NVIDIA Nemotron 3 Super is a 120B-parameter open hybrid MoE model, activating just 12B parameters for maximum compute efficiency and accuracy in complex multi-agent applications. Built on a hybrid Mamba-Transformer Mixture-of-Experts archit…
Model details →NVIDIA: Nemotron 3 Super vs Qwen: Qwen3 235B A22B Instruct 2507
Qwen: Qwen3 235B A22B Instruct 2507 wins on 2 of 6 axes — pricing and capability skew in its favour for most workloads.
Qwen3-235B-A22B-Instruct-2507 is a multilingual, instruction-tuned mixture-of-experts language model based on the Qwen3-235B architecture, with 22B active parameters per forward pass. It is optimized for general-purpose text generation, inc…
Model details →Side-by-side comparison
| Capability | NVIDIA: Nemotron 3 Super | Qwen: Qwen3 235B A22B Instruct 2507 | Winner |
|---|---|---|---|
| Context window Maximum number of input tokens the model can attend to in a single request. | 262K | 262K | Tie |
| Input price (per 1M) Cost per million input tokens billed by the provider. | $0.10 | $0.07 | 🏆 Qwen: Qwen3 235B A22B Instruct 2507 |
| Output price (per 1M) Cost per million output tokens billed by the provider. | $0.50 | $0.10 | 🏆 Qwen: Qwen3 235B A22B Instruct 2507 |
| Tool / function calling First-class support for emitting structured tool calls. | Yes | Yes | Tie |
| Vision input Accepts image inputs alongside text. | — | — | Tie |
| Reasoning mode Internal chain-of-thought / extended-thinking support. | Yes | Yes | Tie |
Frequently asked questions
Is NVIDIA: Nemotron 3 Super better than Qwen: Qwen3 235B A22B Instruct 2507?
Qwen: Qwen3 235B A22B Instruct 2507 wins on 2 of 6 axes — pricing and capability skew in its favour for most workloads.
What's the price difference between NVIDIA: Nemotron 3 Super and Qwen: Qwen3 235B A22B Instruct 2507?
Input: $0.10 vs $0.07 per 1M tokens. Output: $0.50 vs $0.10 per 1M tokens.
What context windows do NVIDIA: Nemotron 3 Super and Qwen: Qwen3 235B A22B Instruct 2507 support?
NVIDIA: Nemotron 3 Super supports up to 262K tokens. Qwen: Qwen3 235B A22B Instruct 2507 supports up to 262K tokens.
Do both NVIDIA: Nemotron 3 Super and Qwen: Qwen3 235B A22B Instruct 2507 support tool calling?
NVIDIA: Nemotron 3 Super: yes. Qwen: Qwen3 235B A22B Instruct 2507: yes.