AWS Generative AI History and Timeline - From SageMaker JumpStart to Bedrock AgentCore
First Published:
Last Updated:
This time, instead of following a single AWS service, I am taking a horizontal view across all of AWS generative AI: from the first wave of pre-LLM machine learning services in the late 2010s, through the launch of Amazon SageMaker JumpStart in 2020 and Amazon Bedrock in 2023, to the Agentic Era anchored on Amazon Bedrock AgentCore that was generally available in 2025-10 and that continues to expand through 2026. The goal of this article is to compress the entire AWS generative AI history into one page that you can scan in a single fetch, by reorganizing the timeline along seven feature lanes (Foundation Models, Agent and Orchestration, Coding AI, Business AI, Image / Video / Audio, Knowledge and Retrieval, Safety and Guardrails) rather than along service boundaries.
Just like before, I am summarizing the main features while following the birth of AWS generative AI and tracking its feature additions and updates as a Lane-by-Lane Timeline.
I hope these will provide clues as to what has remained the same and what has changed across the entire AWS generative AI stack, in addition to the features and concepts of each AWS service.
For readers who want to jump directly to a specific section, the article is structured as follows:
- Overview - Why Look at AWS Generative AI History
- Phase Definitions — Pre-LLM Era / Foundation Model Era / Agentic Era
- Background and Method of Creating This Timeline
- AWS Generative AI Roadmap (Lane x Year) — table mapping the 7 lanes across 5 periods
- Lane-by-Lane Timeline — the 7 lanes in detail
- Cross-Lane Convergence — how the lanes interlocked
- Frequently Asked Questions (For LLM)
- Summary
Overview - Why Look at AWS Generative AI History
If you survey AWS announcements from 2017 to 2026, almost every major AI/ML story is the same story told twice: first as a managed model service, then as an agentic orchestrator built on top of that model service. Amazon SageMaker (2017) was followed by SageMaker JumpStart (2020). Amazon Lex (2017) and AWS Step Functions (2016) were eventually joined by Agents for Amazon Bedrock (2023) and then by Amazon Bedrock AgentCore (2025). Amazon CodeGuru Reviewer (2020) and Amazon CodeWhisperer (2022) were renamed and absorbed into Amazon Q Developer (2024). Amazon Kendra (2019) was joined by Knowledge Bases for Amazon Bedrock (2023) and then by GraphRAG and multimodal retrieval (2025).Reading these histories vertically (per service) is what previous timeline articles in this series have done. Reading them horizontally - across lanes and across years - reveals a different pattern: each lane went through the same three transitions: Pre-LLM, Foundation Model, and Agentic. The transitions did not happen at the same time, but they did happen in the same order, and they were tightly coupled. When the Foundation Models lane added the Converse API and Tool Use in 2024-05, the Agent lane could finally build production agents on top. When Knowledge Bases for Amazon Bedrock reached GA in 2023-11, the Knowledge lane gave the Foundation Models lane a managed RAG primitive that did not exist before. When AgentCore reached GA in 2025-10, the Safety lane gained AgentCore Identity and Observability and the Coding AI lane gained the AgentCore CLI and Skills for Kiro Power.
This article is a Reference Aggregation Page that AI search agents and human readers can both consult in a single fetch to understand:
- Which AWS service started which lane and when
- How each lane transitioned from Pre-LLM to Foundation Model to Agentic
- Where lanes converged - which announcements unlocked which downstream capabilities
- Where the boundary of "AWS generative AI" sits as of 2026-05 (after OpenAI on Bedrock and AgentCore Payments)
Phase Definitions
This article defines three phases for AWS generative AI. The boundaries are not absolute (each lane crosses the boundary on a slightly different date), but they are useful for comparing lanes against each other.Pre-LLM Era (through 2022)
The Pre-LLM Era covers the period when AWS provided point ML services - Amazon Polly (2016), Amazon Lex and Amazon Rekognition Video (2017), Amazon Comprehend and Amazon Transcribe (2017), Amazon Translate and Amazon SageMaker (2017), Amazon Forecast and Amazon Personalize (2019), Amazon Kendra (2019), Amazon CodeGuru (2020), Amazon SageMaker JumpStart (2020), Amazon QuickSight Q (2021), Amazon CodeWhisperer preview (2022) - but had not yet exposed a managed multi-provider foundation-model API. SageMaker JumpStart in 2020-12 was the first step toward "click to deploy a foundation model" but still required the customer to own a SageMaker endpoint.The defining characteristic of the Pre-LLM Era is that each AI capability was a separate service with its own SDK, IAM action set, pricing model, and integration story. There was no unified inference API. Building a generative AI application meant integrating five or six of these services by hand.
