AI model tracker
The latest and most popular AI models for image, video, music and text, each linked to the maker’s own announcement.
Last checked October 18, 2026. Models change fast, so always confirm details on the linked source before you rely on them. Disclosure: Anthropic’s Claude is one of the tools used to help build this site, and Claude models are listed here on the same basis as every other maker.
Image
| Model | Maker | Released | Status | What it is |
|---|---|---|---|---|
| Nano Banana 2.1 | Google DeepMind | October 6, 2026 | Current | Image model built on Gemini 3.6 Flash that takes text and images in and outputs images, with a model card listing the Gemini app, Google AI Studio, the Gemini API, Search AI Mode, Google Ads, Flow and Stitch. Source: Google DeepMind model card ↗ |
| FLUX 3 Image | Black Forest Labs | October 1, 2026 | Current | Generate and edit from a single endpoint, place elements with bounding boxes written into the prompt, combine up to ten reference images, and output up to 4K. Source: Black Forest Labs release notes ↗ |
| GPT Images 2.5 (ChatGPT Images 2.5) | OpenAI | September 8, 2026 | Current | OpenAI's latest image update in ChatGPT and the API, described as faster and sharper, with better fidelity and tools for creating images. Source: OpenAI announcement ↗ |
| Midjourney V8.2 | Midjourney | July 24, 2026 | Current | The default model since July 24, 2026, focused on aesthetics, image quality and personalization. An Edit Model followed on August 27 for editing with instructions, up to four image references, inpainting and outpainting. Source: Midjourney updates ↗ |
| Ideogram 4.0 | Ideogram | June 3, 2026 | Current | An open-weight image model with a commercial license, focused on dense multilingual text, bounding-box layout control and 2K output. Source: Ideogram news ↗ |
| DALL-E 2 and DALL-E 3 | OpenAI | May 12, 2026 (API shutdown date) | Retired | Removed from the OpenAI API on May 12, 2026. OpenAI recommends gpt-image-2, gpt-image-1 or gpt-image-1-mini instead. Source: OpenAI deprecations ↗ |
| GPT Image 2 | OpenAI | April 21, 2026 | Current | OpenAI's image model in the API and Codex, with stronger editing, better layouts and text rendering, and output up to 2K. Source: OpenAI developer docs ↗ |
| Nano Banana 2 | Google DeepMind | February 26, 2026 | Current | Google's image model that combines Nano Banana Pro capabilities with Gemini Flash speed, available in the Gemini app, Search and the Gemini API. Source: Google blog ↗ |
| FLUX.2 family | Black Forest Labs | November 25, 2025 | Current | Variants for different needs: max, pro and flex through the API, a 32B open-weight dev model, and fast klein models, with multi-reference support for up to ten images. Source: Black Forest Labs ↗ |
Video
| Model | Maker | Released | Status | What it is |
|---|---|---|---|---|
| Sora | OpenAI | September 24, 2026 (API shutdown date) | Retired | The Sora app and website closed on April 26, 2026, and the Sora 2 models and Videos API were removed from the API on September 24, 2026. Source: OpenAI deprecations ↗ |
| FLUX 3 Video | Black Forest Labs | August 4, 2026 | Preview | BFL's first video model, with synchronized audio, clips up to 20 seconds, four modes on one endpoint and, since September 10, 2K and 4K output. Source: Black Forest Labs release notes ↗ |
| MiniMax H3 (Hailuo) | MiniMax | July 31, 2026 | Current | A general-purpose multimodal model that generates video with native stereo sound, up to 15 seconds at 2K. MiniMax describes it as an open model. Source: MiniMax research ↗ |
| Runway Gen-4.5 and Aleph 2.0 | Runway | May 21, 2026 (Aleph 2.0 launch) | Current | Gen-4.5 makes 2 to 10 second clips from text or images (API since February 10). Aleph 2.0 edits existing footage up to 30 seconds at 1080p. Source: Runway news ↗ |
| Seedance 2.0 | ByteDance Seed | February 12, 2026 | Current | A multimodal audio-video model that takes text, image, audio and video inputs and generates clips of 4 to 15 seconds with sound. Source: ByteDance Seed ↗ |
