Liste des modèles d’IA : API vidéo, image, musique et chat

Google describes Gemini 4 Hydrogen as a fast, efficient model for everyday coding. It is adapted from Gemini 4 Carbon and succeeds Gemini 3.8 Flash. On Kie.ai it currently supports chat requests only.

Google describes Gemini 4 Carbon as its strongest reasoning model, built for complex software engineering, multimodal knowledge work and scientific reasoning. It is based on Gemini 4 Argon and spends more thinking tokens on hard problems to explore hypotheses and verify its work. On Kie.ai it currently supports chat requests only.

MiniMax H3 Image is an open-source image generation and editing model from MiniMax, part of the MiniMax H3 family. On Kie.ai it generates images at 1K, 2K or 4K (about 16 megapixels) in ten aspect ratios. It also edits images using up to 16 reference images.

The Kie Image Upscaler API is a KIE hosted model that upscales an image by 2×, 4× or 8× up to 67 megapixels, or enhances it at its original size, while keeping text in the image faithful.

Seedream 5.5 is ByteDance Seed's image creation model, released on November 9, 2026. It keeps unedited regions intact across up to 100 rounds of editing, generates images with transparent backgrounds natively, and renders more realistic textures and dense small text.

Claude Fable 5.2 is Anthropic's advanced AI model for demanding technical work, combining stronger agentic coding, multidisciplinary reasoning, and improved vision. It generates outputs up to 30% faster than Fable 5.1 and supports complex, multi-stage tasks.

The Video Upscaler API on Kie.ai is a KIE hosted model that upscales video to a 1080P, 2K or 4K output tier, keeps the original aspect ratio, frame rate and audio, and returns exactly the resolution you pick.

Gemini Omni 1.2 Flash is Google's updated Gemini Omni Flash model for generating and editing video with native, synchronized audio from text, image and video inputs. Google says it extends clips to 15 seconds, adds multi-turn conversational editing and follows references more closely than Gemini Omni 1.1 Flash.

Kling 4.0 Flash is the Flash model in the Kling 4.0 family, which Kling positions for faster generation and greater value. On Kie.ai, the Kling 4.0 Flash API offers two models: Reference to Video, which accepts a prompt with optional reference images, elements, or reference videos, and Keyframe to Video, which generates a video from a single first-frame image. Both output 720p MP4 clips from 3 to 20 seconds.

Built by NVIDIA for speech recognition, Nemotron 3.5 Transcribe can return the transcript, word-level timestamps, speaker labels, and speech translation between English and 31 languages in one API request on Kie.ai.

The Kimi K3.1 Pro API gives you what Moonshot AI calls its most capable 3T-class model to date, post-trained from Kimi K3 for coding, 3D creation, financial research and factual accuracy.

Access OpenAI's GPT Image 2.6 for image generation and editing in two versions: Flare, the fast version, which OpenAI says improves layout quality, and Sunburst, for refined, higher-quality output and consistent edits.

GPT-6.1 Terra is OpenAI's fast, efficient GPT-6.1 model for visual reasoning, precise computer use and clear writing, and a step up from GPT-5.6 Terra.

PixVerse V6 Max is PixVerse's AI video model for cinematic text-to-video and image-to-video generation. It delivers realistic motion and native audio, with up to 15s video at 1080p.

BytePlus OmniHuman 2.0 is an audio-driven AI video generation model that transforms images and audio into lifelike talking avatar videos. It supports natural lip sync, expressive facial movements, body gestures, and video generation up to 5 minutes long.

Kling 4.0 is Kuaishou’s AI video model for professional video creation. It delivers realistic motion and synchronized audio, with up to 30s video in 720p and 1080p; 4K output is coming soon.

Flux 3 is Black Forest Labs' new generation of image models. Kie AI starts with Flux 3 Klein for fast text-to-image generation and multi-reference image editing up to 4K, with more Flux 3 models to follow.

Gemini 4 Argon is Google DeepMind's flagship AI model for advanced reasoning, agentic coding, and complex enterprise tasks. It supports multimodal inputs and long-horizon workflows, with a unified 1M-token context budget.

Vidu Q4 is an AI video generation model designed for cinematic visuals, realistic motion, and synchronized audio. Vidu Q4 Preview supports up to 4K resolution, 16-second videos, 15 image references, and three audio references for consistent characters and voices.

Claude Haiku 5.5 is Anthropic's small model for high-volume, latency-sensitive work. On Kie.ai you call it as claude-haiku-5-5 on POST /claude/v1/messages in the Anthropic Messages format, with tools, streaming and a thinking switch.