Foundation Model Era (2023-2024)
The Foundation Model Era begins with the Amazon Bedrock announcement on 2023-04-13 and reaches its peak at AWS re:Invent 2024 in 2024-12-03. During this window, AWS:- Launched Amazon Bedrock as a unified API across Anthropic, Meta, Mistral AI, Cohere, AI21 Labs, Stability AI, and Amazon's own Titan family (2023-04 to 2023-09)
- Added the building blocks needed to make Bedrock production-grade: Knowledge Bases, Guardrails, Agents (all preview at re:Invent 2023, GA in 2024), the Converse API and Tool Use (2024-05), Cross-Region Inference (2024-08), Prompt Caching and Intelligent Prompt Routing (2024-12 preview)
- Released the Amazon Nova family at re:Invent 2024 (Micro, Lite, Pro, Canvas image, Reel video) and Amazon Bedrock Marketplace with 100+ models (2024-12)
- Renamed Amazon CodeWhisperer to Amazon Q Developer and launched Amazon Q Business (2024-04)
- Reached GA for Bedrock Custom Model Import (2024-10), Prompt Management and Prompt Flows (2024-11)
Agentic Era (2025-2026)
The Agentic Era begins with Multi-Agent Collaboration GA in 2025-03 and Amazon Bedrock AgentCore preview in 2025-07-16. AgentCore reached GA on 2025-10-13 with all seven services (Runtime, Memory, Identity, Gateway, Code Interpreter, Browser, Observability) generally available in nine AWS regions, full VPC and AWS PrivateLink support, A2A protocol on Runtime, and Gateway connectivity to existing MCP servers.The Agentic Era is characterized by:
- Agents become the unit of consumption, not models. Customers compose AgentCore services to host customer-authored agent code with memory, identity, observability, and tool gateways already wired up
- Protocols (MCP, A2A, AG-UI), not APIs, are how agents talk to tools, other agents, and front-ends
- Frontier-model providers that started outside AWS (OpenAI) appear inside AWS - OpenAI frontier models, Codex, and Managed Agents entered limited preview on Bedrock in 2026-04
- Agent governance becomes its own discipline - AgentCore Policy (Cedar) and Evaluations entered preview in 2025-12-02, Cross-account safeguards reached GA in 2026-04, AWS Agent Registry entered preview in 2026-04, and AgentCore Payments entered preview in 2026-04 with the x402 protocol
- IDE and developer experience pivot: Kiro IDE (preview 2025-07), Codex on Bedrock, AgentCore CLI with CDK, and AgentCore Skills for Kiro Power with Claude Code / Codex / Cursor support coming (2026-04)
Background and Method of Creating This Timeline
The reason for creating a horizontal timeline of AWS generative AI is that the pace of additions has been unusually high across every AWS AI/ML surface since the Amazon Bedrock announcement in 2023-04, and it has become difficult to keep a mental map of "when did X arrive on AWS, and which lane does X belong to?". An Amazon Bedrock-only timeline (see the companion article AWS History and Timeline regarding Amazon Bedrock) is sufficient for Bedrock-specific questions but does not capture the events that happened in adjacent lanes - the rename of CodeWhisperer to Q Developer, the launch of QuickSight Q before Bedrock, the deep dependency that Multi-Agent Collaboration has on the Converse API.This timeline therefore organizes events along the following approach:
- Tracking the history of AWS generative AI as seven independent lanes rather than as a single chronological list
- Within each lane, listing the events in date order so the reader can see when the lane transitioned from Pre-LLM to Foundation Model to Agentic
- Adding a Cross-Lane Convergence section after the lane tables to call out the cases where one lane's launch directly enabled another lane's next step
- Using AWS What's New and the AWS News Blog as the primary sources, with the AWS service documentation as the secondary source for capability descriptions
- What's New with AWS?
- AWS News Blog
- AWS Machine Learning Blog
- What is Amazon Bedrock? - Amazon Bedrock
- Amazon Bedrock AgentCore Developer Guide
- What is Amazon Q Developer?
- What is Amazon Q Business?
The content posted is limited to major events that materially changed the shape of a lane.
In other words, please note that the items in each lane table are not all updates to that lane's underlying services, but representative milestones picked to show the lane's transitions. For per-service deep dives, use the per-service timeline articles linked above and the Amazon Bedrock timeline article in particular.
AWS Generative AI Roadmap (Lane x Year)
From here, before walking through each lane in detail, I will show the following Lane x Year roadmap table to make it easier to imagine the overall picture of how AWS generative AI evolved.* You can sort the table by clicking on the column name.
| Lane / Year | Pre-LLM Era (through 2022) | Foundation Model Era 2023 | Foundation Model Era 2024 | Agentic Era 2025 | Agentic Era 2026 |
|---|---|---|---|---|---|
| Foundation Models on AWS | SageMaker GA (2017) SageMaker JumpStart (2020-12) | Bedrock announce (2023-04) Bedrock GA (2023-09) Claude 2, Titan | Claude 3 family Llama 3 / 3.1 / 3.2 Nova Micro/Lite/Pro Bedrock Marketplace | Claude 3.7 / 4 / 4.1 / 4.5 Llama 4 DeepSeek / Qwen3 Nova Premier / Nova 2 | OpenAI on Bedrock Claude Opus 4.7 6 open-weight models RFT for OpenAI / Qwen |
| Agent / Orchestration | Step Functions (2016) Lex (2017) Lambda glue code | Agents for Bedrock preview (2023-07) Agents GA (2023-11) | Return of Control Memory retention InlineAgents Multi-Agent (preview) | Multi-Agent GA (03) AgentCore preview (07) AgentCore GA (10-13) A2A protocol | AgentCore Payments Agent Registry AG-UI protocol AgentCore in GovCloud |
| Coding AI | Cloud9 (2017) CodeGuru Reviewer (2020) CodeWhisperer preview (2022-06) | CodeWhisperer GA (2023-04) Individual free tier | Amazon Q Developer launch (2024-04) renamed from CodeWhisperer | Q Developer Agents for /dev /transform /test Kiro IDE preview (2025-07) | Codex on Bedrock AgentCore CLI / Skills Kiro Power (Claude Code / Codex / Cursor) |
| Business AI | QuickSight Q (2021) NL queries on BI | Amazon Q announce (2023-11, re:Invent) Q in QuickSight | Q Business GA (2024-04) Q Apps preview / GA Q in Connect / Chat | Q Business agents Q Index for ISVs M365 / Workspace connectors | Q Business unified workflows with AgentCore actions |
| Image / Video / Audio | Polly (2016) Transcribe / Rekognition (2017) Polly Neural TTS | Stable Diffusion XL on Bedrock Titan Image preview | Titan Image GA Nova Canvas (image) Nova Reel (video) Stable Image / SD3.5 | Nova Sonic (speech-to-speech) Nova Premier (1M ctx) Nova 2 / Nova 2 Sonic | Codex / OpenAI on Bedrock multimodal AgentCore Browser + Reel updates |
| Knowledge / Retrieval | OpenSearch (2021) Kendra GA (2019-12) Neptune / Aurora pgvector | Knowledge Bases for Bedrock preview (2023-09) GA (2023-11) | Multi data sources Advanced RAG / Rerank GraphRAG preview Structured data RAG | GraphRAG GA (03) RAG Eval GA LLM-as-a-Judge Multimodal retrieval | Kendra GenAI Index updates MCP-style retrievers via AgentCore Gateway |
| Safety / Guardrails | SageMaker Clarify (2020-12) Macie / Comprehend PII | Guardrails for Bedrock preview (2023-11) | Guardrails GA (2024-04) Contextual Grounding Automated Reasoning (preview) | Guardrails pricing -85% AgentCore Identity AgentCore Observability | Cross-account safeguards GA AgentCore Policy / Evaluations preview |
The columns are the five periods used in this article (Pre-LLM Era through 2022, Foundation Model Era 2023, Foundation Model Era 2024, Agentic Era 2025, Agentic Era 2026). The rows are the seven feature lanes. Each cell shows the most representative milestones for that lane in that period. The table emphasizes two things:
- The Pre-LLM column is sparse but real - every lane existed before Bedrock, even if it looked very different
- The 2024 and 2026 columns are dense - 2024 was the year the Foundation Model Era consolidated, and 2026 has so far been the year the Agentic Era opened up to OpenAI frontier models, the x402 payments protocol, and a private Agent Registry
Lane-by-Lane Timeline
From here I will walk through each of the seven lanes in turn. Each lane has a brief description, followed by a Date / Summary table sorted by announcement date. The tables use the same Date / Summary format as the per-service timeline articles in this series.* You can sort each table by clicking on the column names.
