Gridline launches one-API fine-tuning, distillation and batch inference for open models

Scobleizer · x · 2026-11-19

Startup Gridline launches a platform that lets teams fine-tune open models (Llama, Qwen, Mistral, gpt-oss) on their own data, distill a large teacher model into a small owned student model, and run batch inference — all through a single OpenAI-compatible API. Jobs are quoted upfront with before/after eval loss, weights are downloadable, and batch results return within 24 hours at base-model pricing with no endpoints or idle GPUs. Integration is a base URL and key swap if you already use the OpenAI client.

Original post →

More from Infra

Infra channel →