Skip to main content
​
3rd March 2026

​
Gemini 3.1 Flash Lite Preview Introduced

We introduced Gemini 3.1 Flash Lite Preview across AI Stats and rolled out support across product surfaces.Gemini 3.1 Flash Lite PreviewThis rollout includes:
  • Model discovery and availability updates in AI Stats
  • Gateway compatibility and routing coverage updates for Gemini 3.1 Flash Lite Preview
  • Documentation updates to reflect the latest Gemini 3.1 support
​
2nd March 2026

​
Responses API Assistant Phase Support

We added support for OpenAI Responses assistant message phase values:
  • Request handling now accepts assistant message phase in Responses input items (commentary or final_answer)
  • Gateway Responses payloads now preserve and return assistant message phase when present
  • OpenAPI documentation has been updated to document phase in Responses request and response message items

​
How to pass phase across turns

Use phase only on assistant messages:
  • commentary for intermediate assistant steps
  • final_answer for final assistant answers
  • Preserve and resend the original assistant message phase values on follow-up turns

​
Example: turn 1 request

{
  "model": "openai/gpt-5.3-codex",
  "input": [
    { "type": "message", "role": "user", "content": "Plan this refactor." }
  ]
}

​
Example: turn 1 response excerpt

{
  "output": [
    {
      "type": "message",
      "role": "assistant",
      "phase": "commentary",
      "content": [{ "type": "output_text", "text": "First, break this into modules..." }]
    },
    {
      "type": "message",
      "role": "assistant",
      "phase": "final_answer",
      "content": [{ "type": "output_text", "text": "Recommended approach: ..." }]
    }
  ]
}

​
Example: turn 2 request (resend assistant messages with phase)

{
  "model": "openai/gpt-5.3-codex",
  "input": [
    {
      "type": "message",
      "role": "assistant",
      "phase": "commentary",
      "content": [{ "type": "output_text", "text": "First, break this into modules..." }]
    },
    {
      "type": "message",
      "role": "assistant",
      "phase": "final_answer",
      "content": [{ "type": "output_text", "text": "Recommended approach: ..." }]
    },
    { "type": "message", "role": "user", "content": "Now generate the migration checklist." }
  ]
}
​
26th February 2026

​
Nano Banana 2 Introduced

We have introduced Nano Banana 2 across AI Stats, released as Gemini 3.1 Flash Image Preview.Nano Banana 2This rollout includes:
  • Model availability and discovery updates in AI Stats
  • Gateway compatibility coverage updates for Gemini 3.1 Flash Image Preview
  • Documentation refreshes to reflect the latest Nano Banana model support
​
19th February 2026

​
Gemini 3.1 Pro Preview Introduced

We have introduced Gemini 3.1 Pro Preview across AI Stats and surfaced it across our product experiences.Gemini 3.1 Pro PreviewThis rollout includes:
  • Model availability and discovery updates in AI Stats
  • Gateway coverage and pricing surface updates for Gemini 3.1 text-generation support
  • Documentation refreshes to reflect the latest Gemini 3.1 model support
​
18th February 2026

​
Claude Sonnet 4.6 Introduced

We have introduced Claude Sonnet 4.6 across AI Stats and surfaced it across our product experiences.Claude Sonnet 4.6This rollout includes:
  • Model availability and discovery updates in AI Stats
  • Gateway compatibility coverage updates for the latest Anthropic model line
  • Documentation refreshes to reflect the latest model support
​
30th January 2026

​
Major Overhaul Completion and Enhanced Compatibility

We have been tirelessly working on the next phase of AI Stats, and we’re thrilled to announce that this comprehensive overhaul has now been completed. This update represents a significant leap forward in our mission to provide a unified, scalable platform for AI model access and insights.

​
AI Stats Gateway gets another overhaul

The AI Stats Gateway has had another major overhaul behind the scenes, this has allowed us to introduce compatibility with Anthropic Messages API, giving another way for you to use our gateway. We have also enhanced the way that operations are done, meaning it should be faster, and add less latency than ever before, allowing you to build even more! We also are now confident in this design to allow us to expand rapidly, and across new modalities without major changes being required. We hope that we can continue to expand the gateway comprehensively in the coming month or two.