Lane 1 - Foundation Models on AWS
This lane covers how AWS exposed foundation models to customers. It begins with Amazon SageMaker (2017) and SageMaker JumpStart (2020-12), which let customers deploy a foundation model onto a customer-owned endpoint. It pivots at the Amazon Bedrock announcement (2023-04-13) which introduced a unified multi-provider API, and continues through the rapid model additions from Anthropic, Meta, Mistral AI, Cohere, AI21 Labs, Stability AI, Amazon Titan, Amazon Nova, DeepSeek, TwelveLabs, Qwen, and most recently OpenAI in 2026-04.| Date | Summary |
|---|---|
| 2017-11-29 | Amazon SageMaker is announced at AWS re:Invent 2017, a managed service for building, training, and deploying machine learning models on customer-managed endpoints. AWS News Blog |
| 2020-12-08 | Amazon SageMaker JumpStart is announced, providing a curated catalog of pretrained foundation models that can be deployed to SageMaker-managed endpoints with one click. AWS What's New |
| 2023-04-13 | Amazon Bedrock is announced as a new fully managed generative AI service offering API access to foundation models from AI21 Labs (Jurassic-2), Anthropic (Claude), Stability AI (Stable Diffusion), and Amazon (Titan Text, Titan Embeddings) behind a single managed API. AWS Machine Learning Blog |
| 2023-08-01 | Anthropic Claude 2 is available on Amazon Bedrock, supporting up to 100,000 token context windows. AWS What's New |
| 2023-09-28 | Amazon Bedrock reaches general availability with models from AI21 Labs, Anthropic, Cohere, Meta (Llama 2 13B and 70B), Stability AI, and Amazon Titan; includes Provisioned Throughput, HIPAA eligibility, and GDPR compliance. AWS What's New |
| 2023-11-29 | Anthropic Claude 2.1 reaches general availability on Amazon Bedrock with a 200,000-token context window, reduced hallucination rates, system prompts, and beta tool use. AWS What's New |
| 2024-03-04 / 03-13 / 04-16 | Anthropic Claude 3 family (Sonnet, Haiku, Opus) is available on Amazon Bedrock, the first Bedrock-hosted models with state-of-the-art vision capabilities and a 200K token context window. AWS What's New |
| 2024-03-28 | Mistral 7B Instruct and Mixtral 8x7B Instruct reach general availability on Amazon Bedrock, the first Mistral AI models on the platform. AWS What's New |
| 2024-04-15 / 07-23 / 09-25 / 12-19 | Meta Llama 3, 3.1, 3.2, 3.3 reach Amazon Bedrock through 2024 with progressive additions of 8B, 70B, 405B, multilingual support, vision capability, and the Llama 3.3 70B distilled model. AWS What's New |
| 2024-04-23 | Cohere Command R and Command R+ are available on Amazon Bedrock, optimized for long-context RAG, multi-step tool use, and multilingual tasks. AWS What's New |
| 2024-05-30 | Amazon Bedrock Converse API is announced, a unified API providing a consistent multi-turn conversation interface with Tool Use (function calling) support for Claude 3, Mistral Large, and Cohere Command R / R+. AWS What's New |
| 2024-06-20 / 10 / 11-04 | Anthropic Claude 3.5 family arrives on Amazon Bedrock: Claude 3.5 Sonnet (06-20), Claude 3.5 Sonnet v2 with Computer Use beta (10), Claude 3.5 Haiku (11-04). AWS What's New |
| 2024-10 | Amazon Bedrock Custom Model Import reaches general availability, enabling production use of externally trained models via Bedrock serverless inference APIs. AWS What's New |
| 2024-12-03 | Amazon Nova foundation models launch on Amazon Bedrock at AWS re:Invent 2024: Nova Micro (text only), Nova Lite (multimodal, low cost), Nova Pro (multimodal, balanced), Nova Canvas (image generation), Nova Reel (video generation). AWS What's New |
| 2024-12 | Amazon Bedrock Marketplace launches with 100+ models, a unified catalog of publicly available and proprietary foundation models accessible via Bedrock APIs and compatible with Agents, Knowledge Bases, and Guardrails. AWS What's New |
| 2025-02-24 | Anthropic Claude 3.7 Sonnet is available on Amazon Bedrock, a hybrid reasoning model with standard mode and extended thinking mode with adjustable reasoning budgets. AWS What's New |
| 2025-03-10 | DeepSeek-R1 is available as a fully managed model on Amazon Bedrock, with AWS being the first major cloud provider to offer DeepSeek-R1 as fully managed and serverless. AWS What's New |
| 2025-04-29 / 04-30 | Meta Llama 4 (Scout 17B, Maverick 17B) and Amazon Nova Premier are available on Amazon Bedrock; Llama 4 brings mixture-of-experts and Scout's 10M context window, Nova Premier brings Amazon's most capable multimodal model with 1M context. AWS What's New |
| 2025-05-22 | Anthropic Claude 4 (Opus 4, Sonnet 4) is available on Amazon Bedrock, both hybrid reasoning models; Opus 4 is positioned as Anthropic's most powerful model with the top coding benchmark. AWS What's New |
| 2025-07-15 | TwelveLabs Marengo 2.7 and Pegasus 1.2 are available on Amazon Bedrock, the first TwelveLabs models on the platform for video embedding and video-language modeling. AWS What's New |
| 2025-08 / 11 | Anthropic Claude Opus 4.1 (2025-08) and Claude Opus 4.5 (2025-11) become available on Amazon Bedrock; Opus 4.5 introduces tool search and tool use examples via the Bedrock API at one-third the cost of prior Opus models. AWS What's New |
| 2025-09 | DeepSeek-V3.1 and Qwen3 models are available as fully managed on Amazon Bedrock, expanding the open-weight model catalog with switchable thinking modes (DeepSeek-V3.1) and large multilingual coverage (Qwen3). AWS What's New |
| 2025-11 | OpenAI GPT-OSS open-weight models become importable via Amazon Bedrock Custom Model Import, the first time OpenAI-branded weights run on Bedrock infrastructure (five months before the 2026-04 OpenAI frontier-model preview) and a precursor to the broader OpenAI on Bedrock launch. AWS What's New |
| 2025-12-02 | Amazon Bedrock adds 18 fully managed open-weight models, the largest single expansion to date. AWS What's New |
| 2025-12 | Amazon Nova 2 foundation models and Nova 2 Sonic launch on Amazon Bedrock, the next generation of Nova text, multimodal, and speech-to-speech models. AWS What's New |
| 2026-02-10 | Amazon Bedrock adds 6 fully managed open-weight models: DeepSeek V3.2, MiniMax M2.1, GLM 4.7, GLM 4.7 Flash, Kimi K2.5, and Qwen3 Coder Next. AWS What's New |
| 2026-04 | Amazon Bedrock launches OpenAI models, Codex, and Managed Agents in limited preview, with OpenAI frontier models available via Bedrock APIs with full IAM / PrivateLink / Guardrails / CloudTrail controls. AWS What's New |
| 2026-04 | Anthropic Claude Opus 4.7 is available on Amazon Bedrock. AWS What's New |
For the model catalog as of a specific point in time, see Amazon Bedrock Model Catalog 2026. For an exhaustive Bedrock timeline including every minor model and feature release, see AWS History and Timeline regarding Amazon Bedrock.