| Kling 3.0 | Kuaishou | February 5, 2026 | Current | Video, Image and Omni models with clips up to 15 seconds and native multilingual audio. Kling 4.0 Flash opened to limited early access on September 28, 2026, with a wider 4.0 rollout planned for October. Source: Kuaishou and Kling ↗ |
| Veo 3.1 | Google DeepMind | See source | Current | Google's video model, producing 8-second clips at 720p, 1080p or 4K with native audio, plus a lower-cost Lite tier. Available in the Gemini app, Flow, YouTube Shorts, Vids, the Gemini API and Vertex AI. Source: Google DeepMind ↗ |
Music and voice
| Model | Maker | Released | Status | What it is |
|---|---|---|---|---|
| Eleven v4 and v4 Turbo | ElevenLabs | September 28, 2026 | Current | Text-to-speech models for expressive narration (v4) and low-latency agents (v4 Turbo), supporting more than 90 languages. Source: ElevenLabs blog ↗ |
| Suno v6 | Suno | September 9, 2026 | Current | A new model generation developed with Warner Music Group, BMG and Believe. It comes as v6, v6-wild and the faster v6-mini, which is available to everyone. Source: Suno blog ↗ |
| ElevenLabs Music v2.5 | ElevenLabs | September 2026 (September 2026) | Current | The new default music model in ElevenMusic. ElevenLabs says you own what you make on every plan, including Free, with lossless downloads on all plans. Source: ElevenLabs blog ↗ |
| Stable Audio 3.0 | Stability AI | May 20, 2026 | Current | A family trained on licensed data. Small, Small SFX and Medium are open weights on Hugging Face; Large is available through the API. Medium reaches about six minutes per track. Source: Stability AI ↗ |
| Lyria 3.5 | See source | Current | Google's music model for full-length songs with verses, choruses and bridges, producing 44.1 kHz stereo audio from text or up to ten images through the Gemini API. Source: Gemini API docs ↗ | |
| GPT-Live-1 | OpenAI | See source | Current | OpenAI's live voice model in the API. In October 2026 it gained native Decisions integration, prompt-to-voice custom voices and nine new voices in German, French, Hindi, Japanese and Korean. Source: OpenAI API changelog ↗ |
Text
| Model | Maker | Released | Status | What it is |
|---|---|---|---|---|
| Claude Haiku 5.5 | Anthropic | October 7, 2026 | Current | Anthropic's cheapest and fastest small model, with an adjustable effort setting, priced at $0.10 per million input tokens and $0.50 per million output tokens for prompts up to 100,000 tokens. Source: Anthropic ↗ |
| Mistral Large 4 | Mistral AI | October 6, 2026 | Preview | A 1 trillion parameter multimodal model with 49 billion active parameters, in public preview. Mistral says the open weights will be released by the end of October. Source: Mistral AI ↗ |
| Gemini 4 Argon | Google DeepMind | September 30, 2026 | Limited rollout | Google's new frontier model, rolling out first to trusted cyber defenders through its Fairwind Program, with wider release planned after safety testing. Source: Google blog ↗ |
| GPT-6.1 Sol | OpenAI | September 29, 2026 | Current | Available in ChatGPT and the API as gpt-6.1-sol. OpenAI says it nearly matches GPT-6 Astra on agentic coding and professional work at one-fifth of Astra's standard token prices. An Ultrafast mode for the Responses API followed on October 8. Source: OpenAI announcement ↗ |
| Claude Sonnet 5.5 | Anthropic | September 28, 2026 | Current | The second model in the Claude 5.5 family, priced at $2 per million input tokens and $10 per million output tokens. Opus 5.5 is also available; Haiku 5.5 was announced as coming in the following weeks. Source: Anthropic ↗ |
| DeepSeek V4.1-Flash | DeepSeek | September 10, 2026 | Current | A 552 billion parameter mixture-of-experts model with native visual understanding, available in the DeepSeek API as deepseek-flash. DeepSeek is phasing out V4-Pro. Source: DeepSeek API docs ↗ |
| Gemini 3.5 Flash | Google DeepMind | May 19, 2026 | Current | The first model in the Gemini 3.5 series, built for agents and coding and available in the Gemini app, AI Mode in Search and developer platforms. Source: Google blog ↗ |
Missing a model, or spotted a mistake? Tell us. We update this page as announcements land and list the maker’s own page as the source.