Ramp opens its LLM router after cutting AI costs by 30%+
iamrobotbear · x · 2026-07-21
Ramp says its internal LLM router is now being opened up as Ramp Router: one OpenAI-compatible endpoint that sends each request to the cheapest model that still clears the quality bar.
- The system tests new models on real work before routing traffic.
- It also decides when to cache, batch, or escalate to a stronger model.
- Ramp says the approach has already cut costs by 30%+ while keeping output the same or better.
- The alpha group is reportedly running at 2.75T tokens/month with 99.99% uptime.
The post frames model routing as the practical answer to fast-moving prices and capabilities across GPT, Claude, Gemini, Grok, Qwen, DeepSeek, Kimi, and GLM.
Related event: Ramp Opens Internal Multi-Model Router, Claims 30% LLM Cost Reduction(11 posts)→
More from Infra
- Open-source Latch compiles PyTorch into training megakernels, 2.75x faster on H100 — sguada · 2026-10-27
- NVIDIA Claims HGX Rubin NVL8 Delivers 10x More Tokens per Megawatt Than B200 — thursdai_pod · 2026-10-27
- Commoditization dilemma: either human labor or GPUs become the commodity — _AustinCalvert_ · 2026-10-27
- DeepSpeed shows AutoEP/AutoTP/AutoSP parallelism gains, hosts office hours — PyTorch · 2026-10-27
- Running a local 9B as a coding agent: brief the big model to plan, let the small model execute — Electrical_Ninja3805 · 2026-10-27
- Cloudflare Durable Objects validated as top architecture for AI agent WebSocket servers — tristanbob · 2026-10-27