NVIDIA Nemotron 3 Super API Prices Rise 253%: What Developers Need to Know

The Numbers

NVIDIA has repriced the Nemotron 3 Super API. Input tokens move from $0.085 to $0.30 per million — a +253% increase. Output tokens rise from $0.40 to $0.90 per million. Both changes are live now.

Does This Actually Hurt?

It depends on your usage pattern. Output-heavy workloads (long generations, document drafting, agentic chains) feel the most pressure: at $0.90/M output tokens, a pipeline burning 10M output tokens a month jumps from $4.00 to $9.00 on that line alone. Input-heavy workloads — long-context retrieval, classification, embeddings workarounds — take the bigger proportional hit given the 253% input increase, even though the absolute output rate is still the steeper cost.

At $0.30 input / $0.90 output, Nemotron 3 Super is no longer priced as a budget option. Teams that chose it specifically for cost efficiency should re-run their per-workload math before the next billing cycle.

What To Do

• Audit your token split. If you're output-heavy, the output rate is the bigger lever.
• Compare alternatives. Several models in a similar capability tier sit below these new rates. Use Nemotron 3 Super — live specs & price history to stack it against current competitors side by side.
• Check caching options. If your provider supports prompt caching, high input costs make cache hits more valuable than before.

The model itself hasn't changed — only the price. Whether the performance justifies $0.30/$0.90 depends entirely on what you're getting from it relative to cheaper alternatives now on the board.

Ezra, Scout AI Team