# Atlas with Amazon Nova Micro in 2026 > Amazon Nova Micro offers a substantial 128,000 token context window at an input price of just $0.035 per Mtok, making it the most cost-effective option in the Bedrock catalog. Amazon Nova Micro is ideal for absorbing high-volume, low-stakes calls within Atlas due to its ultra-low cost of $0.035 per Mtok input and a substantial 128,000 token context window, making it the cheapest model in the Bedrock catalog for specific tasks like initial triage and routing. ## Key takeaways - Amazon Nova Micro is the cheapest Bedrock model at $0.035 per Mtok input. - It offers a 128,000 token context window, suitable for sizable module triage in Atlas. - The model is text-only, making it unsuitable for tasks requiring visual input like diagrams. - It lacks a reasoning mode and has a small model size, so it should not own Atlas's build loop. - Integrates direct with AWS IAM and VPC endpoints, keeping data within your compliance boundary. - Output pricing for Amazon Nova Micro is $0.14 per Mtok. ## What is Amazon Nova Micro best for in Atlas? Amazon Nova Micro, available in 2026, excels at handling high-volume, low-stakes calls within Atlas, thanks to its industry-leading price point of $0.035 per Mtok input. This makes it the cheapest model in the Bedrock catalog, ideal for tasks where cost efficiency is paramount. Within Atlas, Amazon Nova Micro is specifically designed to absorb high-volume, low-stakes calls at effectively zero marginal cost. Its 128,000 token context window is large enough for Atlas to read a sizable code module for initial triage and routing decisions. This makes it an excellent choice for tasks such as preliminary code analysis, understanding repository structure, or powering `Axis, the hybrid semantic and keyword code search`, where a broad understanding of the codebase is needed without deep, complex reasoning. Atlas's read-only plan agent can leverage Amazon Nova Micro to draft initial plans, ensuring that more expensive, reasoning-capable models are only engaged when necessary, optimizing overall operational costs. ## What are the cost and context tradeoffs of Amazon Nova Micro? Amazon Nova Micro offers an exceptional value proposition in 2026, with an input price of just $0.035 per Mtok and a generous 128,000 token context window. This makes it about 23x cheaper than Nova Pro for input tokens, while still providing ample space for code analysis. The primary strength of Amazon Nova Micro lies in its cost-effectiveness. At $0.035 per Mtok input and $0.14 per Mtok output, it is the cheapest model in the Bedrock catalog, significantly undercutting other options like Nova Pro by approximately 23 times for input tokens. This low cost makes it highly attractive for operations that generate a large volume of token usage but do not require advanced reasoning. Its 128,000 token context window is substantial, allowing Atlas to process extensive code files or multiple related documents to gain context. Furthermore, Amazon Nova Micro runs under the same IAM role and VPC endpoints as the rest of your AWS estate, ensuring that no data leaves your compliance boundary, which is a significant advantage for security-conscious development environments. ## When should I choose a different model over Amazon Nova Micro for Atlas? Developers in 2026 should consider alternative models when Amazon Nova Micro's text-only nature or lack of a reasoning mode becomes a limitation for Atlas tasks. This model should never own Atlas's build loop due to its small size and specific design for low-stakes operations. While Amazon Nova Micro excels in cost-efficiency and context window size, it comes with specific tradeoffs that necessitate choosing a different model for certain Atlas operations. It is a text-only model, meaning it cannot process multimodal inputs such as screenshots, diagrams, or other visual information, unlike models like Nova Lite and Nova Pro. Crucially, Amazon Nova Micro has no dedicated reasoning mode and a small model size. This means it is fundamentally unsuitable for complex problem-solving, intricate code generation, or owning Atlas's build loop, which requires robust reasoning capabilities to navigate compilation errors, test failures, and iterative refinement. For tasks demanding deep logical inference, complex refactoring, or interaction with visual debugging tools, a more capable, albeit more expensive, model should be selected to drive Atlas. ## Setup 1. Set AWS_ACCESS_KEY_ID, AWS_SECRET_ACCESS_KEY, and AWS_REGION, or use AWS_PROFILE with an assumed role. 2. Request access to Amazon Nova Micro in the Bedrock console for that region. 3. Run `atlas models amazon-bedrock` and find `amazon.nova-micro-v1:0`. 4. Put it in "small_model" only: "amazon-bedrock/amazon.nova-micro-v1:0". 5. Use Atlas's allow/ask/deny permission config to deny write tools on this slot. ## FAQ ### What is the primary advantage of using Amazon Nova Micro with Atlas? Amazon Nova Micro's primary advantage is its extremely low cost of $0.035 per Mtok input, making it the cheapest model in the Bedrock catalog for high-volume, low-stakes operations within Atlas, such as initial code triage. ### What is the context window size for Amazon Nova Micro? Amazon Nova Micro offers a substantial 128,000 token context window, which is large enough for Atlas to read and triage sizable code modules effectively, providing broad context for its operations. ### Can Amazon Nova Micro handle complex reasoning tasks in Atlas? No, Amazon Nova Micro lacks a dedicated reasoning mode and has a small model size. It should not be used for complex problem-solving or to own Atlas's build loop, which requires robust logical inference. ### Is Amazon Nova Micro suitable for multimodal inputs like images or diagrams? Amazon Nova Micro is text-only, so it cannot process screenshots or diagrams. For tasks requiring visual input, developers should consider other models like Nova Lite and Nova Pro. ### How does Amazon Nova Micro's pricing compare to other Bedrock models? At $0.035 per Mtok input, Amazon Nova Micro is the cheapest model in the Bedrock catalog, approximately 23 times cheaper than Nova Pro's input pricing. Output is priced at $0.14 per Mtok. ### How does Atlas ensure compliance when using Amazon Nova Micro? Amazon Nova Micro runs under the same IAM role and VPC endpoints as your existing AWS estate, ensuring that no data leaves your compliance boundary and maintaining your established security posture. ### What Atlas features benefit most from Amazon Nova Micro's capabilities? Atlas features like `Axis, the hybrid semantic and keyword code search`, and the read-only plan agent benefit from Amazon Nova Micro's large context window and cost-effectiveness for initial triage, routing, and broad code understanding. ### What is the output token limit for Amazon Nova Micro? Amazon Nova Micro has an 8,192 token output limit, which is suitable for generating concise responses and summaries in low-stakes scenarios within Atlas. --- Canonical HTML: https://seashell.sh/resources/models/amazon-nova-micro Source of truth: aeo_pages row `/resources/models/amazon-nova-micro` (segment: Models) (this file is generated from it, never hand-edited). Licence: SeaShell is proprietary with a free core. It is not open source and there is no public source repository.