Available Models

Agnes 1.5 Flash is a lightweight conversational model by Agnes AI, optimized for fast response times and everyday tasks. Completely free to use, supports 128K context window and Function Calling. An ideal choice for entry-level AI conversations.

Agnes 2.0 Flash is an upgraded version by Agnes AI, balancing conversation quality with fast response times. Completely free to use, supports 128K context window and Function Calling. Significant quality improvement over the 1.5 version.

Agnes-Image-2.0-Flash is a high-performance image generation and editing model by Sapiens AI. Supports text-to-image, image-to-image, and multi-image composition workflows. Ideal for creative production, image optimization, marketing visuals, e-commerce product images, social content, and professional visual content.

Agnes Image 2.1 Flash is the latest image generation model by Agnes AI, supporting text-to-image, image-to-image, and multi-image composition workflows. Significant improvements over 2.0 in image quality, detail rendering, and generation speed. Ideal for creative design, marketing visuals, e-commerce product images, and social media content.

Agnes Video V2.0 is a video generation model by Agnes AI, supporting both text-to-video and image-to-video modes. Generates high-quality video content from text descriptions or reference images. Ideal for short video creation, advertising, product demos, and creative content production.

DeepSeek V4 Flash model available via SenseTime's SenseNova channel. Features MoE architecture, 1M ultra-long context, and excellent reasoning and code generation capabilities. SenseNova provides a stable domestic access channel.

SenseNova 6.7 Flash Lite is a lightweight LLM by SenseTime, part of the SenseNova foundation model family. It delivers fast inference at minimal cost while maintaining solid conversation quality, ideal for daily chat, information queries, and lightweight text processing.

GPT-5.4 is a powerful language model by OpenAI with strong reasoning and text generation capabilities. Supports 128K context window, excelling in complex reasoning, creative writing, code generation, and multilingual understanding tasks.

GPT-5.4 mini is a lightweight, cost-efficient model by OpenAI that maintains solid performance at significantly lower cost. Supports 128K context window with fast response times, ideal for daily chat, text processing, and information extraction tasks. The most cost-effective option in the GPT-5.4 family.

GPT-5.5 is OpenAI's flagship language model, representing the most advanced AI capabilities available. Achieves top-tier performance across complex reasoning, creative generation, coding, multilingual understanding, and long-form text processing. Designed for professional and enterprise applications demanding the highest model capabilities.

TTS v1 is a text-to-speech model by OpenAI that converts text into natural-sounding speech. Supports multiple languages and voice styles with low latency, ideal for real-time voice synthesis, audio content production, and conversational systems.

TTS v1 HD is a high-definition text-to-speech model by OpenAI, offering superior audio quality and more natural voice output compared to the standard version. Supports multiple languages and voice styles, ideal for audiobooks, podcasts, customer service voice, and professional dubbing where high audio quality is essential.

Whisper v1 is a speech recognition model by OpenAI supporting multilingual speech-to-text transcription. Delivers excellent accuracy with automatic language detection across 99 languages. Ideal for meeting transcription, subtitle generation, voice search, and podcast transcription.

MiMo v2.5 is a lightweight LLM by Xiaomi, optimized for cost-effective dialogue and code tasks. Supports 1M ultra-long context with extremely low inference cost. Features decent code generation and reasoning capabilities with transparent thinking process. Ideal for everyday use and development debugging.

MiMo V2.5 ASR is a speech recognition model by Xiaomi, optimized based on the MiMo V2.5 architecture. Supports multilingual speech-to-text with high accuracy and low latency. Ideal for voice assistants, meeting transcription, subtitle generation, and voice search. Supports 1M ultra-long audio input.

MiMo v2.5 PRO is Xiaomi's high-performance LLM with significant improvements in reasoning and Function Calling. Supports 1M ultra-long context with excellent code generation and complex reasoning capabilities. Ideal for professional scenarios requiring tool use, multi-step reasoning, and high-quality code generation.

DeepSeek R1 is a reasoning-focused model trained with reinforcement learning, excelling in math, coding, and complex logic tasks. Performs deep thinking before answering with transparent reasoning process. Ideal for academic research, math competitions, and code review. Available via Volcengine for stable domestic access.