Inference
Baseten
Production model serving that hit $600M ARR in 2026.
Cerebras Inference
1,800+ tokens/sec inference on wafer-scale silicon.
DeepInfra
Open-model inference from $0.02 per million tokens.
Featherless AI
Flat monthly rate for unlimited use of 20,000+ open LLMs
Fireworks AI
Production inference platform for open-weights LLMs
OSChina's serverless hub for open model inference, tuning and apps
GPU marketplace and pay-per-token inference for 25+ open AI models
Lambda Inference
Pay-per-token inference API that Lambda itself is winding down
Acquired by NVIDIA in 2025, now DGX Cloud Lepton.
200+ models and per-second GPU cloud, priced low.
One API key, one bill, around 400 LLMs
Parasail
Brokered GPU inference: pay per token, skip the long-term contract
China's largest independent AI cloud, now filing for a Hong Kong IPO
Replicate
Run any open AI model via API — no infrastructure
SambaNova Cloud
Open models at record tokens-per-second on RDU silicon
SiliconFlow 硅基流动 (SiliconCloud)
OpenAI-compatible API access to 200+ open models, billed per token
Together AI
High-performance inference for open-weights LLMs
Volcengine Ark 火山方舟 (ByteDance)
ByteDance's MaaS platform for Doubao, DeepSeek, GLM and Kimi
VMware's AMD AI Factory Still Requires VMware Cloud Foundation
Supporting AMD's open software stack sounds like a route away from concentration. It requires VMware Cloud Fou...
KE
1 Sep
OpenAI's Jalapeño Beat Blackwell on Work Per Watt, Which Is the Expected Result
A chip that only serves inference outperformed one that also trains, measured on a public third-party benchmar...
EV
27 Aug
Nvidia Is Talking To The Korean Startup Whose Chips Compete Where Nvidia Is Weakest
Jensen Huang met Rebellions' chief executive in Santa Clara. The company designs inference NPUs, is valued nea...
KE
22 Aug
Baseten is in talks to raise US$1 billion at an US$11 billion valuation as inference money keeps flowing
Baseten, which rents Nvidia servers to companies running AI models, is in talks to raise US$1 billion at an US...
AI
1 Jun
Lablup open-sources MLXcel, an Apple-Silicon inference engine, under Apache 2.0
Lablup has released MLXcel, an open-source engine for running AI models on Apple Silicon, under the permissive...
KE
1 Jun