Release Timeline

All 60 tracked AI product releases, grouped by month, newest first.

April 2026

1 release

March 2026

29 releases
ModelMar 26, 2026

Google

Gemini 3.1 Flash Live: Making audio AI more natural and reliable

Gemini 3.1 Flash Live is Google's latest AI audio model.

This release enhances real-time dialogue capabilities, making AI audio interactions more natural and reliable for developers, enterprises, and users worldwide.

AIaudio modelreal-time dialoguemultilingual+6
FeatureMar 25, 2026

DeepMind

Lyria 3 Pro: Create Longer Tracks in More

Introduces Lyria 3 Pro with enhanced capabilities.

This release enhances the creative capabilities of Lyria, allowing for more complex and integrated music production across Google platforms.

DeepMindLyriaMusicAI+2
FeatureMar 24, 2026

OpenAI

Powering Product Discovery in ChatGPT

ChatGPT introduces visually immersive shopping experiences.

This update transforms ChatGPT into a more comprehensive shopping assistant, enhancing user experience and providing merchants with access to high-intent shoppers.

OpenAIChatGPTShoppingAgentic Commerce Protocol+5
FeatureMar 24, 2026

OpenAI

OpenAI Releases Teen Safety Policies for AI Developers

OpenAI introduces prompt-based safety policies for gpt-oss-safeguard.

These policies provide developers with tools to ensure AI systems are safer for teen users, addressing specific age-related risks and promoting responsible innovation.

OpenAITeen SafetyAI Policiesgpt-oss-safeguard+4
BenchmarkMar 24, 2026

ServiceNow-AI

A New Framework for Evaluating Voice Agents (EVA)

Introduces EVA, an end-to-end evaluation framework for voice agents.

EVA is the first framework to jointly score task success and conversational experience, addressing a critical gap in evaluating voice agent quality.

Voice AgentsEvaluation FrameworkConversational AIBenchmarking+6
FeatureMar 23, 2026

OpenAI

Creating with Sora Safely

OpenAI releases Sora 2 model and Sora app with built-in safety features.

This release emphasizes safety in AI-generated video content, addressing privacy and consent while enhancing creative expression.

AIvideo generationsafetyconsent+6
FeatureMar 20, 2026

NVIDIA

Build a Domain-Specific Embedding Model in Under a Day

NVIDIA releases a tutorial for fine-tuning embedding models for domain-specific applications.

This release simplifies the process of creating domain-specific embedding models, making it accessible for enterprises to enhance their retrieval systems with minimal resources.

NVIDIAembedding modelfine-tuningsynthetic data+6
FeatureMar 20, 2026

IBM

What's New in Mellea 0.4.0 + Granite Libraries Release

Release of Mellea 0.4.0 and three Granite Libraries.

These releases improve the maintainability and predictability of AI workflows, enhancing the accuracy and safety of AI models.

IBMMelleaGranite LibrariesAI Workflows+5
BenchmarkMar 19, 2026

NVIDIA

Introducing SPEED-Bench: A Unified and Diverse Benchmark for Speculative Decoding

SPEED-Bench is a new benchmark for evaluating speculative decoding (SD) in AI models.

SPEED-Bench provides a comprehensive and realistic evaluation framework for speculative decoding, enabling better analysis and comparison of SD algorithms and models.

NVIDIABenchmarkSpeculative DecodingAI Models+5
ModelMar 17, 2026

H Company

Holotron-12B - High Throughput Computer Use Agent

Holotron-12B is a multimodal computer-use model optimized for scale and performance.

Holotron-12B offers significant improvements in inference efficiency and agent performance, making it a strong choice for throughput-bound workloads in real-world multimodal applications.

AIModelMultimodalInference+6
ModelMar 17, 2026

OpenAI

Introducing GPT-5.4 mini and nano

GPT-5.4 mini and nano are optimized for coding, tool use, and multimodal reasoning.

These models provide efficient solutions for coding and multimodal tasks, offering significant improvements in speed and cost-effectiveness for developers.

OpenAIGPT-5.4mininano+6
FeatureMar 13, 2026

NVIDIA

NVIDIA KGMON (NeMo Agent Toolkit) Data Explorer

NVIDIA introduces the KGMON (NeMo Agent Toolkit) Data Explorer for dataset exploration and analysis.

This release enhances the capabilities of autonomous data analysis agents, enabling them to handle complex data queries more efficiently.

NVIDIAData AnalysisAI AgentsBenchmark+6
BenchmarkMar 12, 2026

NVIDIA

How NVIDIA AI-Q Reached #1 on DeepResearch Bench I and II

NVIDIA AI-Q achieved first place on DeepResearch Bench I and II.

Leading on both benchmarks demonstrates AI-Q's capability to produce high-quality, well-cited research reports with strong retrieval and reasoning abilities.

NVIDIAAI-QDeepResearch BenchNeMo Agent Toolkit+5
FeatureMar 11, 2026

OpenAI

From model to agent: Equipping the Responses API with a computer environment

OpenAI introduces a computer environment for the Responses API.

This release enhances the capability of AI models to handle complex tasks by integrating a secure execution environment, expanding their utility beyond simple prompts.

