Open-source · Coding · 3 of 10 tools · updated Oct 17, 2026
Open-source AI tools for LLM APIs & model routers
LiteLLM leads the open-source options for LLM APIs & model routers at 92/100, with 61,411 GitHub stars. 3 of the 10 tools tracked for this job publish their source, so you can read it, fork it or self-host it. The highest-scoring tool overall is Hugging Face (94), which does not qualify: Every open model ends up here, with Inference Providers routing to multiple backends.
Same evidence score as the ranking — GitHub stars, Hacker News mentions, release activity, free access — recomputed nightly, never sold. See all 10 tools for LLM APIs & model routers.
Top 3
evidence / 100
+ Self-hosted gateway with spend tracking, virtual keys, and fallbacks across providers.
− Huge dependency surface and frequent releases that occasionally break provider mappings.
+ Serves many concurrent requests on GPUs far faster than llama.cpp-based tools.
− Linux and NVIDIA-centric; painful on Mac or consumer hardware.
+ Unified provider interface and React hooks make streaming chat UIs trivial.
− Major-version upgrades have required significant rewrites.
All 3, side by side
| Tool | Score | Stars | Source | Days idle | Platforms |
|---|---|---|---|---|---|
| LiteLLM | 92 | 61,411 | BerriAI/litellm | 0 | cli, api, linux |
| vLLM | 80 | 93,962 | vllm-project/vllm | 0 | linux, cli, api |
| Vercel AI SDK | 72 | 27,359 | vercel/ai | 0 | api, cli |
Open source here means the tool's own source is public, not that a vendor published an SDK. Self-hosting cost is yours.
Head to head: LiteLLM vs vLLM