AI Tools
vLLM
vLLM: preços, capacidades, atualizações e alternativas, lidos nas páginas do próprio desenvolvedor.

Em inglês
Key facts
- Price
- Not stated on the maker’s pages
- Free plan
- Not stated on the maker’s pages
- Platforms
- Linux · macOS · CLI · API
- API
- Yes Read on github.com
- Open source
- Yes · Apache-2.0 Read on github.com
- Latest version
- v0.34.0
- Last update
- 13 Nov 2026 Read on github.com
vLLM is an open-source engine for running language models and serving their outputs. It supports model loading, request batching, streaming, and deployment across supported accelerator and CPU hardware.
Developers and infrastructure teams use it to host model inference behind APIs, including OpenAI-compatible and Anthropic Messages interfaces. It is software to run on your infrastructure, not a hosted model service; the product pages do not confirm a paid plan or hosting price.
- Category
- Developer Tools
- Read from the maker’s pages
- 2026-11-18
Company
Features
- Uses PagedAttention to manage attention key-value memory.
- Batches incoming requests and supports chunked prefill and prefix caching.
- Supports quantization formats including FP8, INT8, INT4, GPTQ, and AWQ.
- Provides speculative decoding methods including n-gram, suffix, EAGLE, and DFlash.
- Supports tensor, pipeline, data, expert, and context parallelism.
- Serves models through OpenAI-compatible, Anthropic Messages, and gRPC APIs.
- Supports structured outputs, tool calling, and reasoning parsers.
- Open source under the Apache-2.0 license.
Use cases
- Serve language models behind an OpenAI-compatible API.
- Run batched inference for applications with concurrent requests.
- Deploy model inference across multiple GPUs or other supported hardware.
- Stream generated responses to an application.
- Host embedding, retrieval, or multimodal models for internal services.
Pricing
Free
Free
- Unlimited public/private repositories
- Dependabot security and version updates
- 2,000 CI/CD minutes/month
- 500MB of Packages storage
- Issues & Projects
- Community support
Team
$4/user/mo
- Access to GitHub Codespaces
- Repository rules
- Multiple reviewers in pull requests
- Draft pull requests
- Code owners
- Required reviewers
Enterprise
$21/user/mo
- GitHub Copilot access
- Agent governance
- Data residency
- Enterprise Managed Users
- User provisioning through SCIM
- Enterprise Account to centrally manage multiple organizations
Capabilities
- Runs commands — “vllm download-kernels # precompiled kernels for faster startup” source
- Command line — “vllm download-kernels # precompiled kernels for faster startup” source
- Choice of models — “vLLM seamlessly supports 200+ model architectures on Hugging Face, including:” source
- Self-hosted — “Install vLLM with uv (recommended) or pip :” source
- API — “OpenAI-compatible API server, plus Anthropic Messages API and gRPC support” source
- Runs models for you — “A high-throughput and memory-efficient inference and serving engine for LLMs” source
Latest updates
- Version 0.34.0rc3
Made the extensible KV cache exportable for EXPORTS.
- Version 0.34.0rc1
- v0.34.0rc0: [Transformers backend] Migrate Gemma and StableLM (#66393)
Migrated Gemma and StableLM to the Transformers backend.
- Version 0.33.0
Added support for listed models, upgraded to PyTorch 2.14.0, and dropped CUDA 12.9.
- v0.33.0rc2: [ROCm][Bugfix] Skip AITER tuned GEMM when M*N >= 2^31 on gfx950 (#64567)
Skip AITER tuned GEMM when M*N >= 2^31 on gfx950.
Where it works
Linux · macOS · CLI · API
Alternatives to vLLM
- Text Generation Inference
Text Generation Inference is an open-source toolkit for deploying and serving large language models through an API.
- llama.cpp
Run language and vision-language models on CPU or GPU with a C/C++ inference runtime and optional API server.
- LoRAX
LoRAX is an open-source inference server for serving multiple fine-tuned LoRA models through shared GPU infrastructure.
- Ollama
Ollama runs open AI models locally or in Ollama’s cloud, with a CLI and API for using them in apps and coding tools.
- LocalAI
LocalAI runs open AI models on your own hardware, with a web interface and APIs for chat, images, speech, video, and more.
- Baseten
Baseten provides infrastructure to deploy and run AI models, train models, and build agent workflows using managed inference, APIs, training tools, and sandboxes.
Questions about vLLM
Does vLLM have an API?
Yes. The maker’s pages state that vLLM has an API. Read on the maker’s pages on 18 Nov 2026.
Which platforms does vLLM run on?
vLLM runs on Linux, macOS, CLI and API, as its maker’s pages state. Read on the maker’s pages on 18 Nov 2026.
Is vLLM open source?
Yes. The code of vLLM is published at github.com/vllm-project/vllm under the Apache-2.0 licence. Read on the maker’s pages on 18 Nov 2026.
Is vLLM still available?
Yes, as of 18 Nov 2026: its maker’s pages were read that day and carried no notice that vLLM has ended.
What changed recently in vLLM?
The latest entry in its maker’s changelog: 13 Nov 2026, “Version 0.34.0rc3”. Before it: 11 Nov 2026, “Version 0.34.0rc1”. Changelog checked on 14 Nov 2026.
Sources
- https://github.com/vllm-project/vllm — Product description, capabilities, supported hardware, API interfaces, and Apache-2.0 license.
- https://github.com/vllm-project/vllm/releases — Latest release shown as v0.34.0, released November 16, 2026.
- https://vllm.ai/blog/2026-11-18-flydsl-monokernels — November 18, 2026 post about FlyDSL MonoKernels.
Not confirmed
- Prices
- No official pricing page was found in the search.
- Could not read an official pricing page.
- No official pricing was available to confirm.
- No official pricing page was found; tiers and prices are unconfirmed.
- Free plan
- No official pricing page was found to confirm a free tier.