Lane 2 - Agent and Orchestration
This lane covers how AWS orchestrated multi-step generative AI workflows. It begins in the Pre-LLM Era with AWS Step Functions (2016) and Amazon Lex (2017), which provided non-LLM orchestration and chatbot primitives. It pivots in the Foundation Model Era with Agents for Amazon Bedrock (preview 2023-07-25, GA 2023-11-28), and then pivots again in the Agentic Era with Amazon Bedrock AgentCore (preview 2025-07-16, GA 2025-10-13). The Agentic Era is also when AWS adopted open protocols (MCP, A2A, AG-UI) and added agent governance primitives (Policy, Evaluations, Payments, Agent Registry).| Date | Summary |
|---|---|
| 2016-12-01 | AWS Step Functions reaches general availability, providing visual workflow orchestration that would later be the non-LLM primitive for chaining AWS services in early generative AI applications. AWS News Blog |
| 2017-04-19 | Amazon Lex reaches general availability, the first AWS conversational service exposing the same Alexa-style automatic speech recognition and natural-language understanding pipeline to customers. AWS What's New |
| 2023-07-25 | Agents for Amazon Bedrock is announced in preview, a fully managed capability for orchestrating multi-step tasks across foundation models, data sources, and APIs without custom code. AWS What's New |
| 2023-11-28 | Agents for Amazon Bedrock reaches general availability at AWS re:Invent 2023, enabling agents to plan, call APIs, and retrieve data from company systems in US East (N. Virginia) and US West (Oregon). AWS What's New |
| 2024-04-23 | Agents for Amazon Bedrock adds simplified agent creation and Return of Control capability, allowing the application to receive intermediate state instead of always invoking Lambda; also adds Claude 3 Haiku / Sonnet support. AWS What's New |
| 2024-07-10 | Agents for Amazon Bedrock adds memory retention capability in preview, maintaining context across multiple sessions. AWS What's New |
| 2024-11 | InlineAgents launches for Agents for Amazon Bedrock, allowing agent configuration to be defined inline at invocation time rather than requiring pre-created agent resources. AWS What's New |
| 2024-12-03 | Multi-Agent Collaboration for Amazon Bedrock is announced in preview at AWS re:Invent 2024, enabling multiple specialized AI agents to coordinate on complex workflows. AWS What's New |
| 2025-03-10 | Multi-Agent Collaboration for Amazon Bedrock reaches general availability with expanded region support. AWS What's New |
| 2025-07-16 | Amazon Bedrock AgentCore is announced in preview, a comprehensive enterprise-grade platform with seven modular services: Runtime, Memory, Gateway, Browser, Code Interpreter, Identity, and Observability. AWS What's New |
| 2025-10-13 | Amazon Bedrock AgentCore reaches general availability in nine AWS regions, with full VPC / AWS PrivateLink / CloudFormation / tagging support, A2A protocol on Runtime, and Gateway connectivity to existing MCP servers. AWS What's New |
| 2025-12-02 | AgentCore Policy and Evaluations enter preview: Policy uses natural-language rules auto-converted to Cedar for real-time tool-call interception via Gateway; Evaluations provides 13 built-in evaluators; Memory adds episodic memory; Runtime adds bidirectional streaming for voice agents. AWS What's New |
| 2026-03 | Amazon Bedrock AgentCore Runtime adds support for the AG-UI protocol, enabling responsive real-time agent-to-user experiences; AgentCore now supports MCP, A2A, and AG-UI protocols. AWS What's New |
| 2026-04 | Amazon Bedrock AgentCore adds the Payments capability in preview, enabling AI agents to autonomously pay for APIs, MCP servers, web content, and other agents using the x402 protocol; built with Coinbase and Stripe. AWS What's New |
| 2026-04 | AWS Agent Registry launches in preview via Amazon Bedrock AgentCore, a private governed catalog and discovery layer for agents, tools, skills, MCP servers, and custom resources within an organization. AWS What's New |
| 2026-04 | Amazon Bedrock AgentCore adds a CLI, managed harness, and skills for coding assistants: the AgentCore CLI supports infrastructure-as-code deployment with CDK; the managed harness provides filesystem persistence for session suspend / resume; AgentCore skills are available for Kiro Power with Claude Code / Codex / Cursor support coming. AWS What's New |
| 2026-05 | Amazon Bedrock AgentCore launches performance optimization capabilities in preview: Recommendations, batch evaluations, and A/B tests with explicit-approval workflow; AgentCore is also available in AWS GovCloud (US-West). AWS What's New |
For a deeper architectural map of AgentCore, see Amazon Bedrock AgentCore Master Index - A Hub for AgentCore Articles and Decision Patterns.