​
Database Overhaul for Long-Term Expansion

At the core of this update is a complete redesign of our database architecture. We’ve implemented a more flexible and robust schema that ensures our gateway can scale seamlessly to support an ever-growing number of models across all modalities - text, image, video, audio, moderations, embeddings, and beyond. This foundation should allow us to rapidly expand our coverage while maintaining high performance and reliability. This also ties into the Gateway overhaul, ensuring we have the data to offer the best developer experience possible - whilst also being part of our longer term plans to move the data from static JSON files into the database for accuracy, this does not mean we won’t accept contributions via GitHub, we always will!

​
OpenAI & Anthropic Compatibility with New /messages Endpoint

Building on our existing OpenAI compatibility through the /chat/completions and /responses endpoints, we’re now introducing Anthropic compatibility with our new /messages endpoint. This endpoint provides a standardized interface for chat-based interactions, making it easier than ever for developers to integrate with our gateway regardless of their familiarity with specific provider APIs. The /messages endpoint supports advanced features like streaming responses, tool calling, and multi-modal inputs, ensuring a smooth transition for existing OpenAI or Anthropic users. We have done our best to make things completely interoperable, however, please be aware there are some differences in how certain features work between providers, so please review the documentation carefully and report anything that seems off!

​
Website Improvements for Better User Experience

To make AI Stats the “one stop shop” platform for staying up to date with AI developments, we’ve overhauled key sections of the website. These improvements include:
  • Optimized load times and mobile responsiveness for on-the-go access
  • Improved search and filtering capabilities across all data sets
  • Introduced a Playground for easily testing models through the gateway in a simple but powerful interface - more features planned for this!
  • Enhanced model pages with clearer pricing, availability, and performance metrics
These changes create a more intuitive and efficient experience, helping users quickly find the information they need and make informed decisions about AI models.We would appreciate any feedback on these changes or any parts of the website, so please get in touch if you have any suggestions!

​
SDK Developments

We’re overhauling our in house SDKs to move away from AI SDK-style integrations and instead develop our own native implementations. This shift enables us to create a unified interface across all our SDKs, allowing developers to interact with any supported model through a single, consistent method. We’re excited about the possibilities unlocked by our gateway-powered SDKs and believe that making a wide range of models accessible across multiple programming languages will empower the community to build truly innovative applications. We also are working on our own AI SDK (Vercel) to allow even easier integration with our gateway, more details on this will be shared soon!

​
Exploring New Endpoints for Expanded Model Access

To maximize access to cutting-edge AI models, we’re actively exploring and prototyping new endpoints. These will include specialized interfaces for emerging modalities like video generation, embeddings, and audio processing as well as lesser accessible modalities like music and high quality speech generation. Our goal is to provide developers with the broadest possible range of AI capabilities through a single, unified gateway.Our ultimate aspiration is to support every model in our database through the gateway, making one of the most powerful technologies available as accessible as possible. We aim to enable seamless model switching, democratizing AI for developers and researchers across the globe.This update marks a pivotal moment in AI Stats’ evolution. We’re committed to continuing our rapid pace of innovation, driven by community feedback and the ever-advancing AI landscape. Stay tuned for more updates as we roll out these new features and expand our model coverage.Thanks for being part of this journey with us!
​
15th December 2025

​
SDKs Alpha Release

Our SDKs are now available in an early alpha stage in V0.2.0. We initially released them with the expectation that the OpenAPI Generator would work seamlessly, but we quickly realized it wasn’t performing as expected and needs significant review and improvement. The SDKs are still available for use, but we want to be transparent that they are in alpha and may require updates as we refine the generation process.If anyone can help with testing these or has experience with OpenAPI generators, we would be very grateful for your input. We want to build a world-class Developer Experience, where AI becomes something so trivial to use, it can easily be added, experimented, used for research, all through our single platform.We have decided for now to remove the C++ and Rust SDKs from our plans. We might add them back in the future should there be demand, but we are focusing on a smaller set of languages whilst we find the best API generators that we can use and make sure the most used languages are supported. We hope to have all SDKs moved to a V1 ASAP - targeting no later than mid-January 2026.We also plan to unify all SDK versions alongside the API and Website + Docs at some point in the future for better consistency.

