Skip to content
CostPerPrompt

Nemotron 3 Ultra API Pricing

NVIDIA · context window 262K · prices updated 2026-11-22

Input per 1M tokens
$0.5
Output per 1M tokens
$2.20
Cached input per 1M
$0.1
80% cheaper than fresh input

Where Nemotron 3 Ultra sits on price

At $2.20 per million output tokens, Nemotron 3 Ultra is a mid tier model — 1.1× cheaper than the median output price of $2.50, which puts it cheaper than 55% and pricier than 45% of the models we track. It is the most expensive of the 7 NVIDIA models we track. Output costs 4.4× input, the usual spread — trim both, starting with the answer length.

Output tokens per $1
454,545
One full context fill
$0.1311
Cheaper than
55% of tracked models

Cached input is 80% cheaper. On an input-heavy workload you need roughly a 31% cache-hit rate to take 25% off the input line — reachable for chatbots and agents that resend the same system prompt and history.

What real workloads cost on Nemotron 3 Ultra

These three workloads are the ones teams actually run on a mid-priced model — a $2.20/1M model is not bought for the same job as one ten times the price.

Workload Per request Per month
Support chatbot — 500 conversations/day (5K in / 1.4K out) Conversation history re-sent each turn — the classic prompt-caching win. $0.0056 $83.70
Document pipeline — 10K docs/day (3K in / 500 out) Batch-friendly: no user waiting, so a ~50% batch discount usually applies. $0.0026 $780
Small app — 1K requests/day (500 in / 150 out) The baseline most side projects actually run at. $0.0006 $17.40

Model your exact traffic in the API cost calculator — it preloads Nemotron 3 Ultra with caching and batch options.

Nemotron 3 Ultra price history

NVIDIA last cut Nemotron 3 Ultra's output price on Nov 17, 2026 (-8% per output token), from $0.6/$2.40 to $0.5/$2.20 per 1M in/out. That is one of 10 repricings in the 112 days we have tracked it — worth knowing before you hard-code today's rate into a budget.

Date Input /1M Output /1M Output change
Aug 2, 2026 (tracking began) $0.6 $3.60 —
Aug 28, 2026 $0.5 $2.20 -39%
Sep 2, 2026 $0.625 $3.13 +42%
Sep 3, 2026 $0.6 $2.40 -23%
Sep 4, 2026 $0.625 $3.13 +30%
Sep 14, 2026 $0.6 $2.40 -23%
Sep 16, 2026 $0.625 $3.13 +30%
Sep 19, 2026 $0.6 $2.40 -23%
Oct 2, 2026 $0.5 $2.20 -8%
Oct 11, 2026 $0.6 $2.40 +9%
Nov 17, 2026 $0.5 $2.20 -8%

Change-points from our daily price snapshots (tracking since Aug 2, 2026; intraday moves between snapshots are not captured).

Nemotron 3 Ultra vs Mistral Medium 3.1

The closest-priced alternative from another vendor is Mistral Medium 3.1 (Mistral) — priced within a rounding error on output, with $0.06 less per million input tokens. When two models land this close on price, the decision is quality on your own workload, not the price sheet: run 50 real requests through both and compare.

See Mistral Medium 3.1 pricing →

Cheaper alternatives

More NVIDIA models

Frequently asked questions

› How much does the Nemotron 3 Ultra API cost?

Nemotron 3 Ultra costs $0.5 per million input tokens and $2.20 per million output tokens, with cached input at $0.1 per million (80% cheaper). That works out to roughly 454,545 output tokens per dollar.

› What does the support chatbot workload cost on Nemotron 3 Ultra?

Support chatbot — 500 conversations/day (5K in / 1.4K out) costs about $0.0056 per request and $83.70 per month on Nemotron 3 Ultra. Conversation history re-sent each turn — the classic prompt-caching win.

› What does it cost to fill Nemotron 3 Ultra's 262K context window?

Sending 262K of input in a single request costs $0.1311 at $0.5 per million tokens — before any output. With prompt caching that same fill drops to about $0.0262 on repeat requests. This is why large context windows are cheap to advertise and expensive to actually use.

› Is Nemotron 3 Ultra worth the price?

Nemotron 3 Ultra sits in the middle of the market (7 of 7 by price within NVIDIA). The honest test is a routing experiment: send the same 200 real requests to this model and to a tier below, and compare failure rate against the price gap — most teams find a majority of traffic never needed the pricier model.