AI Center › AI Model Registry
1ArtificialIntelligence · The Global AI Ecosystem — living dataset
AI Model Registry
The AI Model Registry is a live view of notable AI models — their developer, release date, parameters and training compute — drawn from the Epoch AI dataset and shown as Epoch reports them, each figure carrying its own as-of date.
It is a displayed-not-interpreted record, not a leaderboard. Models are listed by publication date, not ranked; the underlying figures and classifications are Epoch AI's, used under CC BY 4.0, and each model cites the source Epoch records for it wherever that source is a working link.
Ordered by publication date, most recent first. This is a temporal order, not a ranking: the register does not rank models by size, training compute, or capability. Figures and classifications are Epoch AI’s, displayed as reported.
| Model | Organization | Published | Parameters | Training compute (FLOP) | Accessibility |
|---|---|---|---|---|---|
| GLM-5.3 | Z.ai (Zhipu AI) | 2026-08-14 | 7.44e11 | — | API access |
| Qwen 3.8 27B | Alibaba | 2026-08-14 | 2.70e10 | — | Open weights (unrestricted) |
| DeepSeek-V4-Pro-0813 | DeepSeek | 2026-08-13 | 1.60e12 | — | Open weights (unrestricted) |
| Gemini 3.7 Flash | Google DeepMind | 2026-08-13 | — | — | API access |
| Grok 4.6 | xAI | 2026-08-12 | — | — | API access |
| GPT-5.6 Cyber | OpenAI | 2026-08-11 | — | — | API access |
| GPT-5.5 Cyber | OpenAI | 2026-08-11 | — | — | API access |
| Nemotron 3.5 Lightning | NVIDIA | 2026-08-11 | 3.00e10 | — | Open weights (unrestricted) |
| Muse Glimmer | Meta AI | 2026-08-10 | 3.00e10 | — | Open weights (unrestricted) |
| Motif-3 | Motif Technologies | 2026-08-07 | 3.14e11 | 1.10e24 | Open weights (unrestricted) |
| Muse Spark 1.2 | Meta AI | 2026-08-05 | — | — | API access |
| DeepSeek V4 Flash 0731 | DeepSeek | 2026-07-31 | 2.84e11 | 2.50e24 | Open weights (unrestricted) |
| K-EXAONE 2.0 | LG AI Research | 2026-07-31 | 7.50e11 | 3.55e24 | Open weights (unrestricted) |
| Inkling-Small | Thinking Machines | 2026-07-30 | 2.76e11 | — | Open weights (unrestricted) |
| Gemini Robotics ER 2 | Google DeepMind | 2026-07-30 | — | — | API access |
| GPT Transcribe | OpenAI | 2026-07-29 | — | — | API access |
| GPT Live Transcribe | OpenAI | 2026-07-29 | — | — | API access |
| A.X K2 | SK Telecom | 2026-07-29 | 6.88e11 | 1.80e24 | Open weights (unrestricted) |
| Qwen3.7 Flash | Alibaba | 2026-07-27 | — | — | API access |
| Claude Opus 5 | Anthropic | 2026-07-24 | — | — | API access |
| Gemini 3.6 Flash | Google DeepMind | 2026-07-21 | — | — | API access |
| Gemini 3.5 Flash-Lite | Google DeepMind | 2026-07-21 | — | — | API access |
| Gemini 3.5 Flash Cyber | Google DeepMind | 2026-07-21 | — | — | Unreleased |
| Qwen 3.8 Max | Alibaba | 2026-07-19 | 2.40e12 | — | API access |
| Kimi K3 | Moonshot | 2026-07-16 | 2.80e12 | 2.00e25 | Open weights (non-commercial) |
| Inkling | Thinking Machines | 2026-07-15 | 9.75e11 | 1.84e24 | Open weights (unrestricted) |
| GPT-5.6 Sol | OpenAI | 2026-07-09 | — | — | API access |
| GPT-5.6 Terra | OpenAI | 2026-07-09 | — | — | API access |
| GPT-5.6 Luna | OpenAI | 2026-07-09 | — | — | API access |
| Muse Spark 1.1 | Meta AI | 2026-07-09 | — | — | API access |
| Grok 4.5 | xAI | 2026-07-08 | 1.50e12 | — | API access |
| SWE-1.7 | Cognition | 2026-07-08 | — | — | — |
| Muse Image | Meta AI | 2026-07-07 | — | — | Hosted access (no API) |
| GPT-Realtime-2.1-Mini | OpenAI | 2026-07-06 | — | — | API access |
| Tencent Hy3 | Tencent | 2026-07-06 | 2.95e11 | — | Open weights (unrestricted) |
| Claude Sonnet 5 | Anthropic | 2026-06-30 | — | — | API access |
| Solar Open2 250B | Upstage | 2026-06-28 | 2.50e11 | 1.05e24 | Open weights (unrestricted) |
| Doubao Seed 2.1 Turbo | ByteDance | 2026-06-23 | — | — | API access |
| Doubao Seed 2.1 Pro | ByteDance | 2026-06-23 | — | — | API access |
| GLM-5.2 | Z.ai (Zhipu AI) | 2026-06-16 | 7.44e11 | — | Open weights (unrestricted) |
How this registry is maintained
- The data is drawn from the Epoch AI Notable AI Models dataset, a single canonical CSV, used under the Creative Commons Attribution (CC BY 4.0) licence.
