Open-source Together AI alternatives

vLLM is an open-source library for running and serving large language models efficiently. It originated at UC Berkeley's Sky Computing Lab and is now developed by a large community of contributors from academia and industry. It is written in Python and released under the Apache-2.0 license.

Speed comes from techniques such as PagedAttention for attention key-value memory, batching incoming requests continuously, chunked prefill, prefix caching, CUDA and HIP graphs, and optimized attention and mixture-of-experts kernels. It supports many quantization formats, including FP8, INT8, INT4, GPTQ and AWQ, as well as speculative decoding and disaggregated prefill and decode.

On the usability side, vLLM integrates with Hugging Face models and supports parallel sampling, beam search, streaming outputs, structured outputs and tool calling, plus several parallelism modes (tensor, pipeline, data, expert and context) for distributed inference. It targets NVIDIA, AMD and TPU hardware and suits teams that serve open models in production and need high throughput.

Unsloth is an open-source framework with a desktop app for running and training AI models on your own hardware. The same tool covers inference of local language, image, audio and embedding models as well as fine-tuning them, and it runs on Windows, Linux (including WSL) and macOS, with support for NVIDIA, AMD and Intel GPUs, CPUs and a Vulkan backend.

On the running side it loads GGUF and MLX models, serves them through an OpenAI-compatible API, and lets coding agents such as Claude Code and Codex use local models with tool calling and code execution. It also offers private web search, deep research and retrieval, plus access from other devices over a LAN or a secure remote link. For training it supports LoRA, QLoRA, full fine-tuning, pretraining and reinforcement learning methods such as GRPO and DPO, with export to formats including GGUF.

The project claims fine-tuning that is roughly twice as fast with about 70 percent less VRAM, which was its original draw. Unsloth is Apache-2.0 licensed, can be installed from native packages or a Docker image, and has notebooks and documentation for getting started.

LocalAI is an open-source AI engine for running models of many kinds, including language, vision, voice, image and video models, on your own hardware. A GPU is not required. It is written in Go and released under the MIT license, was started by Ettore Di Giacinto and is looked after by the LocalAI team.

Its design is a small core with separate backends that are pulled on demand, so nothing unused gets installed. Backends wrap engines such as llama.cpp, vLLM, whisper.cpp, stable-diffusion and MLX, and you can write your own in any language against an open interface. It can run on CPU alone, through Vulkan, or on NVIDIA, AMD, Intel and Apple Silicon hardware.

APIs are drop-in compatible with OpenAI, Anthropic and ElevenLabs across every backend. Multi-user features include API key authentication, user quotas and role-based access, and built-in agents support tool use, RAG, MCP and skills. The project stresses that data stays on your infrastructure. It suits developers and teams who want a private, self-hosted replacement for cloud AI APIs.

SGLang is an inference framework, released as open source, for language models, vision-language models and diffusion models. It is optimized for agentic workloads, reinforcement-learning rollouts and large-scale serving. The project includes SGLang Diffusion, a built-in engine for image and video generation that ships in the same repository and Python package. It is released under the Apache-2.0 license.

Getting started takes either a prebuilt Docker image or a Python install using uv, followed by a launch command for the chosen model. A cookbook helps pick a model and hardware pair and generates a ready-to-run command. Supported hardware spans NVIDIA and AMD GPUs, Google TPUs, Intel GPUs and CPUs, Apple Silicon through Metal and MLX, and Huawei Ascend NPUs, with more integrations in progress. The wider SGLang ecosystem adds educational projects and community events.