Side-by-side comparison of pricing, 12 benchmarks, and generation speed.
| Metric | Nemotron 3 Super 120B A12B (Reasoning) | Qwen3 32B (Non-reasoning) |
|---|---|---|
| Input ($/M tokens) | $0.19 | $0.16 |
| Output ($/M tokens) | $0.65 | $0.64 |
Verdict. Nemotron 3 Super 120B A12B (Reasoning) wins the overall benchmark matchup 2–0 across 2 overlapping categories, but raw benchmark score is only one input to the decision.
Pricing. Both models sit in the budget bracket for output-token pricing. At 1.0× the per-million-token cost, Qwen3 32B (Non-reasoning) is meaningfully cheaper if your traffic is output-heavy (long completions, document generation, agent loops). Qwen3 32B (Non-reasoning) makes more sense when output volume is low and absolute reasoning quality justifies the premium.
Strengths. Nemotron 3 Super 120B A12B (Reasoning) is strongest on Coding Index (37.7), Intelligence Index (13.6). Qwen3 32B (Non-reasoning) leads on Intelligence Index (7.3).
Speed. On throughput, Nemotron 3 Super 120B A12B (Reasoning) generates tokens at 173 tok/s versus 106 tok/s — about 39% faster. On time-to-first-token, Nemotron 3 Super 120B A12B (Reasoning) responds in 2390ms vs 2460ms, which matters most for chat-style UIs.
Provider. NVIDIA and Alibaba sell to overlapping but distinct developer audiences: NVIDIA tends to ship frontier reasoning models with premium positioning, while Alibaba often prices more aggressively. Your existing vendor relationships, billing, and SLA preferences may matter as much as the raw numbers above.
Workload cost. Workload scenarios (per million requests at 30M input + 15M output tokens): Nemotron 3 Super 120B A12B (Reasoning) costs $15.45 ($185/year); Qwen3 32B (Non-reasoning) costs $14.40 ($173/year). At a smaller 5M-input/2M-output scale (single-developer tool or prototype): Nemotron 3 Super 120B A12B (Reasoning) ≈ $2.25/run, Qwen3 32B (Non-reasoning) ≈ $2.08/run. At agent/realtime scale (200M input / 100M output per million requests): Nemotron 3 Super 120B A12B (Reasoning) ≈ $103/run, Qwen3 32B (Non-reasoning) ≈ $96/run. Qwen3 32B (Non-reasoning) becomes more attractive at higher volume — the absolute per-token pricing difference compounds when you ship at scale.
Recommendation. Both models have legitimate use cases — the right answer depends on whether you are optimizing for benchmark ceiling, latency, or unit cost. Start with the cheaper / faster model, evaluate against your specific task, and only switch if the upgrade shows a meaningful lift.
Head-to-head deltas
Data from Artificial Analysis API — 12 benchmarks
Qwen3 32B (Non-reasoning) is cheaper overall. Its blended price (3:1 input/output ratio) is $0.28/M tokens vs $0.30/M for Nemotron 3 Super 120B A12B (Reasoning).
Nemotron 3 Super 120B A12B (Reasoning) wins 2 out of 12 benchmarks compared to 0 for Qwen3 32B (Non-reasoning). See the detailed benchmark chart above for per-category results.
Nemotron 3 Super 120B A12B (Reasoning) generates tokens faster at 173 tok/s vs 106 tok/s. Nemotron 3 Super 120B A12B (Reasoning) also has lower time-to-first-token (2.39s vs 2.46s).
Choose based on your priorities: Qwen3 32B (Non-reasoning) for lower cost, Nemotron 3 Super 120B A12B (Reasoning) for stronger benchmark performance, and Nemotron 3 Super 120B A12B (Reasoning) for faster generation. For latency-sensitive apps, check the TTFT comparison above.