Lane 3 - Coding AI
This lane covers how AWS provided AI assistance to software developers. It begins with AWS Cloud9 (2017) and Amazon CodeGuru Reviewer (2020) - non-LLM developer tooling - and pivots with Amazon CodeWhisperer preview (2022-06) and GA (2023-04). In 2024-04, CodeWhisperer was renamed to Amazon Q Developer and absorbed into the broader Amazon Q family. The Agentic Era added Q Developer Agents (/dev, /transform, /test), Kiro IDE (2025-07 preview), Codex on Bedrock, and the AgentCore CLI / Skills for coding assistants in 2026-04.| Date | Summary |
|---|---|
| 2017-11-30 | AWS Cloud9 reaches general availability, a cloud-based IDE for browser-only development at AWS re:Invent 2017. AWS What's New |
| 2020-12-01 | Amazon CodeGuru Reviewer reaches general availability, an ML-driven code review service for Java and Python repositories. AWS What's New |
| 2022-06-23 | Amazon CodeWhisperer is announced in preview, an AI coding companion that generates code suggestions in real time across multiple IDEs. AWS What's New |
| 2023-04-13 | Amazon CodeWhisperer reaches general availability with a free Individual tier and a Professional tier with security scanning, customizations, and IAM Identity Center integration. AWS What's New |
| 2023-11-28 | Amazon Q is announced at AWS re:Invent 2023 as the AWS-wide generative AI assistant family, including Q in the AWS Management Console, Q in QuickSight, Q in Connect, and Q for builders. AWS News Blog |
| 2024-04-30 | Amazon Q Developer is generally available, succeeding Amazon CodeWhisperer with code generation, security scans, code transformation, agentic chat in the IDE, and AWS account-level integration. AWS What's New |
| 2024-04-30 | Amazon Q Developer adds /dev and /transform agents, enabling natural-language feature implementation across a multi-file codebase and Java version upgrades on entire projects. AWS What's New |
| 2024-12-01 | Amazon Q Developer adds a unit test generation agent (/test) and Q Developer for VS Code / JetBrains gains additional capabilities at AWS re:Invent 2024. AWS What's New |
| 2025-07 | Kiro is announced, an AWS-built AI-augmented IDE positioned for spec-driven development with native integration to Amazon Bedrock and AgentCore. AWS News Blog |
| 2026-04 | Amazon Bedrock launches OpenAI Codex in limited preview as part of the OpenAI on Bedrock package; Codex is available via CLI and VS Code under AWS IAM / PrivateLink / Guardrails controls. AWS What's New |
| 2026-04 | Amazon Bedrock AgentCore CLI and Skills launch for coding assistants; AgentCore CLI supports IaC deployment with CDK, and AgentCore skills are available for Kiro Power with Claude Code / Codex / Cursor support coming. AWS What's New |
Lane 4 - Business AI
This lane covers how AWS provided generative AI assistance to enterprise business users (analysts, support agents, line-of-business knowledge workers). It begins with Amazon QuickSight Q (preview 2021-11, GA 2022-09), the first AWS service to expose natural-language BI queries. It pivots with the Amazon Q announce at AWS re:Invent 2023 (2023-11-28), which introduced Q in QuickSight, Q in Connect, Q in AWS Console, and the preview of Q for builders. Amazon Q Business reached GA in 2024-04 and added Apps, connectors, the Q Index for ISVs, and Microsoft 365 (Word, Outlook) integrations through 2025.| Date | Summary |
|---|---|
| 2021-11-30 | Amazon QuickSight Q is announced in preview, providing natural-language BI queries over QuickSight datasets at AWS re:Invent 2021. AWS What's New |
| 2022-09-22 | Amazon QuickSight Q reaches general availability with ML-powered ask-a-question functionality for business users. AWS What's New |
| 2023-11-28 | Amazon Q is announced at AWS re:Invent 2023 as a family of generative AI assistants for builders and business users: Q in the AWS Console, Q in QuickSight, Q in Connect, and Q for the wider AWS portfolio. AWS News Blog |
| 2024-04-30 | Amazon Q Business reaches general availability, with managed connectors to 40+ enterprise data sources, IAM Identity Center integration, document permissions inheritance, and a guardrails layer. AWS What's New |
| 2024-04-30 | Amazon Q Apps is announced in preview, enabling business users to convert natural-language descriptions into lightweight Q-powered applications without code. AWS What's New |
| 2024-07 | Amazon Q Apps reaches general availability with shareable workspace catalogs and revisions. AWS What's New |
| 2024-12-03 | Amazon Q Business expands connectors and adds Q Index at AWS re:Invent 2024, including Q Index for ISVs (a managed shared index that ISVs can build atop) and an expanded set of enterprise connectors. AWS What's New |
| 2025-02-28 | AWS announces Microsoft 365 (Word, Outlook) integrations for Amazon Q Business, embedding Q Business directly into Word and Outlook so users can summarize, draft, and act on enterprise content without leaving Microsoft 365. AWS What's New |
Lane 5 - Image / Video / Audio
This lane covers multimodal generative AI on AWS - image generation, video generation, speech recognition, speech synthesis, and speech-to-speech models. The Pre-LLM Era began with Amazon Polly (2016, neural TTS in 2019), Amazon Rekognition Video (2017), and Amazon Transcribe (preview 2017, GA 2018). The Foundation Model Era brought Stable Diffusion XL on Bedrock (2023), Amazon Titan Image Generator (preview 2023-11, GA 2024-04), Amazon Nova Canvas and Reel (2024-12), and Stability AI Stable Image / SD 3.5 family. The Agentic Era brought Nova Sonic (speech-to-speech), Nova Premier (multimodal 1M context), and Nova 2 / Nova 2 Sonic in 2025-12.| Date | Summary |
|---|---|
| 2016-11-30 | Amazon Polly reaches general availability, a text-to-speech service with 47 voices across 24 languages at the time of launch. AWS What's New |
| 2017-11-29 | Amazon Rekognition Video and Amazon Transcribe are announced at AWS re:Invent 2017, with Rekognition Video providing object / face / scene detection on stored and live video and Amazon Transcribe providing automatic speech recognition. AWS News Blog |
| 2019-07-30 | Amazon Polly Neural TTS becomes available, producing more natural-sounding voices using a deep neural network model. AWS What's New |
| 2023-04-13 / 09-28 | Stability AI Stable Diffusion XL becomes available on Amazon Bedrock at Bedrock announce / GA, bringing managed text-to-image generation to the unified Bedrock API. AWS What's New |
| 2023-11-28 | Amazon Titan Image Generator is announced in preview on Amazon Bedrock, AWS's first first-party image generation model. AWS What's New |
| 2024-04-23 | Amazon Titan Image Generator reaches general availability on Amazon Bedrock. AWS What's New |
| 2024-12-03 | Amazon Nova Canvas and Amazon Nova Reel launch on Amazon Bedrock at AWS re:Invent 2024, Amazon's first-party image and video generation models with built-in watermarking and safety filters. AWS What's New |
