NVIDIA Nemotron-3 Super 120B at a glance
Good enough on 6/54 tasks at the Balanced step. Best value on 0 tasks. Doesn't qualify on any: Financial Analysis & Trading Decisions, Infrastructure & Utility, Long-form Content Generation, Topic Org
Cost vs quality across all tasks
within ~1.3× of the best-value model · 1.3–2× · >2× · ★ this model is the best-value pick on that task. Top-right = best quadrant. Only tasks where this model qualifies at the Balanced step are plotted.
Per-task breakdown
| Task | Category | Quality (% of best) | Confidence | Overpay |
|---|---|---|---|---|
| Evidence Grounded Claim Generation | Content Summarization & Synthesis | 93% | HIGH | 1.2x |
| Profiled Document Structure Extraction | Structured Data & Fact Extraction | 98% | MEDIUM | 2.7x |
| Language Identification | Relevance, Classification & Matching | 99% | RANKED | 3.7x |
| Promotional Message Generation | Social & Promotional Content | 90% | HIGH | 4x |
| Social Post Portfolio Selection | Relevance, Classification & Matching | 93% | MEDIUM | 6.7x |
| Engagement Opportunity Triage | Relevance, Classification & Matching | 96% | RANKED | 10x |
Overpay — how much more you pay by running this model instead of the best-value model that meets the quality step on that task (marked ★). "16x" means you overpay 16× — the same output for 16× the best-value good-enough option; ★ means this model is that option (no overpayment). Confidence — how sure we are about the quality score (more judgments + more agreement = higher confidence): RANKED many independent judges scored this model's outputs and their agreement is very high (most confident) — HIGH many judges have scored it and they mostly agree (well-pinned) — MEDIUM enough judges have weighed in to publish, but they disagree more than we'd like (treat with a small grain of salt). LOW-confidence cells are hidden everywhere on the site. See the methodology for the exact thresholds.