Amazon Nova Lite is Amazon's low-cost multimodal model on Bedrock, and inside Atlas it is the cheap model with a surprisingly large read window. Amazon Nova Lite carries a 300K token context, the largest of the original Nova line, at $0.06 per Mtok input and $0.24 per Mtok output. It is billed and authorized entirely through AWS IAM rather than a separate API key, so there is no new vendor contract to sign. The tradeoff is an 8,192 token max output and no reasoning mode.
What is Amazon Nova Lite best at inside Atlas?
Amazon Nova Lite is best at cheap wide reads inside Atlas. Its 300K token context at $0.06 per Mtok input is more than 2x the 128,000 window of Nova Micro for very little extra cost, which makes Nova Lite the default Bedrock model for background work that still needs real context.
Atlas searches code with hybrid semantic and keyword retrieval fused by reciprocal rank fusion, and Atlas indexes code by AST declarations using tree-sitter, not blind line windows. A retrieval pass over a large repository produces a lot of whole-declaration chunks, and Amazon Nova Lite's 300K token context swallows them at a price where you stop counting. Atlas's documented configuration puts it in the background slot: set "small_model": "amazon-bedrock/amazon.nova-lite-v1:0" in atlas.json for cheap background calls. Atlas fans out work to subagents that can run in the foreground or in parallel background sessions, and Nova Lite at $0.06 per Mtok input is priced to absorb that fan-out without the bill becoming a topic of conversation.
How does Amazon Nova Lite pricing compare to Nova Pro and Nova Micro?
Amazon Nova Lite runs $0.06 per Mtok input, roughly 13x cheaper on input than Nova Pro's $0.8 per Mtok while keeping the same 300,000 token window. Against Nova Micro's 128,000 token context, Nova Lite gives more than 2x the window for very little extra cost.
Amazon Nova Lite sits in the sweet spot of the original Nova line on the context-per-dollar axis. Nova Pro charges $0.8 per Mtok input for the same 300,000 token window that Nova Lite provides at $0.06. Nova Micro is cheaper still, but its 128,000 token context is less than half of Nova Lite's 300K. Output on Nova Lite is $0.24 per Mtok. For an Atlas workload that is retrieval-heavy and generation-light, which describes most background traffic, Nova Lite's ratio is the one that matters: you pay almost nothing to read a great deal. Amazon Nova Lite is also multimodal, so screenshots and diagrams are available to it, unlike text-only Nova Micro.
How does IAM billing change the Atlas setup?
Amazon Nova Lite at $0.06 per Mtok input is billed through your existing AWS account and authorized by IAM, so there is no new vendor contract or API key to manage. Atlas accepts AWS_PROFILE, or AWS_ACCESS_KEY_ID plus AWS_SECRET_ACCESS_KEY, plus AWS_REGION, and container or web identity credentials also work.
For teams already on AWS, the IAM path removes the two slowest steps in adopting a model: procurement and secret distribution. Amazon Nova Lite runs under credentials the organization already issues and audits. Atlas also supports AWS_BEARER_TOKEN_BEDROCK for Bedrock, alongside the standard AWS_PROFILE and AWS_ACCESS_KEY_ID plus AWS_SECRET_ACCESS_KEY plus AWS_REGION combinations. Before Nova Lite will answer, enable Amazon Nova Lite in the Bedrock model access page for your region, because Bedrock gates models per region per account. Then run `atlas models amazon-bedrock` and confirm amazon.nova-lite-v1:0 resolved. If the id is missing, the model access request has not been approved for that region yet.
Why do Bedrock model id prefixes matter for Amazon Nova Lite?
Bedrock ids for Amazon Nova Lite can carry us., eu., or global. prefixes, and regional pricing differs, so match the id to your AWS_REGION. A mismatched prefix is the most common reason amazon.nova-lite-v1:0 fails to resolve or bills at an unexpected rate in Atlas.
Amazon Nova Lite is exposed on Bedrock under inference profile ids that encode geography. The base id is amazon.nova-lite-v1:0, but the id your account can actually invoke may be prefixed for a region group, and regional pricing differs from the $0.06 per Mtok input and $0.24 per Mtok output headline. The fix is mechanical: run `atlas models amazon-bedrock`, read the id that resolved, and set exactly that id in atlas.json against the AWS_REGION you configured. Do not copy an id from a document written for a different region. Atlas will happily call whatever id you pin, and Bedrock will happily bill you at whatever rate that region charges.
When should you pick a different model than Amazon Nova Lite?
Pick a different model than Amazon Nova Lite when the task needs reasoning or a big generation. Amazon Nova Lite has an 8,192 token max output and no reasoning mode, which shows on multi step debugging where a reasoning model would recover. Nova Lite belongs in the small_model slot.
