AWS Transform now supports model-to-model migration assessment for generative AI workloads
Automatically assess and migrate your OpenAI, Gemini, or Anthropic workloads to Amazon Bedrock with production-ready code and cost comparisons.
View original announcement →Visual Summary
graph TD
A{{AWS Transform Model Migration}}:::announced
B((Developer Codebase)):::external
C([Codebase Scanning]):::feature
D([Model Mapping & Cost]):::feature
E([Code Generation]):::feature
F(Amazon Bedrock):::compute
G(Amazon CloudWatch):::compute
H([Bedrock Guardrails]):::feature
I((Third-Party Providers)):::external
B ==>|"ATX CLI scan"| A
I -.->|"source models"| A
A -->|"detects SDKs"| C
A -->|"maps equivalents"| D
A -->|"produces changes"| E
E ==>|"migrates to"| F
F -->|"observability"| G
F -->|"safety"| H
classDef announced fill:#ff9900,stroke:#ec7211,color:#fff,font-weight:bold
classDef compute fill:#e3f2fd,stroke:#1565c0,color:#1565c0
classDef storage fill:#e8f5e9,stroke:#2e7d32,color:#2e7d32
classDef feature fill:#fff3e0,stroke:#e65100,color:#e65100
classDef external fill:#f5f5f5,stroke:#616161,color:#616161
What's New
AWS Transform has added a model-to-model migration custom transformation that automates the assessment and planning of generative AI workload migrations from third-party providers (OpenAI, Google Gemini, Anthropic, LiteLLM, Ollama) to Amazon Bedrock. The AI-powered agent scans codebases, identifies all AI SDKs and models in use, and produces a comprehensive migration plan including model mappings, cost comparisons, and production-ready code changes. It is available at no additional charge beyond standard AWS Transform pricing across all supported AWS Regions.
How It Works
- Codebase scanning: The ATX CLI runs the
mke-genai-model-migrationcustom transformation against your repository, automatically detecting every AI SDK, model reference, and integration pattern present in the code. - Interactive requirements gathering: The agent asks targeted questions about migration priorities, cost sensitivity, and operational requirements before generating recommendations, ensuring the plan reflects your specific context.
- Model mapping and equivalency analysis: Each detected third-party model is mapped to a functionally equivalent Amazon Bedrock model, with transparent side-by-side cost comparisons to support informed decision-making.
- Code generation: The agent produces production-ready code changes that swap the model layer while preserving the existing application architecture, including support for direct SDK calls, LangChain, LlamaIndex, CrewAI, LangGraph, and multi-provider routing via LiteLLM.
- Zero-code-change path: For eligible workloads, the agent recommends Amazon Bedrock's OpenAI-compatible endpoints, enabling migration without any code modifications.
- Intelligent cost optimization: Tiered model routing recommendations, prompt caching analysis, and model lifecycle awareness (excluding models within 90 days of end-of-life) are built into the recommendation engine.
- Output artifacts: The transformation delivers a detailed migration assessment, an SDK migration guide, model mapping documentation, cost estimates, and a step-by-step operational runbook.
Why It's Important
- Reduces migration risk and effort: Manually migrating AI workloads across frameworks like LangChain, CrewAI, or LangGraph can take weeks per service; automation compresses this to hours while reducing the risk of breaking production systems.
- Unlocks enterprise-grade AWS security: Migrating to Bedrock enables IAM-based access control, VPC endpoint isolation, and Amazon Bedrock Guardrails — capabilities that are difficult or impossible to replicate with third-party API providers.
- Consolidates operational tooling: Organizations gain unified observability through Amazon CloudWatch and a single control plane for all AI workloads, reducing the complexity of managing multiple vendor dashboards and billing relationships.
- Addresses cost visibility: Transparent cost comparisons at the model level give engineering and finance teams the data needed to justify migration and optimize spend through features like prompt caching.
