Side-by-side benchmark comparison across agentic, coding, multimodal, knowledge, reasoning, and math workflows.
GPT-5.2-Codex
82
Winner · 1/8 categoriesHolo3-35B-A3B
~78
0/8 categoriesGPT-5.2-Codex· Holo3-35B-A3B
Pick GPT-5.2-Codex if you want the stronger benchmark profile. Holo3-35B-A3B only becomes the better choice if you want the cheaper token bill or you would rather avoid the extra latency and token burn of a reasoning model.
GPT-5.2-Codex is clearly ahead on the aggregate, 82 to 78. The gap is large enough that you do not need to squint at the spreadsheet to see the difference.
GPT-5.2-Codex's sharpest advantage is in agentic, where it averages 87 against 77.8. The single biggest benchmark swing on the page is OSWorld-Verified, 85% to 77.8%.
GPT-5.2-Codex is also the more expensive model on tokens at $2.00 input / $8.00 output per 1M tokens, versus $0.25 input / $1.80 output per 1M tokens for Holo3-35B-A3B. That is roughly 4.4x on output cost alone. GPT-5.2-Codex is the reasoning model in the pair, while Holo3-35B-A3B 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. GPT-5.2-Codex gives you the larger context window at 400K, compared with 64K for Holo3-35B-A3B.
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 | GPT-5.2-Codex | Holo3-35B-A3B |
|---|---|---|
| AgenticGPT-5.2-Codex wins | ||
| Terminal-Bench 2.0 | 90% | — |
| BrowseComp | 85% | — |
| OSWorld-Verified | 85% | 77.8% |
| Coding | ||
| HumanEval | 95% | — |
| SWE-bench Verified | 76% | — |
| LiveCodeBench | 66% | — |
| SWE-bench Pro | 86% | — |
| SWE-Rebench | 56.8% | — |
| Multimodal & Grounded | ||
| MMMU-Pro | 84% | — |
| OfficeQA Pro | 92% | — |
| Reasoning | ||
| MuSR | 93% | — |
| BBH | 90% | — |
| LongBench v2 | 90% | — |
| MRCRv2 | 91% | — |
| Knowledge | ||
| MMLU | 99% | — |
| GPQA | 97% | — |
| SuperGPQA | 95% | — |
| HLE | 26% | — |
| FrontierScience | 86% | — |
| SimpleQA | 95% | — |
| Instruction Following | ||
| IFEval | 92% | — |
| Multilingual | ||
| MGSM | 91% | — |
| MMLU-ProX | 87% | — |
| Mathematics | ||
| AIME 2023 | 99% | — |
| AIME 2024 | 99% | — |
| AIME 2025 | 98% | — |
| HMMT Feb 2023 | 95% | — |
| HMMT Feb 2024 | 97% | — |
| HMMT Feb 2025 | 96% | — |
| BRUMO 2025 | 96% | — |
GPT-5.2-Codex is ahead overall, 82 to 78. The biggest single separator in this matchup is OSWorld-Verified, where the scores are 85% and 77.8%.
GPT-5.2-Codex has the edge for agentic tasks in this comparison, averaging 87 versus 77.8. Inside this category, OSWorld-Verified 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.