Amazon Nova Micro is the cheapest model in the Bedrock catalog at $0.035 per Mtok input and $0.14 per Mtok output. It is text only, with a 128,000 token context and an 8,192 token output. Amazon Nova Micro exists to absorb high-volume, low-stakes calls at effectively zero marginal cost, and that is exactly the role it should play in Atlas: the small_model slot, with write tools denied. It has no reasoning mode and a small model size, so it should never own Atlas's build loop.
What is Amazon Nova Micro best at inside Atlas?
Amazon Nova Micro is best at triage and routing inside Atlas. At $0.035 per Mtok input, the lowest price of any model in the Bedrock catalog, Nova Micro exists to absorb high-volume, low-stakes calls at effectively zero marginal cost. Its 128,000 token context still reads a sizable module.
Atlas generates a constant stream of small, unglamorous calls: name this session, summarize this file, decide which subagent should handle this. Atlas fans out work to subagents that can run in the foreground or in parallel background sessions, and every one of them adds to that stream. Amazon Nova Micro is priced so that stream costs effectively nothing. The 128,000 token context is still large enough to read a sizable module for triage and routing, which is the meaningful distinction: Nova Micro is small, but it is not blind. Atlas's documented configuration is exact about where it goes. Put it in "small_model" only: "amazon-bedrock/amazon.nova-micro-v1:0".
How much cheaper is Amazon Nova Micro than Nova Pro?
Amazon Nova Micro runs $0.035 per Mtok input, about 23x cheaper than Nova Pro's $0.8 per Mtok. Output is $0.14 per Mtok. Nova Micro is the lowest price of any model in the Bedrock catalog, which is the entire reason to reach for it in an Atlas configuration.
The 23x gap against Nova Pro is not a marginal saving, it is a different category of spending. A workload that costs a noticeable amount of money on Nova Pro costs a rounding error on Amazon Nova Micro. What the 23x buys back is capability and context: Nova Micro carries 128,000 tokens against the larger windows further up the Nova line, and it is text only. For Atlas's background traffic, neither limit bites. For Atlas's build loop, both do. That asymmetry is why the small_model slot exists as a separate configuration key at all, and Amazon Nova Micro is the most extreme, and most correct, thing to put in it on Bedrock.
Why is Amazon Nova Micro being text only a real constraint?
Amazon Nova Micro is text only, so screenshots and diagrams cannot be used, unlike Nova Lite and Nova Pro. In an Atlas workflow where a developer pastes a screenshot of a failing UI, Amazon Nova Micro cannot see it, no matter how much of its 128,000 token context is free.
Text only is a hard boundary, not a quality gradient. If your Atlas workflow involves handing the agent an image, Amazon Nova Micro is disqualified and Nova Lite or Nova Pro is the answer. Where Nova Micro's text-only nature costs nothing is the work it was built for: reading source code, summarizing a diff, naming a session, routing a query. All of that is text. Atlas indexes code by AST declarations using tree-sitter, not blind line windows, and every chunk that index produces is text Nova Micro can read. So the constraint is real but narrowly scoped. Know whether images are in your loop, and pick accordingly.
How do you keep Amazon Nova Micro away from your source code?
Use Atlas's allow/ask/deny permission config to deny write tools on the Amazon Nova Micro slot. At $0.035 per Mtok input Nova Micro is cheap, but it has no reasoning mode and a small model size, so it should never own Atlas's build loop. A deny rule enforces that.
Every Atlas tool call is permission-gated against allow, ask, and deny rules before it runs, which means the protection is structural rather than a matter of prompting carefully. Denying write tools on the Amazon Nova Micro slot lets it keep doing everything it is good at, reading, summarizing, routing, while making it physically unable to edit a file. Atlas computes a unified diff for every file edit and surfaces it for approval before writing, and Atlas snapshots file changes as git patches so edits can be diffed and rolled back, but neither of those is a reason to let a text-only, non-reasoning model near the write path. Deny is cheaper than review.
How does Amazon Nova Micro fit an AWS compliance boundary?
Amazon Nova Micro runs under the same IAM role and VPC endpoints as the rest of your AWS estate, so no data leaves your compliance boundary. For teams whose blocker on AI coding agents is data egress rather than cost, Nova Micro at $0.035 per Mtok is a low-risk starting point.
