New AI models by release month
What the vendors shipped recently and what you can already run here. Every model on this list works in Studio and with the same API key.
Newest month: October 2026 · past 3 months: video 4, images 8, chat 7
October 2026
4 models| Model | Vendor | Type | Price |
|---|---|---|---|
| FLUX 3 ImageSwap objects between photos and place products exactly where you want them, from up to ten reference images | Black Forest Labs | Image | from $0.0984 / image |
| MAI Image 2.6Clean, photo-studio style pictures and reference-guided edits from Microsoft's own image model. | Microsoft AI | Image | from $0.0934 / image |
| Nano Banana 2.1Google's follow-up to Nano Banana 2, where a 4K picture costs exactly what a 1K one does. | Image | $0.0699 / image | |
| Claude Sonnet 5.5The newest Sonnet for daily coding and tool-calling agents, at a lower token rate than Sonnet 5 | Anthropic | Chat & code | $1.89 / $9.45 · 1M tok |
September 2026
9 models| Model | Vendor | Type | Price |
|---|---|---|---|
| GPT Image 2.5Fast OpenAI images for ads and product shots with readable text, from a prompt or up to 16 reference photos | OpenAI | Image | $0.05 / image |
| GPT Image 2.5 SunburstEdit the same picture again and again while the face, product or layout stays exactly as it was | OpenAI | Image | $0.05 / image |
| Qwen Image 2.1Alibaba's image model for placing products into new scenes from several references, prompted in any language. | Alibaba | Image | from $0.05 / image |
| Seedream 5.0 FlashByteDance's image model that edits just the area you mark and charges one price from 1K to 2K. | ByteDance | Image | $0.0343 / image |
| Claude Opus 5.5Near top-tier Claude quality for repo-wide changes and long coding sessions, at a lower rate than Fable | Anthropic | Chat & code | $3.78 / $18.90 · 1M tok |
| GPT-6 SolThe GPT-6 tier for hard work tasks you want to iterate on many times without paying flagship rates | OpenAI | Chat & code | $2.16 / $10.80 · 1M tok |
| Claude Fable 5.1The Claude to call when a task has already beaten cheaper models and a wrong answer would cost you real money | Anthropic | Chat & code | $9.45 / $47.25 · 1M tok |
| GPT-6 AstraThe strongest GPT-6 model, for research, long software projects and answers that must be right first time | OpenAI | Chat & code | $10.80 / $54.00 · 1M tok |
| GPT-6 LunaQuick, cheap GPT-6 for chatbots and message routing, with a context window as large as the pricier tiers | OpenAI | Chat & code | $0.108 / $0.54 · 1M tok |
August 2026
6 models| Model | Vendor | Type | Price |
|---|---|---|---|
| Wan 3.0Alibaba's video model for half-minute clips in one pass, built from your images, clips and voices. | Alibaba | Video | from $0.0777 / s |
| Gemini Omni 1.1Short 16:9 or 9:16 clips from a prompt, a start image, three fused images or a reference video, up to 4K | Video | from $1.02 / video | |
| MiniMax H3 MaxA lighter MiniMax H3 for animating a still or bridging two frames into a clip of up to 15 seconds. | MiniMax | Video | from $0.104 / s |
| Wan 3.0 Video PrimeThe Wan 3.0 lane for when someone is waiting: same inputs, same 30-second and 4K ceiling, shorter turnaround. | Alibaba | Video | from $0.124 / s |
| Grok Imagine Image 2.0xAI images that use your prompt word for word, up to four per call, for precise tuning and quick variants | xAI | Image | $0.045 / image |
| DeepSeek V4 ProDeepSeek's big open-weight mixture-of-experts model for agent work and very long outputs, billed per token | DeepSeek | Chat & code | $0.54 / $1.22 · 1M tok |
July 2026
12 models| Model | Vendor | Type | Price |
|---|---|---|---|
| MiniMax H32K clips up to 15 s with built-in stereo sound and references that keep faces, motion and voices consistent. | MiniMax | Video | from $0.161 / s |
| Seedance 2.5ByteDance's long-clip model: up to 30 seconds, 30 reference images, and tools to edit or lengthen footage. | ByteDance | Video | from $0.261 / s |
| FLUX 3 VideoClips of 5 to 20 seconds with built-in audio, steered by up to ten keyframes or continued from your own video | Black Forest Labs | Video | from $0.344 / s |
| Seedance 2.5 EcoThe budget lane of Seedance 2.5 for long clips, edits and extensions that can render in the background. | ByteDance | Video | from $0.44 / s |
| Seedream 5.0 ProByteDance's top Seedream for posters, covers and key visuals where the title is part of the image. | ByteDance | Image | from $0.0783 / image |