​
Python SDK Alpha

pip install --upgrade ai-stats-py-sdk==0.2.0a1

​
TypeScript / JavaScript SDK Alpha

npm install @ai-stats/sdk@0.2.0-alpha.1
​
12th December 2025

​
Improving Our API

​
API

The API has received a major breaking change update, introducing OpenAI Compatibility, with the /responses endpoint added for unified model responses supporting text, images, and other modalities. We have done this to align with the OpenAI API structure, making it easier for us to add more models, whilst also making it easier for developers to integrate with our gateway if they are already familiar with the OpenAI API. We will always maintain a superset of the OpenAI API to ensure we can support all models and features available on our gateway.Additionally, the existing /chat/completions, /moderations and /embeddings endpoints have been updated to align with the OpenAI API structure. Other endpoints will be coming soon to provide full OpenAI API compatibility while maintaining the gateway’s unified interface and metadata. To be transparent about our future plans, we have added placeholder endpoints and their corresponding documentation. However, models for these endpoints have yet to be enabled. We are looking to bring as many of these endpoints online as soon as possible, providing as many models and as many providers for these models too.

​
SDKs

The SDKs have been updated to reflect breaking changes made to the API, this means if you are using the SDKs, you will need to update to the latest major versions and adapt. If any help is needed with this migration, please get in touch and we will be happy to help!We have also made some substantial progress on SDKs for other languages, with initial versions hoped to be released by the end of 2025 - these will likely be spotted around the website, and documentation, whilst they aren’t all available yet, we are working as hard as we can to get them available as soon as possible.

​
Updating SDKs

Due to breaking changes, please update your SDKs to the latest versions:

​
Python SDK

pip install --upgrade ai-stats-py-sdk

​
TypeScript / JavaScript SDK

npm update @ai-stats/sdk

​
Reviewing API Implementations

This is a major milestone on our journey to a V1 API. We hope this will be the last breaking change to the API as we take the decision to align with OpenAI’s Standard. Please review your REST API implementations and refer to the updated API reference for the latest schema and endpoints. The changes are rather minor, however, please review them.

​
Website

The website has undergone significant updates to the Model Gateway Page and Model Pricing Page. The Model Gateway Page is now a Quickstart page, supporting more code samples to get up and running quicker, as well as trimming of unnecessary information on the page. This information has largely been moved to the updated Model Pricing Page, which has been enhanced to provide clearer pricing information and set up for future enhancements across all providers.We have also added full Cookie Preferences for better user control over privacy settings.Additionally, we fixed an issue where Model Updates were incorrectly showing as “Today” even when they occurred on historical dates due to our Upgrade to Next.js 16.We have also tweaked Web Watchers to be more useful.

​
Documentation

We have expanded the SDK documentation in preperation for launching these shortly, with more examples and improved guides for each endpoint to help developers integrate our APIs more easily.

​
Infrastructure

We have implemented a Nightly CI process to automatically update model statuses based on their release dates, ensuring the data remains current and accurate.
​
5th December 2025

​
UI Improvements

Dark Mode has been implemented, offering users a darker theme to reduce eye strain and improve comfort in low-light environments.
​
1st December 2025

​
Introducing The New AI Stats

We are excited to announce a major update to AI Stats. This is the culmination of months of work to bring a major overhaul to our website and offering. Today, we are introducing a unified model gateway, massively expanded data coverage, and a revamped user experience, as well as much more. This release marks a significant step towards our vision of making AI integration a seamless experience for developers and researchers alike.

​
Rebranding

We have rebranded slightly, with adjusted logos and wordmarks to better reflect our mission and vision. The new design is more modern, clean, and versatile, allowing us to better represent our brand across various platforms and media. We hope you like the new look as much as we do!

​
Data coverage

At the heart of AI Stats is our database. It is designed to be a highly expansive and flexible store of model data. If you have used AI Stats before, you may know we only used to have text-based models. However, we always wanted to change that, so today we are announcing that we have grown our entire data coverage dramatically to allow us to track more models, such as image, video, and embedding models, in as much detail as possible.There is still a long road to capture every organisation, model, and benchmark result. We currently sit at ~30% coverage, but we are prioritising expanding this as quickly as possible and aim to reach at least 60% over the next year. We want our database to be the gold standard, whilst still keeping it open source so anyone can contribute. We always welcome contributions to help us get there faster!