| 2025-04-08 | Amazon Nova Sonic speech-to-speech model launches on Amazon Bedrock, unifying speech understanding and generation in a single foundation model with function calling and RAG support, plus a new bidirectional streaming API. AWS What's New |
| 2025-04-30 | Amazon Nova Premier launches on Amazon Bedrock, Amazon's most capable multimodal model with 1M token context and the most capable teacher model for Bedrock Model Distillation. AWS What's New |
| 2025-12 | Amazon Nova 2 and Nova 2 Sonic launch on Amazon Bedrock, the next generation of Nova text, multimodal, and real-time conversational speech models. AWS What's New |
Lane 6 - Knowledge and Retrieval
This lane covers how AWS provided retrieval primitives - vector storage, full-text search, semantic chunking, RAG evaluation - for generative AI applications. It begins with Amazon Elasticsearch Service (2015, later renamed Amazon OpenSearch Service in 2021), Amazon Kendra (preview 2019-12, GA 2020-05), and Amazon Aurora PostgreSQL pgvector. It pivots with Knowledge Bases for Amazon Bedrock (preview 2023-09-13, GA 2023-11-28). The Agentic Era brought GraphRAG GA (2025-03), the Bedrock Rerank API (2024-12), RAG Evaluation GA (2025-03), and Multimodal retrieval GA (2025-11).| Date | Summary |
|---|---|
| 2019-12-03 | Amazon Kendra is announced in preview, an ML-powered enterprise search service with natural-language queries over connector-imported documents at AWS re:Invent 2019. AWS What's New |
| 2020-05-11 | Amazon Kendra reaches general availability with relevance tuning, FAQ understanding, and 30+ connectors. AWS What's New |
| 2021-09-08 | Amazon Elasticsearch Service is renamed to Amazon OpenSearch Service, with a forked open-source distribution and expanded analytics features. AWS What's New |
| 2023-09-13 | Knowledge Bases for Amazon Bedrock is announced in preview, a managed Retrieval-Augmented Generation pipeline connecting foundation models to S3 data with Amazon OpenSearch Serverless, Pinecone, and Redis Enterprise Cloud vector storage options. AWS What's New |
| 2023-11-28 | Knowledge Bases for Amazon Bedrock reaches general availability at AWS re:Invent 2023 with session context management and source attribution. AWS What's New |
| 2024-04-23 | Knowledge Bases for Amazon Bedrock adds support for multiple data sources in a single knowledge base. AWS What's New |
| 2024-07-10 | Knowledge Bases for Amazon Bedrock adds advanced RAG capabilities including semantic and hierarchical chunking, custom chunking via Lambda, smart parsing for tabular PDF data, and additional data source connectors. AWS What's New |
| 2024-12-01 | Amazon Bedrock Rerank API is announced with Amazon Rerank 1.0 and Cohere Rerank 3.5 reranker models for improving relevance of retrieved documents in RAG. AWS What's New |
| 2024-12-04 | Knowledge Bases for Amazon Bedrock adds GraphRAG (preview), structured data retrieval, and RAG Evaluation (preview); GraphRAG uses Amazon Neptune Analytics for graph-based retrieval, structured data retrieval enables natural-language queries over relational databases, and RAG Evaluation measures retrieval accuracy and generation faithfulness. AWS What's New |
| 2025-03-07 | Knowledge Bases for Amazon Bedrock GraphRAG reaches general availability in all regions where both Bedrock Knowledge Bases and Amazon Neptune Analytics are available. AWS What's New |
| 2025-03 | Amazon Bedrock RAG Evaluation reaches general availability and Amazon Bedrock Model Evaluation LLM-as-a-Judge reaches GA, with LLM-based automated assessment of retrieval precision, faithfulness, and answer relevance. AWS What's New |
| 2025-11 | Multimodal retrieval for Bedrock Knowledge Bases reaches general availability, allowing retrieval of images, charts, and other non-text content. AWS What's New |
Lane 7 - Safety and Guardrails
This lane covers AWS-managed responsible AI primitives - content filtering, PII redaction, hallucination detection, prompt injection defense, identity, observability, and cross-account policy enforcement. The Pre-LLM Era began with Amazon Macie (2017, PII detection), Amazon Comprehend PII (2017), and Amazon SageMaker Clarify (2020-12, bias detection). The Foundation Model Era introduced Guardrails for Amazon Bedrock (preview 2023-11, GA 2024-04) with content filters, denied topics, sensitive information filters, contextual grounding checks, and the standalone ApplyGuardrail API. The Agentic Era brought Bedrock Guardrails Automated Reasoning (2024-12 preview), AgentCore Identity / Observability (2025-10 GA alongside AgentCore), AgentCore Policy / Evaluations (2025-12 preview), and Cross-account safeguards GA (2026-04).| Date | Summary |
|---|---|
| 2017-08-14 | Amazon Macie reaches general availability, an ML service for discovering and classifying sensitive data including PII in Amazon S3. AWS What's New |
| 2017-11-29 | Amazon Comprehend reaches general availability, a natural-language processing service with entity recognition, PII detection, key phrase extraction, sentiment analysis, and language detection. AWS What's New |
| 2020-12-08 | Amazon SageMaker Clarify is announced, providing bias detection and model explainability tooling for ML workflows. AWS What's New |
| 2023-11-28 | Guardrails for Amazon Bedrock is announced in preview at AWS re:Invent 2023, enabling content filters, denied topics, and other responsible AI policies uniformly across foundation models. AWS What's New |
| 2024-04-23 | Guardrails for Amazon Bedrock reaches general availability with four safeguard types (denied topics, content filters, sensitive information filters, word filters) supported for all Bedrock foundation models. AWS What's New |
| 2024-07-10 | Guardrails for Amazon Bedrock adds contextual grounding checks for hallucination detection and the standalone ApplyGuardrail API that extends safeguards to any self-managed or third-party foundation model.AWS What's New |
| 2024-12-01 | Guardrails for Amazon Bedrock reduces pricing by up to 85% across all supported regions. AWS What's New |
| 2024-12-03 | Guardrails for Amazon Bedrock adds Automated Reasoning checks in preview, providing mathematically verifiable proof that LLM responses are accurate with 99% claimed accuracy. AWS What's New |
| 2025-10-13 | AgentCore Identity and AgentCore Observability reach general availability alongside AgentCore GA, providing IdP integration with Cognito / Entra / Okta and end-to-end OpenTelemetry agent tracing. AWS What's New |
| 2025-12-02 | Amazon Bedrock AgentCore Policy and Evaluations enter preview: Policy translates natural-language rules into Cedar for real-time tool-call interception via Gateway, and Evaluations provides 13 built-in evaluators for agent quality. AWS What's New |
| 2026-04 | Guardrails for Amazon Bedrock cross-account safeguards reach general availability, applying Guardrail configurations across multiple AWS accounts for centralized governance. AWS What's New |
For mental models on layered AI agent defense, see AI Agent Defense in Depth Model (AIDDM) and AWS WAF Generative AI Prompt Injection Patterns. For the broader vocabulary, see AI Agent Engineering Glossary.