| Qwen Image 3.0Alibaba's third Qwen image model, able to lay out an entire page with text readable down to 10 px. | Alibaba | Image | from $0.0576 / image |
| Qwen Image 3.0 ProThe pro tier of Qwen Image 3.0, tuned for pages full of small type that all has to read correctly. | Alibaba | Image | from $0.0857 / image |
| Gemini 3.6 FlashFeed it a huge document and get a tight, polished answer back, at a Flash-tier token rate | Chat & code | $0.79 / $3.95 · 1M tok | |
| GPT-5.6 LunaOpenAI's quickest GPT-5.6 tier for chat, sorting and field extraction at very high volume | OpenAI | Chat & code | $0.216 / $1.30 · 1M tok |
| GPT-5.6 SolThe top GPT-5.6 model, at its best when an agent works in a shell and fixes code over several steps | OpenAI | Chat & code | $4.32 / $21.60 · 1M tok |
| GPT-5.6 TerraThe middle GPT-5.6 tier, around GPT-5.5 quality at a lower price, for everyday code and reasoning | OpenAI | Chat & code | $2.16 / $12.96 · 1M tok |
| Kimi K3Open weights for hours-long engineering in large repositories, with built-in understanding of images and video | Moonshot AI | Chat & code | $3.38 / $16.88 · 1M tok |
June 2026
8 models| Model | Vendor | Type | Price |
|---|---|---|---|
| Grok Imagine Video 1.5Animate one photo into a clip of up to 15 seconds with a synced soundtrack and spoken lines you write | xAI | Video | from $0.088 / s |
| HappyHorse 1.1Multi-character clips where you tell the model which reference image is who, in nine aspect ratios | Alibaba | Video | from $0.399 / s |
| Kling 3.0 TurboQuick Kling clips without sound, for edits where you add your own music and voice later | Kuaishou | Video | from $0.277 / s |
| Seedance 2.0 MiniThe entry-level Seedance 2.0 for short silent clips in volume, with sound only when you ask for it. | ByteDance | Video | from $0.0784 / s |
| Seedance 2.0 Mini EcoThe slowest and most budget-minded Seedance 2.0 Mini lane, with 1080p for jobs that can wait in a queue. | ByteDance | Video | from $0.0789 / s |
| Nano Banana 2 LiteThe quickest member of Google's Nano Banana family, for lots of 1K pictures and simple edits. | Image | $0.0375 / image | |
| Claude Sonnet 5A mid-priced Claude that plans, drives terminals and browsers, and codes close to Opus 4.8 results | Anthropic | Chat & code | $3.24 / $16.20 · 1M tok |
| GLM-5.2Z.ai's open-weight model that plans first, then builds and checks multi-file work over a 1M-token context | Z.ai | Chat & code | $1.22 / $4.05 · 1M tok |
May 2026
3 models| Model | Vendor | Type | Price |
|---|---|---|---|
| Gemini OmniMerge three photos into one scene, animate a still or re-imagine a clip, in 4 to 10 second videos | Video | from $1.45 / video | |
| Claude Opus 4.8Keep a pipeline that was tuned on Opus 4.8 running exactly as it is while you test newer models next to it | Anthropic | Chat & code | $3.78 / $18.90 · 1M tok |
| MiniMax M3An inexpensive open-weight model for coding agents that can write very long answers in one go. | MiniMax | Chat & code | $0.81 / $3.24 · 1M tok |
April 2026
6 models| Model | Vendor | Type | Price |
|---|---|---|---|
| HappyHorse 1.0Short clips from text, one image or up to nine references, plus a mode that restyles your own footage | Alibaba | Video | from $0.322 / s |
| Wan 2.7 VideoOne Alibaba model that creates clips, continues them, moves a still to a soundtrack and edits footage. | Alibaba | Video | from $0.161 / s |
| GPT Image 2OpenAI images in almost any shape, from square to ultra-wide 3:1 banners, up to 4K | OpenAI | Image | from $0.0625 / image |
| Wan 2.7 ImageAlibaba's image model for consistent product shots and edits, with a seed and up to four takes per run. | Alibaba | Image | $0.0611 / image |
| Wan 2.7 Image ProThe higher Wan 2.7 Image tier, for 4K pictures from a prompt and 2K edits with up to nine references. | Alibaba | Image | $0.154 / image |
| DeepSeek V4 FlashLow-cost DeepSeek for agents and bulk jobs that read huge inputs and still need to stay on budget | DeepSeek | Chat & code | $0.135 / $0.27 · 1M tok |
March 2026
4 models| Model | Vendor | Type | Price |
|---|---|---|---|
| Kling 3.0 Motion ControlYour mascot or character performs the exact moves from a real video, with a consistent look from start to end | Kuaishou | Video | $0.249 / s |
| PixVerse V6Flexible clips from cheap 360p drafts to 1080p, with transitions and fusion of up to 7 images. | PixVerse | Video | from $0.0405 / s |
| Veo 3.1 LiteThe entry ticket to Veo footage: write a prompt, get an 8-second clip, no images involved. | Video | from $0.177 / video | |
