NewChat Models

NVIDIA

Nemotron 3.5 Lightning

Efficiency-focused Nemotron model for low-cost text workflows and rapid iteration, using a paid 256K OpenRouter route.

Efficient256K contextText
Open AppBrowse All Models

At a glance

Know the model before you prompt.

OpenRouter API specifications
Context window
262,144 tokens
Maximum output
Provider-dependent
Inputs → output
Text → Text
Knowledge cutoff
Not verified

API model ID: nvidia/nemotron-3.5-lightning

Capabilities & boundaries

What it supports. Where the limits are.

Tool support requires the appropriate API integration; a supported tool is not automatically active in every chat.

Supported API features and tools

  • NVIDIA’s efficiency-oriented 30B-total, 3B-active model, with architecture and serving optimizations described in its card
  • Text transformation, coding assistance and instruction-following on a 256K OpenRouter catalog route
  • Reasoning support through compatible hosted endpoints
  • Tool requests and structured-output support listed by the reviewed OpenRouter providers

Before you choose

  • No fixed tokens-per-second guarantee is established here. Queue time, output length, reasoning and provider selection affect end-to-end speed.
  • The official card separates pre-training through September 2025 from post-training through May 2026. Supply current sources rather than treating these training dates as current web knowledge.
  • The OpenRouter catalog reference is 262,144 tokens (256K). NVIDIA’s underlying model card describes up to 1M in suitable deployments; that is not the app’s confirmed route limit. Individual hosts can impose smaller windows. Leave room for the reply and reasoning.
  • This route accepts text and returns text, not image, audio or video input/output. Provide extracted text for document work; a prompt cannot add vision support.
  • Tool calls and structured outputs require a compatible host and integration. The model does not independently browse the web, execute code or access private files.
  • Answers, extracted fields and generated code can be wrong. Supply evidence and acceptance criteria, check cited passages and test code before use. A long context does not guarantee complete recall.

Extended thinking

The reviewed OpenRouter metadata supports reasoning, but does not establish a universal default or named effort scale for this route. NVIDIA’s underlying model supports reasoning on/off via its self-hosted chat template. Confirm the hosting API’s accepted controls before assuming those flags or another Nemotron model’s effort levels apply.

  • Use nvidia/nemotron-3.5-lightning exactly; the decimal point belongs in this paid API ID even though the website URL uses hyphens.
  • io.net lists 262,144 context and 131,072 maximum output tokens. DeepInfra lists the same context with 32,768 output, while other hosts publish different ceilings.
  • Compare hosts using both input and output costs, accepted structured-output schemas and measured task accuracy. Server-side speed techniques are not user-visible app switches.

Put it to work

Start with a more useful prompt.

Original examples from EZ Ai Assist. Adapt these to your task and the features available in your workspace.

Workflow 01

Normalize product records

Make a repeated transformation deterministic and reviewable.

Normalize these product records to the supplied schema. Preserve IDs exactly, convert only the units with explicit conversion rules, and use null for missing fields. Return one record per input plus a separate validation report for duplicates, ambiguous values and rejected conversions. Do not fill gaps with plausible product facts.

Workflow 02

Draft a minimal parser test set

Use lightweight coding help for a well-defined boundary.

Given this parser and its documented input format, propose a compact test set covering valid cases, empty input, malformed delimiters and unexpected encoding. Include exact input, expected output or error and the reason each test matters. Do not change the implementation or claim any test has been executed.

Workflow 03

Compare compact-model candidates

Measure value using a consistent sample rather than a model label.

Using the supplied sample tasks and candidate outputs, score each result for correctness, schema compliance and unnecessary content. Cite evidence for every deduction and mark unscorable cases. Create a blank table for measured latency and billed tokens. Recommend a decision rule that favors lower cost only after the required quality threshold is met.

Developer reference

OpenRouter API pricing

These are named-provider OpenRouter API reference prices, not EZ Ai Assist subscription prices.

View EZ Ai Assist plans
io.net route reference on OpenRouter · USD per 1,000,000 tokens
Token typePrice
Input$0.0469
Output$0.134
Cached input$0.02345
  • Rates were checked October 8, 2026. The named host is a pricing reference, not confirmation that the app pins that host. Routing, availability, cache eligibility and promotions can change the actual bill.
  • The io.net endpoint currently lists $0.0469 input (about $0.047 rounded) and $0.134 output per million tokens, with $0.02345 cached input. Five-decimal formatting preserves that cache rate.
  • The reviewed io.net listing marks these rates as a 33% discount. Treat them as promotional reference prices, not a permanent promise or proof the app uses that host.
  • Other hosts may be cheaper or more expensive. For example, DeepInfra lists $0.06 input and $0.16 output; evaluate route quality and limits as well as price.
  • For a rough uncached estimate, multiply input tokens by the input rate and output tokens by the output rate, then divide by one million. Inspect the provider’s usage and billing records for reasoning and any additional charges.

Common questions

A few things worth knowing.

Is Lightning the same model as Nemotron 3 Nano?

No. This guide uses the separate nvidia/nemotron-3.5-lightning ID and NVIDIA’s Lightning 3.5 model card. Both are compact Nemotron options, but their versions, active parameter figures, pricing and hosting configurations differ.

Why does the table show $0.0469 instead of $0.047?

The reviewed io.net endpoint publishes $0.0469 per million input tokens; $0.047 is its rounded presentation. The more precise table also preserves the $0.02345 cached-input rate. The listing marks these prices as discounted.

Is it always the fastest or cheapest model?

No. Lightning is an efficiency-focused option, but hosting, queueing, reasoning and answer length affect speed and total cost. Measure representative tasks and compare input, output and applicable cache rates before selecting a route.

Can I request a particular reasoning effort?

The reviewed metadata does not establish one shared named effort scale or default across hosts. Verify the specific endpoint and app controls rather than copying medium/high settings from Ultra or assuming a prompt changes the API configuration.

Are these the app’s confirmed model IDs?

Yes. The app owner confirmed these exact paid OpenRouter IDs, with text-only 256K routes and no :free variants. That confirmation does not test a live request or guarantee that every hosting feature and reasoning control is exposed in your workspace.

Are the token prices EZ Ai Assist subscription prices?

No. These are named-host developer reference rates on OpenRouter. Hosting choices, discounts and supported features can change; your EZ Ai Assist plan and any limits displayed in the app are separate.

Can I use this model to inspect an image or search the web?

Not from the model alone. These routes are text-only. Supply extracted text or use a verified vision model for images. Web search and other tools require an enabled integration; a generated tool request is not proof an action happened.

Check the source

Official documentation

Specifications and API prices checked on . Example prompts and workflow advice are editorial guidance from EZ Ai Assist.

Same provider

More from NVIDIA

NVIDIA

Nemotron 3.5 Super VL

NVIDIA’s vision-language reasoner for text, images and visual documents. The app uses paid OpenRouter hosting; context limits vary by serving host.

NVIDIA

Nemotron 3 Ultra

NVIDIA’s flagship Nemotron reasoner for complex text, code and multi-step analysis, using a paid 256K OpenRouter route.

NVIDIA

Nemotron 3 Super

Balanced Nemotron workhorse for text, coding and tool-assisted workflows, using a paid 256K OpenRouter route.

NVIDIA

Nemotron 3 Nano

Compact Nemotron model for short replies, extraction and bounded reasoning, using a paid 256K OpenRouter route.