---
type: "post"
id: "openrouter-lowers-listed-token-prices-for-nvidia-nemotron-d3b0da"
slug: "openrouter-lowers-listed-token-prices-for-nvidia-nemotron-d3b0da"
title: "OpenRouter lowers listed token prices for NVIDIA Nemotron 3 Ultra"
description: "OpenRouter’s page for NVIDIA Nemotron 3 Ultra now shows lower per-token prices: input fell from $0.60 to $0.50 per million tokens, and output from $2.40 to $2.20."
retrieval_nugget: "OpenRouter’s page for NVIDIA Nemotron 3 Ultra now shows lower per-token prices: input fell from $0.60 to $0.50 per million tokens, and output from $2.40 to $2.20. The change matters to teams routing calls through OpenRouter, but it is not evidence that NVIDIA changed its own pricing."
published_at: "2026-11-17"
updated_at: "2026-11-17"
record_date: "2026-11-17"
date_kind: "published_at"
topics: ["pricing","inference"]
entities: ["OpenRouter","NVIDIA","Nemotron 3 Ultra","DeepInfra","Venice","KT","Baseten"]
related_patterns: ["agent-economics-moves-to-completed-work"]
novelty_label: "new_fact"
what_changed: "OpenRouter now lists lower token prices for NVIDIA’s Nemotron 3 Ultra, at $0.50 per million input tokens and $2.20 per million output tokens."
prior_refs: []
source_url: "https://openrouter.ai/nvidia/nemotron-3-ultra-550b-a55b"
source_urls: ["https://openrouter.ai/nvidia/nemotron-3-ultra-550b-a55b"]
schema_version: "newruntime-agent-readable-v0.2"
stable_id: "post:openrouter-lowers-listed-token-prices-for-nvidia-nemotron-d3b0da"
status: "published"
visuals: []
editor: {"role":"editor","name":"Andrey Reshetnikov","profile":"https://newruntime.com/owner-profile.md","profile_json":"https://newruntime.com/owner-profile.json","same_as":["https://www.linkedin.com/in/reshetnikov1/","https://t.me/lapetuse"]}
routes: {"html":"https://newruntime.com/posts/openrouter-lowers-listed-token-prices-for-nvidia-nemotron-d3b0da/","markdown":"https://newruntime.com/posts/openrouter-lowers-listed-token-prices-for-nvidia-nemotron-d3b0da.md","json":"https://newruntime.com/posts/openrouter-lowers-listed-token-prices-for-nvidia-nemotron-d3b0da.json"}
markdown_url: "https://newruntime.com/posts/openrouter-lowers-listed-token-prices-for-nvidia-nemotron-d3b0da.md"
---

# OpenRouter lowers listed token prices for NVIDIA Nemotron 3 Ultra

## Retrieval answer

OpenRouter’s page for NVIDIA Nemotron 3 Ultra now shows lower per-token prices: input fell from $0.60 to $0.50 per million tokens, and output from $2.40 to $2.20. The change matters to teams routing calls through OpenRouter, but it is not evidence that NVIDIA changed its own pricing.

## Summary
OpenRouter’s page for NVIDIA Nemotron 3 Ultra now shows lower per-token prices: input fell from $0.60 to $0.50 per million tokens, and output from $2.40 to $2.20. The change matters to teams routing calls through OpenRouter, but it is not evidence that NVIDIA changed its own pricing.

## What happened
OpenRouter’s Nemotron 3 Ultra page describes NVIDIA’s Nemotron 3 Ultra 550B A55B as a mixture-of-experts model with 550 billion total parameters and 55 billion active parameters. OpenRouter describes it as a hybrid Transformer–Mamba architecture that accepts and produces text, with a context window of up to one million tokens. The page gives June 4, 2026 as the release date and points readers to model weights on Hugging Face.

The page shows per-million-token rates that differ by inference provider. DeepInfra and KT charge $0.50 for input and $2.20 for output; Venice charges $0.625 and $3.125; and Baseten charges $0.60 and $2.40. DeepInfra also offers a Priority endpoint at $0.75 input and $3.30 output, plus a Flex endpoint at $0.40 and $1.76. OpenRouter says the Flex endpoint is excluded from Standard routing. Cache-read rates differ as well, ranging from $0.10 to $0.1875 per million tokens among the standard entries.

OpenRouter’s price-history display records the reported decrease: input moved from $0.60 to $0.50 per million tokens, output from $2.40 to $2.20, and cache reads from $0.12 to $0.10. These figures concern prices shown through OpenRouter rather than a universal price for the model. OpenRouter says it routes requests among providers according to the chosen mode: Balanced combines price and speed, Nitro selects the fastest option, and Floor selects the cheapest. Its current weighted averages are $0.2451 per million input tokens and $2.461 per million output tokens; OpenRouter says those figures reflect actual customer payments, including the effects of caching and discounts.