| Mureka V9Write the words, describe the sound, and get back a full sung song as one audio file. | Mureka | Audio | $0.15 / song |
February 2026
8 models| Model | Vendor | Type | Price |
|---|---|---|---|
| Kling 3.0Talking characters, first-to-last-frame transitions and 4K ad shots from one Kling model with optional sound | Kuaishou | Video | from $0.0963 / s |
| Grok Imagine 1.0 VideoLonger AI clips, up to half a minute, from a prompt or from a set of up to seven reference images | xAI | Video | from $0.0299 / s |
| Seedance 2.0ByteDance's all-round video model for clips up to 4K, with voices, effects and music from the same render. | ByteDance | Video | from $0.157 / s |
| Seedance 2.0 EcoThe economy tier of Seedance 2.0 for clips you do not need right away, such as overnight batches. | ByteDance | Video | from $0.248 / s |
| Seedance 2.0 FastThe quicker, cheaper way to run Seedance 2.0 when 720p is enough, sound included. | ByteDance | Video | from $0.124 / s |
| Nano Banana 2Google's Gemini 3.1 Flash Image for quick variant rounds, extreme banner shapes and sizes up to 4K. | Image | from $0.07 / image | |
| Qwen Image 2Alibaba's model for pictures where the lettering has to be right: posters, slides, comics and infographics. | Alibaba | Image | $0.077 / image |
| Seedream 5.0 LiteByteDance's model for high-resolution pictures and series of up to four images that share one style. | ByteDance | Image | from $0.077 / image |
January 2026
3 models| Model | Vendor | Type | Price |
|---|---|---|---|
| Vidu Q3 ProShengShu's sharper Vidu Q3 for clips from one to sixteen seconds, with dialogue and effects at no extra cost. | ShengShu | Video | from $0.142 / s |
| Vidu Q3 TurboThe quick Vidu Q3 for lots of short clips, with dialogue and sound effects already in the price. | ShengShu | Video | from $0.081 / s |
| Grok ImagineUp to four images per prompt in 13 shapes, including tall phone formats, at a flat rate per quality tier | xAI | Image | $0.0246 / image |
December 2025
2 models| Model | Vendor | Type | Price |
|---|---|---|---|
| Kling 2.6 Motion ControlMake a character from any photo copy the dance, gesture or walk from a short clip, at a budget rate | Kuaishou | Video | $0.138 / s |
| FLUX.2 [max]The FLUX.2 tier for final campaign images where fabric, skin and surface detail have to hold up at large size | Black Forest Labs | Image | from $0.154 / image |
November 2025
4 models| Model | Vendor | Type | Price |
|---|---|---|---|
| FLUX.2 [pro]Photoreal product and lifestyle shots with readable in-image text, built from up to eight reference images | Black Forest Labs | Image | from $0.0688 / image |
| Nano Banana ProGoogle's Gemini 3 Pro Image for putting real people, products and styles into one polished 4K scene. | Image | from $0.07 / image | |
| FLUX.2 [flex]Stylized posters and illustrations from text or up to eight reference images, at 1K or 2K | Black Forest Labs | Image | from $0.193 / image |
| Z-Image TurboAlibaba Tongyi's quick text-to-image model for bulk pictures, with short English or Chinese text in the frame. | Alibaba | Image | $0.02 / image |
October 2025
3 models| Model | Vendor | Type | Price |
|---|---|---|---|
| Veo 3.1 (Quality)The top Veo 3.1 tier for the render you actually deliver: ads, trailers and music video shots. | Video | from $2.53 / video | |
| Veo 3.1 FastThe quicker Veo 3.1 for 8-second shots in vertical or horizontal format, from text or up to three images. | Video | from $0.354 / video | |
| Claude Haiku 4.5Quick replies and cheap parallel sub-agents from the smallest Claude, billed per token from one balance | Anthropic | Chat & code | $1.35 / $6.75 · 1M tok |
August 2025
1 model| Model | Vendor | Type | Price |
|---|---|---|---|
| Nano BananaThe first Nano Banana, still handy for simple 1K edits and for pipelines already tuned to its look. | Image | $0.0359 / image |
May 2025
3 models| Model | Vendor | Type | Price |
|---|---|---|---|
| FLUX.1 Kontext [max]Edit an image by describing the change, with lettering and multi-part instructions that come out right | Black Forest Labs | Image | $0.176 / image |
| FLUX.1 Kontext [pro]Change clothes, backgrounds or text in a photo just by describing it, while the person stays recognizable | Black Forest Labs | Image | $0.088 / image |
| Imagen 4.0Google's text-only image model for headers, stock-style photos and pipelines where a prompt is the only input | Image | $0.115 / image |
A new model shows up here once its vendor makes it available to developers. Check what a run costs.