Nemotron 3 Super 120B A12B (Reasoning) vs Qwen3 235B A22B 2507 (Reasoning)
Which is better for the money: Nemotron 3 Super 120B A12B (Reasoning) by NVIDIA or Qwen3 235B A22B 2507 (Reasoning) by Alibaba? Intelligence, total benchmark cost, and end-to-end time from Artificial Analysis & other sources, snapshot 2026-10-27.
Depends on your budget
Nemotron 3 Super 120B A12B (Reasoning) scores 0.1 points higher (12.8 vs 12.7) but costs 5.6× as much to run the benchmark ($1319 vs $234; premium vs mid-range cost classes). They are not alternatives at one budget: pick Nemotron 3 Super 120B A12B (Reasoning) when you need intelligence above 12.7, and Qwen3 235B A22B 2507 (Reasoning) when 12.7 is enough.
| Metric | Nemotron 3 Super 120B A12B (Reasoning) | Qwen3 235B A22B 2507 (Reasoning) | Difference |
|---|---|---|---|
| Intelligence (higher is better) | 12.8 | 12.7 | −0.8% Qwen3 235B A22B 2507 … is 0.8% less intelligent |
| Total AA cost (lower is better) | $1,318.62 | $233.56 | −82.3% Qwen3 235B A22B 2507 … is 82.3% cheaper |
| End-to-end time (lower is better) | 17.13s | 21.63s | +26.3% Qwen3 235B A22B 2507 … is 26.3% slower |
| Price per 1M tokens (in / out) | $0.30 / $0.90 | $0.23 / $2.30 | |
| Input modalities | Text | Text |
Qwen3 235B A22B 2507 (Reasoning) vs Nemotron 3 Super 120B A12B (Reasoning): 0.8% less intelligence, 82.3% less cost to run the AA benchmark, and 26.3% slower end-to-end.
Against the frontier
- Nemotron 3 Super 120B A12B (Reasoning) is off the frontier: at the same or lower cost, GPT-6.1 Sol (Max) scores 51.80 for $1082.
- Qwen3 235B A22B 2507 (Reasoning) is off the frontier: at the same or lower cost, MiMo-V2.6-Pro scores 46.30 for $207.
Compare something else
Accept: text/markdown or ?format=md returns this page as Markdown. JSON: /api/v1/compare/nvidia-nemotron-3-super-120b-a12b/qwen3-235b-a22b-instruct-2507-reasoning · Markdown: /api/v1/compare/nvidia-nemotron-3-super-120b-a12b/qwen3-235b-a22b-instruct-2507-reasoning?format=md · llms.txt · agent skill