{"updated":"2026-11-07T18:59:28Z","models":[{"id":"glm-5.3-flash-uncensored","name":"glm-5.3-flash-uncensored","description":"GLM-5.3 Flash community fine-tune with safety refusals removed. 32K context. Intended for research, red-teaming and creative work where refusals get in the way. No content filtering is applied; use is subject to the Terms of Service.","modalities":{"input":["text"],"output":["text"]},"context_length":32768,"max_output_length":32768,"pricing":{"unit":"per_million_tokens","prompt_usd":0.3,"completion_usd":1.0},"billing_note":null,"ready":false,"uptime":0.78962},{"id":"bge-reranker-v2-m3","name":"bge-reranker-v2-m3","description":"BGE-reranker-v2-m3 cross-encoder reranking, served on the Thailand B300 fleet. Billed per input token like embeddings.","modalities":{"input":["text"],"output":["score"]},"context_length":8192,"max_output_length":null,"pricing":{"unit":"per_million_tokens","prompt_usd":0.02,"completion_usd":0.0},"billing_note":null,"ready":false,"uptime":0.7476},{"id":"flux-schnell","name":"flux-schnell","description":"FLUX.1-schnell text-to-image (Apache 2.0). Billed at a 1000-token equivalent per image, i.e. USD 0.01 per image at the configured rate.","modalities":{"input":["text"],"output":["image"]},"context_length":512,"max_output_length":null,"pricing":{"unit":"per_million_tokens","prompt_usd":0.0,"completion_usd":10.0},"billing_note":"1000-token equivalent per image","ready":false,"uptime":0.74748},{"id":"chroma1-hd","name":"chroma1-hd","description":"Chroma1-HD text-to-image (Apache 2.0), de-distilled from FLUX.1-schnell. Unlike schnell it takes real classifier-free guidance, so negative_prompt and guidance_scale actually steer it; costs ~35 steps against schnell's 4. Billed at a 1000-token equivalent per image.","modalities":{"input":["text"],"output":["image"]},"context_length":512,"max_output_length":null,"pricing":{"unit":"per_million_tokens","prompt_usd":0.0,"completion_usd":10.0},"billing_note":"1000-token equivalent per image","ready":false,"uptime":0.74746},{"id":"indextts2","name":"indextts2","description":"IndexTTS-2 text-to-speech via /v1/audio/speech. Speaker and emotion are separate inputs, so one cloned character voice can be gentle or excited on demand - pass the app's mood through `instruct` and keep `emo_alpha` near 0.6, since full strength distorts articulation on short lines. Licensed under bilibili's agreement, free below 100M monthly actives. Billed at a 1000-token equivalent per 20 seconds of audio.","modalities":{"input":["text"],"output":["audio"]},"context_length":2000,"max_output_length":null,"pricing":{"unit":"per_million_tokens","prompt_usd":0.0,"completion_usd":10.0},"billing_note":"1000-token equivalent per 20 seconds of audio","ready":false,"uptime":0.7473},{"id":"whisper-large-v3","name":"whisper-large-v3","description":"Whisper large-v3 speech-to-text via /v1/audio/transcriptions (multipart upload). Mandarin output is converted to Taiwanese traditional orthography. Billed at a 1000-token equivalent per 20 seconds of audio.","modalities":{"input":["audio"],"output":["text"]},"context_length":null,"max_output_length":32768,"pricing":{"unit":"per_million_tokens","prompt_usd":0.0,"completion_usd":3.33},"billing_note":"1000-token equivalent per 20 seconds of audio","ready":false,"uptime":0.74767},{"id":"bge-m3","name":"bge-m3","description":"BGE-M3 multilingual embedding model (1024-dim) served on the Thailand B300 fleet.","modalities":{"input":["text"],"output":["embedding"]},"context_length":8192,"max_output_length":null,"pricing":{"unit":"per_million_tokens","prompt_usd":0.02,"completion_usd":0.0},"billing_note":null,"ready":false,"uptime":0.74748},{"id":"glm-5.2-fp8","name":"glm-5.2-fp8","description":"","modalities":{"input":["text"],"output":["text"]},"context_length":131072,"max_output_length":32768,"pricing":{"unit":"per_million_tokens","prompt_usd":0.93,"completion_usd":3.0},"billing_note":null,"ready":false,"uptime":0.78487},{"id":"qwen3-vl-30b","name":"qwen3-vl-30b","description":"Qwen3-VL-30B-A3B-Instruct (Apache 2.0, FP8) vision-language model. Same /v1/chat/completions as the text model, with image_url content parts; up to 4 images per prompt. Capacity for the app rather than inventory for sale.","modalities":{"input":["text","image"],"output":["text"]},"context_length":32768,"max_output_length":4096,"pricing":{"unit":"per_million_tokens","prompt_usd":0.3,"completion_usd":1.2},"billing_note":null,"ready":false,"uptime":0.74702},{"id":"wan2.2-ti2v-5b-720p","name":"wan2.2-ti2v-5b-720p","description":"Wan 2.2 TI2V-5B text/image-to-video, 720p (1280x704) 24 fps, 1-10 s, async job via /v1/videos","modalities":{"input":["text","image"],"output":["video"]},"context_length":null,"max_output_length":null,"pricing":{"unit":"per_second","prompt_usd":0.0,"completion_usd":10.0,"per_second_usd":"0.05"},"billing_note":"Billed per second of generated video: 5000 token-equivalents per second.","ready":false,"uptime":0.61491}]}