Lambda
On-demand GPU instances billed by the minute with no egress fees, from a company that also builds the hardware.
Every product in the index, with what each plan costs and when it was last verified.
On-demand GPU instances billed by the minute with no egress fees, from a company that also builds the hardware.
Serverless GPUs for your own inference and training jobs, billed per second with a free starter allowance.
Run open models locally on your own machine.
One API key, hundreds of models, automatic failover.
Run thousands of community models by the second — image, video, audio, text.
Per-second GPU rental across Community and Secure Cloud tiers, plus serverless inference workers.
Fast serverless inference and fine-tuning for open-weight models, billed per token.
Enterprise-leaning models built for RAG and search, with open weights released for research use.
Free chat with open-weight reasoning models, and the cheapest serious API anywhere — though DeepSeek announced a substantial price rise in August 2026.
China's most-used consumer assistant, with aggressive metered pricing via Volcano Engine.
Baidu's assistant, tied into its search and maps ecosystem.
Where open weights live.
Hybrid architecture built for very long documents and grounded answers.
Million-token context and strong agentic browsing, with tiers named after tempo markings.
Open weights you host yourself.
Text, voice and video models under one roof, with open-weight text releases and a pay-as-you-go developer platform.
The most widely forked open-weight family.
Hunyuan models wrapped in a WeChat-native assistant.
GLM sold as a coding plan that heavily undercuts Western agents.
Searches peer-reviewed studies and shows where the evidence leans.
Systematic literature review across 138M papers — extracts findings into comparison tables with PRISMA support.
General agent that plans, browses and returns finished deliverables.
Upload your sources, get grounded answers and audio overviews.
Answer engine with citations on everything, and a model picker so you choose who answers each query.