Amazon Nova Lite's 8,192 token max output means large refactors have to be chunked across multiple turns, and its lack of a reasoning mode means multi step debugging goes sideways in ways a reasoning model would catch and correct. Neither is a defect in a model designed for cheap high-volume calls. Both are disqualifying for a build agent. The right structure is to keep Nova Lite as "small_model" for cheap background calls and give "model" to something with reasoning. If your constraint is purely cost and you can live with 128,000 tokens instead of 300K, Nova Micro is cheaper still. Atlas lets you switch the active model and provider on the fly with favorites and recents, so testing the ladder is a keystroke.
Setup
- 01Configure AWS auth: AWS_PROFILE, or AWS_ACCESS_KEY_ID plus AWS_SECRET_ACCESS_KEY, plus AWS_REGION (AWS_BEARER_TOKEN_BEDROCK and container or web identity creds also work).
- 02Enable Amazon Nova Lite in the Bedrock model access page for your region.
- 03Run `atlas models amazon-bedrock` and confirm amazon.nova-lite-v1:0 resolved.
- 04Set "small_model": "amazon-bedrock/amazon.nova-lite-v1:0" in atlas.json for cheap background calls.
- 05Note that Bedrock ids can carry us., eu., or global. prefixes and regional pricing differs, so match the id to your AWS_REGION.
Frequently asked questions
- how much does Amazon Nova Lite cost per million tokens
- Amazon Nova Lite costs $0.06 per Mtok input and $0.24 per Mtok output. Regional pricing differs on Bedrock, so match the model id to your AWS_REGION.
- what is Amazon Nova Lite's context window
- Amazon Nova Lite has a 300K token context, the largest of the original Nova line and more than 2x Nova Micro's 128,000. Its max output is 8,192 tokens.
- how do I use Amazon Nova Lite with a terminal coding agent
- Configure AWS auth with AWS_PROFILE or AWS_ACCESS_KEY_ID plus AWS_SECRET_ACCESS_KEY plus AWS_REGION, enable Nova Lite in Bedrock model access, then set "small_model": "amazon-bedrock/amazon.nova-lite-v1:0" in atlas.json.
- Amazon Nova Lite vs Nova Pro pricing
- Amazon Nova Lite is roughly 13x cheaper on input than Nova Pro's $0.8 per Mtok, at $0.06 per Mtok, while keeping the same 300,000 token context window.
- do I need an API key for Amazon Nova Lite
- No. Amazon Nova Lite is billed through your existing AWS account and authorized by IAM, so there is no new vendor contract or API key to manage.
- why is amazon.nova-lite-v1:0 not resolving in Atlas
- Either Nova Lite is not enabled in the Bedrock model access page for your region, or the id prefix is wrong. Bedrock ids can carry us., eu., or global. prefixes.
- is Amazon Nova Lite good for debugging
- Not for hard cases. Amazon Nova Lite has no reasoning mode, which shows on multi step debugging where a reasoning model would recover. Keep it in the small_model slot.
Try Atlas in your terminal
The terminal-native AI coding agent. Free core, single binary.
Install AtlasRelated guides
Atlas vs Greptile: Terminal AI Coding Agents in 2026
Comparing Atlas and Greptile in 2026. Atlas offers terminal-native AI coding with permission-gated tools. Greptile reviews code with sandbox execution, catching 20% more bugs.
Atlas for Django in 2026
Atlas, the terminal-native AI coding agent, empowers Django developers in 2026. Boost productivity across models, views, and migrations with secure, reviewable AI assistance.
Atlas for R: A Terminal-Native AI Coding Agent for tidyverse, roxygen2, and testthat in 2026
Atlas is a terminal-native AI coding agent for R in 2026. It reads roxygen2 docblocks and renv.lock, rewrites loops as dplyr or purrr pipelines, and runs devtools::test().
Atlas for PHP in 2026
Atlas, the terminal-native AI coding agent, empowers PHP developers in 2026 with intelligent code understanding, secure workflows, and direct integration for Composer and PSR standards.
Atlas for Elixir in 2026
Adopt Atlas, the terminal-native AI coding agent, for Elixir development in 2026. Enhance productivity with deep code understanding, safety features, and direct integration into mix projects and OTP applications.
Atlas for Phoenix in 2026
Atlas is a terminal-native AI coding agent for Phoenix in 2026. It reads contexts, LiveView modules, and Ecto changesets, then runs mix test behind a prompt.
Atlas for Astro: Islands, Content Collections, and Zero JS by Default in 2026
Atlas is a terminal-native AI coding agent for Astro in 2026. It reads astro.config.mjs, src/pages, and content collection schemas, drops needless client:load directives, and runs astro check.
Atlas for Electron: Terminal-Native AI Coding for Main, Preload, and Renderer in 2026
Atlas is a terminal-native AI coding agent for Electron in 2026, where the main and renderer split, contextIsolation, and preload bridges are the security model.