Community-submitted OpenAI-compatible keys. Unlocking a key costs points; opening the same key again is free.
Use the shared Base URL and API key together in an OpenAI-compatible client.
Log in and spend 20 points to reveal the full API key. Opening the same key again does not cost more points.
Copy the Base URL and the full API key. Most clients need both values to route requests correctly.
Start with a lightweight model request and expect community keys to have rate limits or temporary downtime.
Answers about login, verification, model lists, and safe use of shared keys.
The list is public, but full keys require an account and points so claim activity can be tracked and obvious abuse is easier to limit.
Yes. LMSpeed calls the OpenAI-compatible models endpoint, selects a likely chat model, and sends a one-token chat completion. A key is marked available only when both requests succeed.
The model list comes from the submitted Base URL and key at submission time. Providers may later add, remove, or restrict models.
No. Treat shared free keys as testing resources only. They may have strict rate limits, quota caps, or become unavailable without notice.
Explore the latest published LLMs tracked by LMSpeed, then open a model to compare providers, pricing, benchmarks, and free availability.
DeepSeek V4 Flash 0731 is a sparse mixture-of-experts model from DeepSeek, with 13B active parameters out of 284B total. This re-post-trained revision is suited for coding, reasoning, and agent workflows.
Inkling Small is an open-weight multimodal mixture-of-experts model from Thinking Machines Lab, with 12B active parameters out of 276B total. It is positioned as the smaller, more efficient member of...
Qwen3.7 Flash is a vision-language reasoning model from Alibaba. It is suited for multimodal agents, visual coding, search, and computer interaction, with strengths in object recognition, spatial understanding, and real-world...
Voyage 4 Lite is an efficiency-focused general-purpose embedding model from Voyage AI by MongoDB, optimized for low-latency and cost-sensitive retrieval. It supports Matryoshka embeddings at 2048, 1024, 512, and 256...
Voyage 4 is a general-purpose multilingual embedding model from Voyage AI by MongoDB. It is suited for retrieval, semantic search, and RAG applications, with Matryoshka embeddings at 2048, 1024, 512,...
Voyage 4 Large is a general-purpose multilingual embedding model from Voyage AI by MongoDB, optimized for retrieval quality. It supports Matryoshka embeddings at 2048, 1024, 512, and 256 dimensions, with...
