// head_to_head
Hermes 3 405B Instruct vs Nemotron 3 Ultra
Published LiveBench scores across all seven categories, live list pricing, context windows, and the measured cost of a point of capability — for both models, side by side.
Hermes 3 405B Instruct and Nemotron 3 Ultra are priced within ~10% of each other.
Hermes 3 405B Instruct does not have a published LiveBench run, so this comparison covers price, context and declared capabilities only. A missing score means "not evaluated", not "worse" — the right way to separate these two is an eval on your own workload.
nousresearch
Hermes 3 405B Instruct
- Blended / 1M
- $1.00
- Context
- 131K
- Released
- Aug 16, 2024
- Overall score
- Not evaluated
nvidia
Nemotron 3 Ultra
- Blended / 1M
- $1.05
- Context
- 262K
- Released
- Jun 4, 2026
- Overall score
- 67.4
Specs and pricing
| Metric | Hermes 3 405B Instruct | Nemotron 3 Ultra |
|---|---|---|
| LiveBench overall Mean of the seven LiveBench category scores, 0–100. Higher is better. | — | 67.4 |
| Cost per point Measured benchmark spend divided by overall score — dollars per point of capability. | — | $0.2118 |
| Blended price / 1M 3:1 input:output mix, the usual shape of production traffic. | $1.00 | $1.05 |
| Input price / 1M | $1.00 | $0.600win |
| Output price / 1M | $1.00win | $2.40 |
| Cached input / 1M Price of an input token served from the prompt cache, where the provider publishes one. | — | $0.120 |
| Context window | 131K | 262Kwin |
| Max output tokens | 16K | 183Kwin |
Benchmarks by category
An overall score averages away the thing you probably care about. These are the seven LiveBench categories scored separately — Hermes 3 405B Instruct on top, Nemotron 3 Ultra below, both out of 100.
What each one costs to run
Per-token prices are hard to feel. These are monthly list costs for both models across five workload shapes, using each provider's published cached-input rate where there is one.
| Workload | Hermes 3 405B Instruct | Nemotron 3 Ultra |
|---|---|---|
| Support chatbot 1.2K in / 400 out × 200K requests | $320.00/mo | $301.44/mo |
| RAG assistant 8K in / 600 out × 100K requests | $860.00/mo | $432.00/mo |
| Coding agent 40K in / 4K out × 20K requests | $880.00/mo | $403.20/mo |
| Document extraction 20K in / 1.5K out × 50K requests | $1,075/mo | $756.00/mo |
| Bulk classification 500 in / 20 out × 5M requests | $2,600/mo | $1,500/mo |
Which should you pick?
You need to fit large documents in one call
Nemotron 3 Ultra
Wider context window — 262K against 131K.
Hermes 3 405B Instruct vs Nemotron 3 Ultra FAQ
Which is better, Hermes 3 405B Instruct or Nemotron 3 Ultra?
Hermes 3 405B Instruct and Nemotron 3 Ultra are priced within ~10% of each other. Hermes 3 405B Instruct does not have a published LiveBench run, so this comparison covers price, context and declared capabilities only. A missing score means "not evaluated", not "worse" — the right way to separate these two is an eval on your own workload.
Is Hermes 3 405B Instruct cheaper than Nemotron 3 Ultra?
They cost about the same. Both land near $1.00 per million tokens on a 3:1 input:output blend, so price is unlikely to be the deciding factor between them.
Does Hermes 3 405B Instruct or Nemotron 3 Ultra have a bigger context window?
Nemotron 3 Ultra has the larger context window: 131K for Hermes 3 405B Instruct against 262K for Nemotron 3 Ultra. Note that a window you can fill is not a window you should fill — retrieval quality usually degrades well before the limit, and you pay for every token you put in it.
Do Hermes 3 405B Instruct and Nemotron 3 Ultra support prompt caching?
Nemotron 3 Ultra publishes a cached-input rate of $0.120 per million tokens against a full input rate of $0.600. The catalogue lists no separate cached rate for Hermes 3 405B Instruct, which means the provider does not price it separately here — not that caching is unavailable.
Related comparisons
How these numbers are produced
- Price — provider list price from OpenRouter, refreshed every 15 minutes. “Blended” is a 3:1 input:output mix.
- Scores — LiveBench release 2026-06-25, using their own category map. Each model shows its strongest published run. A blank means “not evaluated”, never “bad”.
- Cost per point — the measured dollars LiveBench spent on the run, divided by the score it earned.
- “Win” — awarded only past a threshold: one full point on a benchmark score, 10% on a price, 25% on a context window. Anything tighter reports as a tie, because effort settings alone move a LiveBench score by more than that.
Published benchmarks rank models on someone else's tasks. Before committing, see LLM & agent evaluation for building an eval on your own.