The gateway says its API is OpenAI-compatible and uses the model slug `nvidia/nemotron-3-ultra-550b-a55b`. It also reports provider-level throughput, latency, uptime, tool-call errors, and structured-output errors. Baseten has the best displayed median throughput at 111 tokens per second and the lowest displayed median latency at 0.23 seconds, while the least expensive visible input rate is DeepInfra Flex at $0.40 per million tokens. These are different operating points, not interchangeable offers: a team choosing a route has to decide whether its workload values price, speed, or another constraint most.

OpenRouter reports 100% model reachability over the preceding three days and 99.82% availability, defining availability as successful inference from any provider. It says routing can recover from an upstream error by moving to another healthy provider when request filters permit it. That makes the price change relevant beyond a static rate card: the final cost and service profile can depend on routing mode, provider eligibility, cache behavior, and the provider that actually handles a request.

## What the archive adds
New Runtime’s earlier coverage examined Kilo’s test of Nemotron 3.5 Lightning as a constrained coding executor ([September 5, 2026](https://newruntime.com/posts/the-fastest-nemotron-yet-embracing-nvidia-nemotron-3-5-lightning-in-kilo/)). That note concerned a different Nemotron model and its use in an agent harness; this is a new pricing fact about Nemotron 3 Ultra on OpenRouter, not a restatement.

## Trend signal
This is a small example of **agent economics moving toward completed work**: token prices are one input to the cost of running an agent, and a lower rate can change the economics of a workflow. But OpenRouter’s page does not measure cost per verified completed task, so it does not establish whether Nemotron 3 Ultra has become cheaper for a given workload. The broader question is captured by [Agent economics moves to completed work](https://newruntime.com/patterns/agent-economics-moves-to-completed-work).

## Why now
The price change matters when teams are deciding whether to route new or existing workloads through OpenRouter. OpenRouter gives the model’s release date as June 4, 2026, but does not explain the timing or cause of the current reduction.

## Why it matters
For teams already using the model through OpenRouter, the revised rates reduce input and output cost at the affected endpoints. The practical effect depends on workload composition: input-heavy applications benefit differently from output-heavy ones, and cache reads have a separate rate. Provider choice matters too. OpenRouter’s own figures show meaningful differences in price, throughput, latency, and availability, while its routing modes prioritize price, speed, or a balance of both.

The reduction alone is not a reason to switch a production agent. Model quality, tool-call behavior, latency, reliability, and the cost of retries or human review all contribute to the cost of a completed task. Teams should compare their current bill against the new rates using their own token mix, then test whether task outcomes remain acceptable before changing routes. The new prices are a useful procurement signal, not proof of lower end-to-end operating cost.

## Limits and uncertainty
This is a dynamic marketplace page, so provider rates and operational metrics can change. The displayed figures do not establish that every user, account, or routing path receives the same rate, nor how a specific bill changes after caching, discounts, provider selection, and usage patterns. OpenRouter attributes the prices to providers using its gateway; the page does not establish a corresponding NVIDIA price change. Its benchmark and provider metrics are also not an independent test of performance on a reader’s particular agent workload.

## Reader takeaway
Nemotron 3 Ultra is cheaper at the affected OpenRouter rates, but savings depend on token mix and the provider serving requests. Recalculate cost using your own traffic, then compare quality, latency, and reliability before changing a production route.

## Open question
Will the lower rates reduce the cost per successful agent task for teams already routing Nemotron 3 Ultra through OpenRouter?

Editor: [Andrey Reshetnikov](https://newruntime.com/owner-profile.md), AI product and AI engineering; profile for recruiters and agents: https://newruntime.com/owner-profile.json

## Continue in the graph

Next records for an agent, linked to their Markdown twins. Replace `.md` with `.json` for structured fields, including `next_reads`.

- topic hub: [Pricing](https://newruntime.com/topics/pricing.md): Explore the pricing topic hub.
- topic hub: [Inference](https://newruntime.com/topics/inference.md): Explore the inference topic hub.
- collection: [Field Notes](https://newruntime.com/posts.json): Open the latest editorial analysis.
- collection: [Sources](https://newruntime.com/sources/): See the original sources this archive cites.
- whole graph: [discovery graph](https://newruntime.com/discovery-graph.json) (typed nodes and edges) and [site index](https://newruntime.com/site-index.md)