OpenAIResponses APIAI AgentsShell Tool+4
APIMar 11, 2026

OpenAI

Rakuten fixes issues twice as fast with Codex

Rakuten uses OpenAI's Codex to reduce mean time to recovery by 50%.

This integration demonstrates how AI can significantly enhance software development efficiency and safety, enabling faster and more reliable product delivery.

OpenAICodexRakutenAPI+6
FeatureMar 10, 2026

Amazon

Accelerate custom LLM deployment: Fine-tune with Oumi and deploy to Amazon Bedrock

Fine-tune a Llama model using Oumi on Amazon EC2.

This release streamlines the process of customizing and deploying large language models on Amazon's infrastructure, enhancing accessibility and efficiency for developers.

AmazonLLMOumiAmazon Bedrock+5
FeatureMar 10, 2026

Hugging Face

Introducing Storage Buckets on the Hugging Face Hub

Hugging Face introduces Storage Buckets for managing ML artifacts.

This feature enhances the Hugging Face Hub by providing efficient storage solutions for intermediate ML artifacts, optimizing both speed and cost for users.

Hugging FaceStorageMachine LearningXet+6
ModelMar 10, 2026

OpenAI

Improving Instruction Hierarchy in Frontier LLMs

OpenAI introduces IH-Challenge, a training dataset for instruction hierarchy.

This release strengthens AI models' ability to handle conflicting instructions, enhancing safety and reliability in real-world applications.

AI SafetyInstruction HierarchyPrompt InjectionReinforcement Learning+4
FeatureMar 10, 2026

OpenAI

New ways to learn math and science in ChatGPT

ChatGPT introduces interactive visual explanations for math and science.

This feature enhances learning by allowing users to interact with and visualize complex concepts, improving understanding and retention.

OpenAIChatGPTInteractive LearningMath Education+6
FeatureMar 9, 2026

Amazon

Access Anthropic Claude models in India on Amazon Bedrock with Global cross-Region inference

Amazon Bedrock now supports Anthropic Claude models in India.

This release expands the availability of Anthropic Claude models, enabling developers in India to leverage advanced AI capabilities through Amazon Bedrock.

AmazonAnthropicClaudeAI+5
FeatureMar 9, 2026

Hugging Face

Ulysses Sequence Parallelism: Training with Million-Token Contexts

Ulysses Sequence Parallelism enables training on million-token contexts.

This feature allows for efficient training of large language models on extensive sequences, crucial for tasks like document analysis and complex reasoning.

UlyssesSequence ParallelismHugging FaceLong Contexts+5
ModelMar 9, 2026

IBM

Granite 4.0 1B Speech: Compact, Multilingual, and Built for the Edge

Granite 4.0 1B Speech is a compact, multilingual speech-language model.

Granite 4.0 1B Speech offers high performance in speech recognition and translation with fewer parameters, making it ideal for edge applications.

IBMGraniteSpeech RecognitionMultilingual+6
ModelMar 9, 2026

Hugging Face

LeRobot v0.5.0: Scaling Every Dimension

LeRobot v0.5.0 introduces full Unitree G1 humanoid support.

This release significantly expands LeRobot's capabilities, supporting more complex robotics tasks and improving real-world deployment responsiveness.

LeRobothumanoidroboticssimulation+6
FeatureMar 6, 2026

OpenAI

Descript enables multilingual video dubbing at scale

Descript uses OpenAI models to optimize multilingual video dubbing.

This release enables scalable and natural-sounding multilingual dubbing, addressing a major challenge in video localization.

DescriptOpenAImultilingualvideo dubbing+5
FeatureMar 6, 2026

OpenAI

Codex Security: now in research preview

Codex Security is an AI application security agent.

Codex Security enhances application security by providing high-confidence findings and actionable fixes, reducing the burden on security teams.

AISecurityVulnerability DetectionOpenAI+4
FeatureMar 6, 2026

Descript

Descript Enables Multilingual Video Dubbing at Scale

Descript uses OpenAI models to optimize multilingual video dubbing.

This release enhances the scalability and naturalness of multilingual dubbing, making it more accessible for large content libraries.

DescriptOpenAIMultilingualVideo Dubbing+5
FeatureMar 6, 2026

OpenAI

Descript Enables Multilingual Video Dubbing at Scale

Descript uses OpenAI models to optimize multilingual video dubbing.

This release enables high-quality, scalable video dubbing, making multilingual content more accessible and natural-sounding.

AIvideo dubbingmultilingualtranslation+6
FeatureMar 5, 2026

Hugging Face

Introducing Modular Diffusers - Composable Building Blocks for Diffusion Pipelines

Modular Diffusers allow building diffusion pipelines using reusable blocks.

This release provides developers with greater flexibility and customization in building AI workflows, enhancing efficiency and creativity.

AIDiffusionModularPipeline+6
ModelMar 5, 2026

OpenAI

Introducing GPT-5.4

GPT-5.4 is OpenAI's most capable and efficient model for professional work.

GPT-5.4 enhances professional workflows by providing more reliable, efficient, and accurate outputs, making it a valuable tool for developers and professionals across various industries.

OpenAIGPT-5.4AI ModelProfessional Work+5

Showing 130 of 60 releases