TextSynth

TextSynth gives developers API access to open models for text completion, chat, translation, embeddings, image generation, speech synthesis and transcription. Its catalogue includes Mistral, Llama, Gemma, Stable Diffusion and Whisper, served from its own inference code on servers in France.

A browser playground lets you test models before writing code, and the REST API is documented for integration. The company says its custom inference code runs faster on standard GPUs and CPUs, which keeps per-request prices low.

The free plan allows rate-limited use of all language models, with at most 200 generated tokens per request and speech-to-text capped at three minutes, and requests are logged. The standard plan removes the rate limits and charges per million tokens, per image or per audio minute.