Playground TextSynth

TextSynth is a hosted service that gives developers and curious users access to open and third party AI models through a REST API and a browser playground. The catalog covers large language models such as Mistral 7B, Llama 3 and 3.1 8B, Llama 3.3 70B and Gemma 3 27B, along with MADLAD400 for translation, a BGE embeddings model, Stable Diffusion for image generation, Whisper Large V3 for speech to text and Parler TTS for speech synthesis. Typical uses listed by the vendor include text completion, question answering, classification, chat, translation, image and speech generation, and transcription. The company says it was founded in 2020, was among the first sites to offer GPT 2, runs custom inference code intended to cut cost on standard GPUs and CPUs, and keeps all of its servers in France. The free plan allows access to all language models with rate limits, periodic captcha completion, at most 200 generated tokens per request and a three minute limit on transcription, and request content is logged. The Standard plan removes those limits, keeps no log of request content and is credit based, so spending is capped by the balance. Rates are per million tokens: Mistral 7B and Llama 3 8B cost $0.20 for input and $2.00 for output, Gemma 3 27B costs $0.40 and $4.00, and Llama 3.3 70B costs $0.70 and $7.00. Stable Diffusion is $0.005 per image, Whisper is $0.003 per minute and Parler TTS is $14.00 per million characters. Prices can be shown in credits, dollars or euros. Custom support is available by email.