Amazon Nova Micro's compliance story is inherited, not bolted on. Set AWS_ACCESS_KEY_ID, AWS_SECRET_ACCESS_KEY, and AWS_REGION, or use AWS_PROFILE with an assumed role, and Nova Micro is governed by whatever IAM policy already governs your other Bedrock traffic. Request access to Amazon Nova Micro in the Bedrock console for that region, then run `atlas models amazon-bedrock` and find amazon.nova-micro-v1:0. On the Atlas side, Atlas can build its code index with local Ollama embeddings, keeping code off third-party servers, so the embedding layer and the model layer can both be kept inside the perimeter. That combination lets a security-conscious team run Atlas's whole background layer without a single byte crossing a boundary they do not control.
Setup
- 01Set AWS_ACCESS_KEY_ID, AWS_SECRET_ACCESS_KEY, and AWS_REGION, or use AWS_PROFILE with an assumed role.
- 02Request access to Amazon Nova Micro in the Bedrock console for that region.
- 03Run `atlas models amazon-bedrock` and find amazon.nova-micro-v1:0.
- 04Put it in "small_model" only: "amazon-bedrock/amazon.nova-micro-v1:0".
- 05Use Atlas's allow/ask/deny permission config to deny write tools on this slot.
Frequently asked questions
- what is the cheapest model on Amazon Bedrock
- Amazon Nova Micro at $0.035 per Mtok input is the lowest price of any model in the Bedrock catalog, about 23x cheaper than Nova Pro's $0.8 per Mtok. Output is $0.14 per Mtok.
- does Amazon Nova Micro support images
- No. Amazon Nova Micro is text only, so screenshots and diagrams cannot be used, unlike Nova Lite and Nova Pro. Its 128,000 token context handles source code fine.
- what is Amazon Nova Micro's context window
- Amazon Nova Micro has a 128,000 token context and an 8,192 token output. The context is still large enough to read a sizable module for triage and routing.
- how do I add Amazon Nova Micro to Atlas
- Set AWS_ACCESS_KEY_ID, AWS_SECRET_ACCESS_KEY, and AWS_REGION, request access in the Bedrock console, run `atlas models amazon-bedrock`, then put "amazon-bedrock/amazon.nova-micro-v1:0" in "small_model" only.
- can Amazon Nova Micro write code
- It should not. Amazon Nova Micro has no reasoning mode and a small model size, so it should never own Atlas's build loop. Deny write tools on that slot.
- does Bedrock keep my code inside my AWS account
- Amazon Nova Micro runs under the same IAM role and VPC endpoints as the rest of your AWS estate, so no data leaves your compliance boundary.
- Amazon Nova Micro vs Nova Lite for a coding agent
- Amazon Nova Micro is cheaper at $0.035 per Mtok input but text only with a 128,000 token context. Nova Lite is multimodal with a 300K context at $0.06 per Mtok.
Try Atlas in your terminal
The terminal-native AI coding agent. Free core, single binary.
Install AtlasRelated guides
Research a Third-Party API Before Integrating It with Atlas in 2026
How to research a third-party API with Atlas in 2026: websearch finds the current docs, webfetch pulls the page as markdown or text, and grep checks repo conventions.
Atlas for PowerShell: A Terminal-Native AI Coding Agent for Modules and Pester in 2026
Atlas is a terminal-native AI coding agent for PowerShell in 2026. It reads your .psm1 module and .psd1 manifest, adds SupportsShouldProcess, and runs Invoke-Pester behind a prompt.
Atlas for Ruby on Rails in 2026
Atlas is a terminal-native AI coding agent for Ruby on Rails in 2026. Run it in a Rails app with a config/application.rb and review every diff before it lands.
Automate GitHub Issue and Pull Request Triage with Atlas (2026 Workflow)
How to automate GitHub issue and pull request triage with Atlas in 2026: the atlas github command checks the actor has admin or write permission before it does anything.
Atlas for Swift in 2026
Atlas for Swift in 2026 empowers developers with a terminal-native AI coding agent. Index code by AST, ensure privacy with local embeddings, and review changes with unified diffs.
Atlas vs Windsurf: Terminal AI Coding Agents in 2026
Atlas and Windsurf offer distinct AI coding experiences in 2026. Atlas provides a terminal-native TUI with permission-gated tools, while Windsurf features an in-editor live preview and a dedicated GUI.
Atlas vs Jules: Terminal AI Coding Agents in 2026
Atlas and Jules comparison for 2026. Atlas offers terminal-native TUI and local code indexing. Jules provides cloud VMs and a Critic agent for PR review.
Atlas for Deno: Terminal-Native AI Coding in 2026
Atlas is a terminal-native AI coding agent for Deno. Tighten --allow-all down to real permission scopes, add JSR deps, and run deno test and deno check in 2026.