Nemotron 3.5 Super VL (120B A12B) vs Claude Opus 5.5
How Nemotron 3.5 Super VL (120B A12B) and Claude Opus 5.5 stack up across benchmarks, pricing, speed, and the workloads that matter.
Nemotron 3.5 Super VL (120B A12B)
NVIDIA · Open weights
Claude Opus 5.5
Anthropic · Proprietary
Nemotron 3.5 Super VL (120B A12B) is the cheaper of the two at $0.29 per million blended tokens, about 20.0x less than Claude Opus 5.5 at $5.78. On math, Claude Opus 5.5 leads by about 22 points in our normalized benchmark average.
The table below breaks down every workload axis where both models have published results. Each score is a normalized 0 to 1 average across the benchmarks tagged with that category. A gap of a few points is noise; a gap of ten or more is a real difference in capability. Where one model has not reported a result, the cell shows n/a and the missing benchmark does not drag its average down.
Benchmark scores by workload
| Workload | Nemotron 3.5 Super VL (120B A12B) | Claude Opus 5.5 | Edge |
|---|---|---|---|
| Reasoning | 55% | 74% | +19.4 pts |
| Coding | 61% | 73% | +12.4 pts |
| Agents & tool use | 47% | 63% | +15.7 pts |
| Math | 52% | 74% | +22.1 pts |
| Vision & multimodal | 65% | 80% | +15.1 pts |
| Long context | 66% | n/a | n/a |
| Writing | n/a | 61% | n/a |
| Knowledge & factuality | 55% | 77% | +21.5 pts |
The headline benchmarks below are the most widely cited individual tests. GPQA Diamond measures graduate-level reasoning in physics, chemistry, and biology. SWE-Bench Verified tests whether a model can fix real GitHub issues. MMMU-Pro covers college-level multimodal understanding across six disciplines. AIME and FrontierMath push competitive and research math. BrowseComp measures web research ability.
Headline benchmarks
| Benchmark | Nemotron 3.5 Super VL (120B A12B) | Claude Opus 5.5 |
|---|---|---|
| GPQA | 90% | n/a |
| SWE-Bench Verified | 71% | n/a |
| BrowseComp | 78% | n/a |
Benchmarks tell you what a model can do in a controlled setting. They do not tell you whether it will work on your specific task, with your specific data, at your specific scale. The recommendations below map each common workload to whichever of these two models scores higher on the relevant axis. Use them as a starting point, not a final answer.
Which should you choose?
Hard reasoning
Multi-step analysis, research, and problems that need sustained thought.
Claude Opus 5.5Coding & software
Writing, reviewing, and debugging code across a real codebase.
Claude Opus 5.5Agents & tool use
Long-running agents that call tools, browse, and act on their own.
Claude Opus 5.5Math & science
Formal math, competitive problems, and quantitative science.
Claude Opus 5.5Vision & documents
Reading images, screenshots, charts, and dense documents.
Claude Opus 5.5Long documents
Whole codebases, books, and transcripts that fill the context window.
Nemotron 3.5 Super VL (120B A12B)Writing & drafting
Copy, emails, and long-form drafting where tone matters.
Claude Opus 5.5High-volume & cost-sensitive
Cheap, repetitive calls where the blended token price dominates.
Nemotron 3.5 Super VL (120B A12B)Scores are normalized from public benchmarks published by llm-stats.com and averaged per workload. Pricing is the blended input/output cost per million tokens at an 8:1 mix. Refreshes daily. How this works.