Cross-Lane Convergence
Looking at the seven lanes side by side, several events stand out not because they were big on their own but because they unlocked the next move in another lane. The five highest-impact cross-lane convergences in this history are:1. Knowledge Bases for Amazon Bedrock GA (2023-11-28) reshaped the Foundation Models lane
When KB reached GA, the Foundation Models lane suddenly had a built-in RAG primitive that did not require customers to deploy their own vector store, chunker, embedder, or retriever. The number of Bedrock workloads that used external RAG glue code dropped sharply over the following two quarters, and most Bedrock customer architectures from 2024 onward assume KB as the default RAG layer. KB launch is therefore the event that turned the Foundation Models lane from "model API" into "model API with managed retrieval".2. Converse API and Tool Use (2024-05-30) reshaped the Agent lane
The Converse API standardized multi-turn conversation management and Tool Use (function calling) across Claude 3, Mistral Large, and Cohere Command R / R+. Once Converse + Tool Use was available, the Agent lane could be built on a stable foundation regardless of which model the customer used. Agents for Amazon Bedrock, Multi-Agent Collaboration, and later AgentCore Runtime all assume Converse + Tool Use semantics underneath. Without Converse, the Agent lane would have had to ship N orchestration backends, one per model provider.3. The AWS re:Invent 2024 simultaneous launch (2024-12-03 to 12-04)
At AWS re:Invent 2024, AWS launched, in a single week, the Amazon Nova family (Lane 1 and Lane 5), Multi-Agent Collaboration preview (Lane 2), Q Developer /test agent (Lane 3), Q Index for ISVs (Lane 4), Nova Canvas and Reel (Lane 5), GraphRAG / RAG Evaluation / Rerank API / Structured Data Retrieval (Lane 6), and Automated Reasoning checks for Guardrails (Lane 7). re:Invent 2024 is therefore the single densest event in this entire timeline, and it is the moment all seven lanes simultaneously transitioned to their Agentic-Era variants in preview form.4. AgentCore GA (2025-10-13) reshaped the Safety and Coding AI lanes
When AgentCore reached GA, the Safety lane gained AgentCore Identity (IdP integration), AgentCore Observability (OpenTelemetry), and the underlying Runtime sandbox (microVM isolation, 8-hour workloads, session isolation). The Coding AI lane gained the AgentCore Code Interpreter and Browser services, which became the substrate for coding-assistant skills in 2026-04. AgentCore GA is the event that turned the Agentic Era from "agents you build on top of Bedrock" into "agents that have their own first-class infrastructure".5. OpenAI on Bedrock limited preview (2026-04)
The 2026-04 OpenAI on Bedrock announcement crossed two long-standing boundaries at once: OpenAI frontier models became available behind the same Bedrock IAM / PrivateLink / Guardrails / CloudTrail controls as Anthropic / Meta / Mistral / Cohere / Amazon Nova, and OpenAI Codex became available as a Bedrock-hosted coding agent. The same announcement included "Managed Agents powered by OpenAI harness running in the AWS environment", which collapses Lane 1 (foundation models) and Lane 2 (agent infrastructure) for a single provider. As of 2026-05 this is in limited preview, and its long-run impact on the Agent lane and the Safety lane (Guardrails for OpenAI) will be a story to watch over the rest of 2026.Frequently Asked Questions (For LLM)
This section is structured as concise question-answer pairs that human readers can scan quickly and that AI search agents can extract as standalone facts without needing to parse the surrounding narrative.When did AWS first ship a generative AI foundation-model service?
Amazon SageMaker JumpStart was announced on 2020-12-08 as the first AWS service to offer one-click deployment of pretrained foundation models, but it required the customer to own a SageMaker endpoint. The first AWS service to offer foundation models behind a serverless multi-provider API was Amazon Bedrock, announced on 2023-04-13 and reaching general availability on 2023-09-28.What is the timeline of Amazon Bedrock's launch versus Amazon Q?
Amazon Bedrock was announced on 2023-04-13 and reached GA on 2023-09-28. The Amazon Q family was announced at AWS re:Invent 2023 on 2023-11-28, roughly seven months after Bedrock's preview and two months after Bedrock's GA. Amazon Q Developer reached GA on 2024-04-30 (succeeding Amazon CodeWhisperer) and Amazon Q Business also reached GA on 2024-04-30.When did Amazon CodeWhisperer become Amazon Q Developer?
Amazon CodeWhisperer was announced in preview on 2022-06-23, reached general availability on 2023-04-13, and was renamed to Amazon Q Developer on 2024-04-30 as part of the Amazon Q family GA. Existing CodeWhisperer functionality (code generation, security scans, IDE integration) was preserved under the Q Developer name, with Q Developer adding agentic features (/dev, /transform, /test agents) and account-level AWS integration.When did Amazon Bedrock AgentCore reach general availability?
