Video model · Oct 15, 2026
Kling 4.0 is an announced video model with stronger motion, multimodal references, and integrated audio handling. Its description lists 3 to 30 seconds, up to ten keyframes, 4K, and generated stereo audio with lip synchronization.
Video model · Sep 28, 2026
Kling 4.0 Flash is the faster variant of Kling 4.0, generating up to 20 seconds of 720p video in 8-bit SDR. It supports text, image, and Omni references, but not first/last frames or multiple keyframes. Access for Ultra Yearly subscribers began on September 28, 2026.
Language model · Jul 10, 2026
KAT-Coder-Air V2.5 is the fast API variant in a family trained around executable environments and verifiable agent tasks. The technical report uses AutoBuilder and KwaiClawEnv to represent that focus on real repository steps. Air offers a 256K context, up to 80K output tokens, and tool calling; public StreamLake documentation gives neither a parameter count nor a weight release.
Language model · Jul 10, 2026
KAT-Coder-Pro V2.5 is trained for executable repository tasks rather than isolated code snippets. Its technical report names fail-to-pass and pass-to-pass verification, tool-use trajectories, and multi-teacher post-training. Publicly, Kuaishou and KwaiPilot offer it as a proprietary endpoint rather than a downloadable checkpoint.
Video model · Jun 17, 2026
Kling 3.0 Turbo is a video model from Kuaishou. Kling 3.0 Turbo supports text-to-video and image-to-video.
Video model · Feb 5, 2026
Kling 3.0 Omni is a video model from Kuaishou. Kling 3.0 Omni accepts text, image, and video inputs. Kling’s native model map also describes audio output and storyboards.
Video model · Feb 5, 2026
Kling 3.0 Standard API is a video model from Kuaishou. This row represents the separately selectable standard API route; it does not claim a separately released base model or weight checkpoint.
Image model · Feb 5, 2026
Kling Image 3.0 is an image model from Kuaishou. Kuaishou announced Kling Image 3.0 on February 5, 2026.
Image model · Feb 5, 2026
Kling Image 3.0 Omni is an image model from Kuaishou. Its hosted TokenHub identifier is kling-image-v3-omni; the table lists it separately for T2I, image references, and image series.
Video model · February 2026
New in Kling 3.0 Pro are native audio, multi-shot storyboards, reference video, and in-video editing alongside text-to-video and image-to-video. Kuaishou describes consistent characters from image or video references, with up to 4K and 15 seconds.
Video model · January 2026
Kling 2.6 Pro was Kuaishou’s high-resolution Kling generation for API and web access. The proprietary model generated up to 1080p and included native audio in the clip.
Video model · Dec 15, 2025
Kling Video O1 is a video model from Kuaishou. Kling Video O1 is a multimodal model for text-to-video, image and element references, first and last frames, reference video, video extension, and editing.
2 more current models are in the model list above.