime metrics show N/A when BenchLM does not have a sourced snapshot for that exact model. The scoring rules and freshness policy are documented on the methodology page.
| Benchmark | DeepSeek V3.2 (Thinking) | GLM-5V-Turbo |
|---|---|---|
| AgenticDeepSeek V3.2 (Thinking) wins | ||
| Terminal-Bench 2.0 | 71% | — |
| BrowseComp | 70% | 51.9% |
| OSWorld-Verified | 67% | 62.3% |
| BrowseComp-VL | — | 51.9% |
| OSWorld | — | 62.3% |
| AndroidWorld | — | 75.7% |
| WebVoyager | — | 88.5% |
| Coding | ||
| HumanEval | 79% | — |
| SWE-bench Verified | 48% | — |
| LiveCodeBench | 45% | — |
| SWE-bench Pro | 58% | — |
| Multimodal & Grounded | ||
| MMMU-Pro | 66% | — |
| OfficeQA Pro | 77% | — |
| Design2Code | — | 94.8% |
| Flame-VLM-Code | — | 93.8% |
| Vision2Web | — | 31.0% |
| ImageMining | — | 30.7% |
| MMSearch | — | 72.9% |
| MMSearch-Plus | — | 30.0% |
| SimpleVQA | — | 78.2% |
| Facts-VLM | — | 58.6% |
| V* | — | 89.0% |
| Reasoning | ||
| MuSR | 81% | — |
| BBH | 86% | — |
| LongBench v2 | 78% | — |
| MRCRv2 | 78% | — |
| ARC-AGI-2 | 4% | — |
| Knowledge | ||
| MMLU | 87% | — |
| GPQA | 85% | — |
| SuperGPQA | 83% | — |
| MMLU-Pro | 73% | — |
| HLE | 22% | — |
| FrontierScience | 77% | — |
| SimpleQA | 83% | — |
| Instruction Following | ||
| IFEval | 85% | — |
| Multilingual | ||
| MGSM | 84% | — |
| MMLU-ProX | 79% | — |
| Mathematics | ||
| AIME 2023 | 87% | — |
| AIME 2024 | 89% | — |
| AIME 2025 | 88% | — |
| HMMT Feb 2023 | 83% | — |
| HMMT Feb 2024 | 85% | — |
| HMMT Feb 2025 | 84% | — |
| BRUMO 2025 | 86% | — |
| MATH-500 | 84% | — |
DeepSeek V3.2 (Thinking) is ahead overall, 68 to 58. The biggest single separator in this matchup is BrowseComp, where the scores are 70% and 51.9%.
DeepSeek V3.2 (Thinking) has the edge for agentic tasks in this comparison, averaging 69.4 versus 58. Inside this category, BrowseComp is the benchmark that creates the most daylight between them.
Get notified when new models drop, benchmark scores change, or the leaderboard shifts. One email per week.
Free. No spam. Unsubscribe anytime. We only store derived location metadata for consent routing.