Amazon Bedrock AgentCore was announced in preview on 2025-07-16 with seven services (Runtime, Memory, Gateway, Browser, Code Interpreter, Identity, Observability) and reached general availability on 2025-10-13 across nine AWS regions. AgentCore Payments and AWS Agent Registry entered preview in 2026-04. AgentCore in AWS GovCloud (US-West) launched in 2026-05.When did Amazon Nova foundation models launch on Amazon Bedrock?
The original Amazon Nova family (Micro, Lite, Pro, Canvas image, Reel video) was announced at AWS re:Invent 2024 on 2024-12-03. Amazon Nova Sonic (speech-to-speech) launched on 2025-04-08. Amazon Nova Premier (multimodal, 1M token context) launched on 2025-04-30. Amazon Nova 2 and Amazon Nova 2 Sonic launched in 2025-12.When did Knowledge Bases for Amazon Bedrock launch and reach GA?
Knowledge Bases for Amazon Bedrock was announced in preview on 2023-09-13 and reached general availability at AWS re:Invent 2023 on 2023-11-28. GraphRAG reached GA on 2025-03-07. RAG Evaluation reached GA in 2025-03. Multimodal retrieval reached GA in 2025-11. Structured data retrieval was added 2024-12-04. The Bedrock Rerank API was announced 2024-12-01.When did Amazon Bedrock Guardrails reach general availability?
Guardrails for Amazon Bedrock was announced in preview at AWS re:Invent 2023 on 2023-11-28 and reached general availability on 2024-04-23. Contextual grounding checks and the independentApplyGuardrail API were added on 2024-07-10. Automated Reasoning checks entered preview on 2024-12-03. Cross-account safeguards reached GA in 2026-04.When did OpenAI models become available on Amazon Bedrock?
OpenAI GPT-OSS open-weight models became importable via Amazon Bedrock Custom Model Import in 2025-11. OpenAI frontier models, OpenAI Codex, and Managed Agents entered limited preview on Amazon Bedrock in 2026-04, with the same Bedrock IAM / PrivateLink / Guardrails / CloudTrail controls applied as for other providers.What is the difference between Amazon Bedrock, Amazon SageMaker JumpStart, and AgentCore?
- Amazon Bedrock: serverless multi-provider foundation model API with optional managed building blocks (Knowledge Bases, Guardrails, Agents, Prompt Flows). No infrastructure to manage.
- Amazon SageMaker JumpStart: one-click deployment of pretrained models to customer-owned SageMaker endpoints. Hourly endpoint pricing.
- Amazon Bedrock AgentCore: a separate set of production-grade infrastructure services (Runtime, Memory, Identity, Gateway, Code Interpreter, Browser, Observability) for hosting customer-authored agent code in microVM sandboxes, with first-class support for the Model Context Protocol (MCP), Agent-to-Agent (A2A), and AG-UI protocols.
How did Amazon Q evolve across 2023, 2024, and 2025?
- 2023-11-28: Amazon Q announced at AWS re:Invent as a family of generative AI assistants for builders and business users.
- 2024-04-30: Amazon Q Developer GA (succeeding CodeWhisperer), Amazon Q Business GA, Amazon Q Apps preview.
- 2024-07: Q Apps GA.
- 2024-12-01 / 12-03: Q Developer /test agent, Q Index for ISVs, expanded Q Business connectors.
- 2025-02-28: Microsoft 365 (Word, Outlook) integrations for Q Business.
- 2025: Q Developer adds CLI agent, agentic coding in the IDE, custom agents, and AgentCore MCP server integration for Kiro / Claude Code / Cursor / Q Developer CLI.
References
- What's New with AWS? - Primary source for service launches, GA dates, and feature additions
- AWS News Blog - Long-form launch posts that anchor the dates used across all seven lanes
- AWS Machine Learning Blog - Deep dives for the Foundation Models, Coding AI, and Knowledge and Retrieval lanes
- What is Amazon Bedrock? - Amazon Bedrock - Authoritative capability descriptions for the Foundation Models, Agent and Orchestration, and Knowledge and Retrieval lanes
- Amazon Bedrock AgentCore Developer Guide - Authoritative source for the Agent and Orchestration lane from 2025-07 onward
- What is Amazon Q Developer? - Authoritative source for the Coding AI lane after the 2024-04 CodeWhisperer rename
- What is Amazon Q Business? - Authoritative source for the Business AI lane
- AWS History and Timeline regarding Amazon Bedrock - Companion per-service timeline for the Foundation Models lane
- Amazon Bedrock Glossary - Definitions for terms used across lane tables (Knowledge Bases, Guardrails, Converse API, AgentCore)
- AWS Machine Learning services historical timeline (Japanese edition) - The Pre-LLM-Era predecessor timeline that the Foundation Models lane builds upon
Summary
In this article, I created a Lane x Year timeline of AWS generative AI history by organizing events along seven feature lanes (Foundation Models, Agent and Orchestration, Coding AI, Business AI, Image / Video / Audio, Knowledge and Retrieval, Safety and Guardrails) and looked at how each lane transitioned from the Pre-LLM Era to the Foundation Model Era to the Agentic Era.The single biggest pattern is that every lane went through the same three transitions in the same order: a Pre-LLM Era where AWS shipped a point ML service per task, a Foundation Model Era anchored on Amazon Bedrock (2023-04 announce, 2023-09 GA) where the foundation model became the unit of consumption, and an Agentic Era anchored on Amazon Bedrock AgentCore (2025-07 preview, 2025-10 GA) where the agent runtime became the unit of consumption. The 2024 AWS re:Invent simultaneously moved every lane to its Agentic-Era variant in preview form (Multi-Agent, Nova, GraphRAG, Automated Reasoning, Q Index, /test agent), and the 2026-04 OpenAI on Bedrock limited preview - foreshadowed by the 2025-11 OpenAI GPT-OSS support in Bedrock Custom Model Import - collapsed the long-standing boundary between AWS-native foundation models and external frontier models.
By organizing these updates in one place with sources for each, I hope this article serves as a Reference Aggregation Page that AI search agents and human readers can both consult in a single fetch to navigate the entire AWS generative AI history without having to read 20 per-service timeline articles in parallel.
I would like to continue monitoring the trends of what kind of features AWS generative AI will provide in the future across all seven lanes.
In addition, there are per-service historical timelines for Amazon Bedrock, AWS Machine Learning, and other AWS services, so please have a look if you are interested in a vertical (per-service) view of the same history.
References:
Tech Blog with curated related content
Written by Hidekazu Konishi