- The feed is fetched, checked and snapshotted off the live site; only a verified snapshot is shown. A failed or malformed fetch keeps the last good snapshot rather than publishing an error.
- Figures — parameters, training compute in FLOP, release date — are Epoch AI's own values. Nothing is re-estimated, re-aggregated or re-scored; parameters and training compute are displayed in scientific notation to three significant figures for readability, so the page shows 7.44e11 where the dataset records 744,000,000,000.
- Models are listed by publication date. They are not ranked by size, compute or capability, and the page presents no assessment of which model is better.
- Each entry carries its own as-of date; the dataset's own freshness stamp is shown in the source line above the table.
Where does this data come from?
The AI Model Registry draws on the Epoch AI Notable AI Models dataset, an openly licensed research dataset of notable machine-learning models. Epoch AI publishes it as a single downloadable CSV under the Creative Commons Attribution (CC BY 4.0) licence.
Epoch AI is a research organisation that documents trends in machine learning, including the compute, parameters and datasets used to train notable models. This page shows a dated view of that dataset; the figures shown are Epoch AI's, not 1BusinessWorld's, and are reproduced with attribution as the licence requires. Where Epoch AI records a usable source for a model — usually the developer's own announcement or the model's paper — the model name links to it. A model name appears as plain text in two cases: Epoch's source field holds no web address for that model, or the address it holds no longer resolves, in which case the row's data is kept and the dead link removed rather than left to fail. Links removed for that second reason are marked on the row and listed beneath the table.
What is training compute, measured in FLOP?
Training compute is the total amount of computation used to train a model, measured in floating-point operations (FLOP). Epoch AI records an estimate of training compute for models where it can be determined, and it is shown here as Epoch reports it.
Training compute has become a common way to describe the scale of a model's training run. In EU law, the amount of computation used for training is also a regulatory threshold: Article 51 of Regulation (EU) 2024/1689 presumes a general-purpose AI model to have systemic-risk capabilities when the cumulative computation used for its training, measured in floating-point operations, is greater than 10^25. The figures on this page are estimates published by Epoch AI and are displayed without adjustment.
How current is this registry?
This view was retrieved from Epoch AI on 2026-08-19T06:18:40Z and shows the 40 most recently published models in the dataset, the newest dated 2026-08-14. The full dataset holds 3,592 notable models and is updated by Epoch AI on a rolling basis.
The registry is refreshed from Epoch AI's canonical dataset and shows the dataset's own most-recent publication date as its freshness stamp. Because the dataset is curated by Epoch AI to its own inclusion criteria, this page attributes the selection and figures to Epoch AI and does not add or re-weight entries.
Cite this dataset
1BusinessWorld AI Center, "AI Model Registry" (data: Epoch AI, CC BY 4.0), updated daily. https://1businessworld.com/ai-center/ai-model-registry/ Version as of August 19, 2026.
The AI Center is informational only. It is provided by 1BusinessWorld strictly for general informational and educational purposes. Nothing in the AI Center constitutes, or should be construed as, legal, regulatory, compliance, technical, engineering, security, investment, financial, or other professional advice, or a recommendation, endorsement, solicitation, or offer regarding any technology, product, model, provider, framework, or course of action. 1BusinessWorld is not a law firm, regulatory authority, standards body, conformity-assessment or certification body, or investment adviser, and nothing in the AI Center creates any advisory, fiduciary, attorney-client, or other professional relationship with 1BusinessWorld. Although the AI Center references official materials published by legislatures, regulators, standards bodies, research organizations, and other named authorities, 1BusinessWorld makes no representation or warranty, express or implied, as to the accuracy, completeness, timeliness, or fitness for any purpose of any content, and, to the fullest extent permitted by law, disclaims all liability for any loss or damage of any kind arising directly or indirectly from the use of, or reliance on, any information presented. Laws, regulations, standards, technical practices, and AI capabilities change frequently and differ by jurisdiction; readers must verify all information against the current official text or source and consult qualified legal, compliance, technical, and other professional advisors before acting. Any decision relating to the development, deployment, procurement, or governance of AI systems is made solely at the reader's own risk. Last reviewed: August 19, 2026.