River Inference Public production inference for open-weight models, starting with DeepSeek V4.1 Flash and GLM 5.3 Flash at 20% below official provider rates. The service offers an OpenAI-compatible API, 1M-token context, vision, and streaming, with a custom training and inference stack written in Rust. New signups get $25 in free credits.
Visit productFeatured media
Loading media…
View on XCosigns Received
“Try River not just for inference, but for training GLM/Kimi/DeepSeek, and running it there. Unlike Runpod serverless, your finetuned model can typically almost instantly respond to your prompts. And it's reasonable pricing. There I run Kimi and GLM trained on Sonnet 4.5 outputs”