Side-by-side benchmark comparison across agentic, coding, multimodal, knowledge, reasoning, and math workflows.
GPT-5.4 mini
68
0/8 categoriesHolo3-122B-A10B
~79
Winner · 1/8 categoriesGPT-5.4 mini· Holo3-122B-A10B
Pick Holo3-122B-A10B if you want the stronger benchmark profile. GPT-5.4 mini only becomes the better choice if you need the larger 400K context window or you want the stronger reasoning-first profile.
Holo3-122B-A10B is clearly ahead on the aggregate, 79 to 68. The gap is large enough that you do not need to squint at the spreadsheet to see the difference.
Holo3-122B-A10B's sharpest advantage is in agentic, where it averages 78.9 against 65.6. The single biggest benchmark swing on the page is OSWorld-Verified, 72.1% to 78.8%.
GPT-5.4 mini is also the more expensive model on tokens at $0.75 input / $4.50 output per 1M tokens, versus $0.40 input / $3.00 output per 1M tokens for Holo3-122B-A10B. GPT-5.4 mini is the reasoning model in the pair, while Holo3-122B-A10B 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-5.4 mini gives you the larger context window at 400K, compared with 64K for Holo3-122B-A10B.
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-5.4 mini | Holo3-122B-A10B |
|---|---|---|
| AgenticHolo3-122B-A10B wins | ||
| Terminal-Bench 2.0 | 60% | — |
| OSWorld-Verified | 72.1% | 78.8% |
| MCP Atlas | 57.7% | — |
| Toolathlon | 42.9% | — |
| tau2-bench | 93.4% | — |
| Coding | ||
| SWE-bench Pro | 54.4% | — |
| Multimodal & Grounded | ||
| MMMU-Pro | 76.6% | — |
| MMMU-Pro w/ Python | 78% | — |
| OmniDocBench 1.5 | 0.1263 | — |
| Reasoning | ||
| MRCRv2 | 40.7% | — |
| MRCR v2 64K-128K | 47.7% | — |
| MRCR v2 128K-256K | 33.6% | — |
| Graphwalks BFS 128K | 76.3% | — |
| Graphwalks Parents 128K | 71.5% | — |
| Knowledge | ||
| GPQA | 88% | — |
| HLE | 41.5% | — |
| HLE w/o tools | 28.2% | — |
| Instruction Following | ||
| IFEval | 87.4% | — |
| Multilingual | ||
| Coming soon | ||
| Mathematics | ||
| MATH-500 | 97.4% | — |
Holo3-122B-A10B is ahead overall, 79 to 68. The biggest single separator in this matchup is OSWorld-Verified, where the scores are 72.1% and 78.8%.
Holo3-122B-A10B has the edge for agentic tasks in this comparison, averaging 78.9 versus 65.6. 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.