​
Model Availability

With the introduction of the AI Stats Gateway, we have also reworked how we track model availability. A full subscription plan section now allows you to see all the subscription plans we track, the models they include, and the features they offer, as well as clearly seeing the price. We have also adjusted how we display availability to make it clearer and easier to see where models are available. We plan to continue tweaking this section to make it as useful as possible.

​
Data Overhaul

We have revisited what data we store, how we store it, and how we present that data, so we have built a smoother, more reliable experience across every page. It is not perfect yet, and we have plenty of optimisations ahead. We are working hard to adjust certain pages to make them clearer and easier to use. We know there is more to do, and we plan to make trends easier to track, comparisons easier to make, and data more accessible. If you have any feedback on how we can improve - whether that is certain pages and their designs or layouts, or even tracking new data points - please reach out!

​
User Experience Improvements

We have redesigned many parts of the website to make it easier to navigate, find information, and understand the data presented. We have worked on improving load times and responsiveness, so you can keep up to date on the go, as well as overall usability. We are committed to making AI Stats a joy to use, no matter how much data we have, and we will continue to refine the user experience based on your feedback.

​
Benchmark pages

We have revamped and overhauled the benchmark pages to make them clearer and easier to see which models are performing best, and how scores were achieved, with more detailed information on the methodology used. We plan to continue to work on these pages to make them as easy as possible to interpret, and to filter the data so you can conduct your own analysis.

​
API provider pages

We have also redesigned the API provider pages to give a clear overview of the performance data we get from our gateway, so you can see exactly how each provider is performing and which models are being used most. We then also give a breakdown of each set of models provided by the provider, so you can see what models are available per provider, as well as easily see if they are available on the gateway. We plan to adjust this shortly to include detailed pricing information, so you can easily compare models’ pricing.

​
Updates

I am excited to announce an expansion of the Latest Updates section on the homepage. This section will now feature not only the latest model changes, but also the latest news and YouTube releases, alongside the model updates. With new pages inside the Updates section too, there has never been a better place to stay updated in the ever-changing world of AI! Stay tuned for even more updates to this section in the near future, as we plan to add more to keep you up to date with everything AI and AI Stats.

​
Our Model Gateway - Alpha

We are also introducing a brand-new unified gateway to use any supported AI model with a single API interface. We are expanding coverage as fast as possible while keeping the service reliable at scale - please stress-test it and tell us what breaks. The integration steps are all covered per model page, and you can find the full API reference in our docs.We would not yet recommend the API for production use, but we are excited to get feedback from the community to help us improve it rapidly. We will take any feedback as seriously as possible to make the gateway as reliable and useful as possible.
Please also be aware we are yet to fully finalise the schema for requests and responses, so these may change in the coming weeks.

​
SDKs - Alpha

Today, we are also announcing official SDKs for Python and TypeScript/JavaScript. Pick your preferred package manager and run one of the commands below to get up and running.

​
Python SDK

The Python SDK provides a simple interface to interact with the Model Gateway, manage API keys, and handle requests/responses.
pip install ai-stats-py-sdk

​
TypeScript / JavaScript SDK

The TypeScript/JavaScript SDK makes it easy to integrate AI models into web and backend applications, with first-class async support and built-in error handling.
npm install @ai-stats/sdk
We plan to expand SDK support to other languages based on community feedback, so please let us know what you would like to see next!

​
Our Future Plan

Right from the start of AI Stats, we believe that a completely public, open source approach is the best way to build a reliable, trustworthy service. We want to continue this with our new version of AI Stats, which is why the entire source code for the website, gateway, and database is available on GitHub. We welcome contributions, feature ideas, and performance testing - help us push the limits of infra together to build a better gateway, and a better AI Stats. Open to everyone, built by the community.We plan to currently work at a rapid pace to expand data coverage, refine the user experience and expand our gateway. We have a public roadmap of features and improvements we want to make, but we are always open to feedback from our users to help us prioritise what matters most to them.
Last modified on March 3, 2026