- Scales across complex, real-world codebases: The agent handles multi-model, multi-framework architectures — not just simple single-provider integrations — making it viable for production enterprise systems rather than toy projects.
- Accelerates AI-readiness on AWS: For organizations already on AWS, consolidating AI workloads on Bedrock reduces vendor sprawl and aligns generative AI infrastructure with existing cloud governance and compliance frameworks.
How It's Different
- Architecture-preserving transformation: Unlike manual rewrites or generic refactoring tools, this agent swaps only the model layer, leaving application logic, agentic architectures, and framework integrations intact.
- Multi-framework awareness: The tool understands not just raw SDK calls but also higher-order patterns like LangChain chains, LlamaIndex pipelines, CrewAI crews, and LangGraph graphs — a level of semantic understanding that generic code migration tools lack.
- Lifecycle-aware recommendations: The 90-day end-of-life exclusion filter ensures recommendations are forward-looking and won't route workloads onto models that are about to be deprecated, a safeguard absent from manual migration approaches.
- Zero-code-change option: The recommendation of Bedrock's OpenAI-compatible endpoints as a migration path is unique — it allows teams to migrate without touching application code at all, a capability no manual process can match in speed.
- Integrated cost intelligence: Cost comparison is embedded directly in the migration plan rather than being a separate exercise, enabling cost-aware model selection as part of the migration workflow itself.
- AWS-managed and continuously improving: As an AWS-managed transformation in the Transform catalog, it benefits from continual learning across all executions and AWS's accumulated migration expertise, unlike one-off internal scripts or consulting engagements.
When to Prefer It
- When migrating startups or teams from OpenAI/Gemini to AWS: Organizations joining AWS or consolidating under an AWS commercial agreement need a fast, low-risk path to move AI workloads without large engineering investments.
- When codebases use multiple AI providers or models: If your application routes different tasks to different models (e.g., GPT-4 for reasoning, an embedding model for retrieval), the agent's multi-model detection and mapping capabilities are essential.
- When agentic frameworks are in use: Teams using LangChain, LlamaIndex, CrewAI, or LangGraph should prefer this tool over manual migration, as it understands framework-specific patterns and generates correct Bedrock-compatible replacements.
- When security and compliance requirements demand AWS-native controls: Enterprises in regulated industries that need IAM-based access, VPC isolation, and audit logging should use this to accelerate the move to Bedrock's enterprise security model.
- When cost optimization is a migration driver: If reducing inference costs or gaining prompt caching benefits is a primary goal, the agent's tiered routing recommendations and cost comparison outputs provide actionable guidance.
- When engineering capacity is limited: Small teams or organizations without dedicated platform engineers can use this managed transformation to produce a migration plan and code changes without deep AWS Bedrock expertise.
- When a zero-downtime, zero-code-change migration is required: For workloads already using OpenAI-compatible API patterns, the Bedrock-compatible endpoint recommendation enables migration with no code changes and minimal operational risk.
Availability
- GA status: Generally available as of June 16, 2026, as an AWS-managed custom transformation in the AWS Transform catalog.
- Regional availability: Available in all AWS Regions where AWS Transform is currently offered; no additional regional opt-in is required.
- Pricing: No additional charge beyond standard AWS Transform pricing; the model-to-model migration transformation is included in the existing Transform pricing model.
- Getting started: Requires installation of the ATX CLI; the transformation is invoked by running
mke-genai-model-migrationagainst the target codebase. - Supported source providers: OpenAI (direct SDK and ChatGPT API), Google Gemini (
@google/genai), direct Anthropic SDK, and open-source models via LiteLLM or Ollama. - Supported integration patterns: Direct SDK calls, LangChain, LlamaIndex, CrewAI, LangGraph, custom agent loops, and multi-provider routing layers.
- Limitation — lifecycle filter: Models within 90 days of end-of-life are explicitly excluded from all recommendations, which may limit options for workloads requiring specific model versions.
