AI Ownership Index
The library
Every open model and inference provider we grade, ranked by the AI Ownership Index. The letter and score are the analytical input; the ownership verdict, green through red, is the thing to read.
Models
B80.8Ai2 OLMoOLMo is the Allen Institute for AI's "fully open / open science" language-model family (OLMo 2 and OLMo 3, the latter including a 32B).fully_open · Apache License 2.0substantial75%B78.8OpenAI gpt-ossgpt-oss is OpenAI's first open-weight release since GPT-2 (August 2025): two mixture-of-experts reasoning models - gpt-oss-120b (~117B total / 5.1B active, 128 experts) and gpt-oss-20b (~21B total / 3.6B active, 32 experts) - both with a 128K context window, configurable low/medium/high reasoning effort, native agentic tool use (function calling, browsing, Python) and structured outputs, shipped under a clean OSI-approved Apache-2.0 license with no click-through gate.open_weights · Apache License 2.0substantial85%B73.2Mistral AIMistral AI is the Paris-based EU lab whose open-weight releases are the leading European alternative to the US open-model families.open_weights · Apache 2.0 (flagship open-weight models); Mistral Research License / MNPL for some modelspartial85%B73.2SoofiSoofi-S is an open-science project in the OLMo mould whose pretraining report commits it to the Open Source AI Definition (OSAID 1.0) - a pledge to release weights, intermediate checkpoints, full per-source data accounting, hyperparameters and training+eval code - but as of this grounding that release has not landed: the base weights are in a closed-beta phase ("open model weights coming soon"), the open licence is not yet finalised (the LICENSE file is access-restricted and carries no SPDX), and it remains an early preview/beta with a custom architecture and incomplete safety hardening.open_weights · Permissive open-source committed (OSAID 1.0) but not yet finalised - SPDX pending, LICENSE access-restrictedpartial75%B72.4Meta LlamaMeta's Llama family is the most widely adopted open-weight LLM lineage, spanning dense Llama 3.1 (8B/70B/405B, 128K context) and Llama 3.3 70B through the Llama 4 mixture-of-experts models (Scout, Maverick, Behemoth), shipped alongside the companion Llama Guard and Prompt Guard safety classifiers.gated_open · Llama Community Licensepartial91%C69.2EuroLLMEuroLLM trades frontier scale for a genuinely European, Apache-2.0, all-EU-language model with the cleanest possible EU AI Act posture - an open-weight model designed in and for the EU, well under any systemic-risk threshold, whose natural home is multilingual EU-facing workloads rather than beating the largest labs on English reasoning.open_weights_recipe · Apache License 2.0substantial79%C67.6QwenQwen is Alibaba's large, fast-moving open-weight family.open_weights · Apache License 2.0 (mainstream Qwen3 sizes; verify per checkpoint)partial74%C65.2GLMGLM is Zhipu AI's open-weight family - Zhipu AI is the Beijing company (spun out of Tsinghua's THUDM lab) that ships internationally under the "Z.ai" brand.open_weights · MIT License (GLM-4.5 / GLM-4.6 / GLM-4-0414 series; verify per checkpoint)partial74%C64.4DeepSeekDeepSeek's V3 and R1 are large mixture-of-experts models (V3 is ~671B total / ~37B active) whose open weights are released under the OSI-approved MIT license - an unusually permissive posture that allows commercial use and even distillation, and which has produced a large family of R1 distills into Llama and Qwen bases.open_weights · MIT Licensepartial80%C62.8KimiKimi is Moonshot AI's (Beijing) model/product line; its confirmed open release is Kimi K2 - a very large Mixture-of-Experts model with 1T total / 32B active parameters (384 experts, 8+1 selected), a 128K context window, and an explicit agentic/coding focus (trained with the MuonClip optimiser).open_weights · Modified MIT License (MIT plus a large-scale attribution clause)partial75% Inference providers
B75.2Together AIA US-headquartered inference provider serving a broad open-model catalogue (Llama 3.x/4, DeepSeek V3/R1, Qwen3 incl.United Statespartial56%B74InfercomA Luxembourg-headquartered EU-sovereign Inference-as-a-Service provider running open-weight models in native BF16 on SambaNova RDU dataflow hardware, with EU-routed inference hosted in Munich, Germany, under 100% EU jurisdiction.Luxembourgsubstantial52%B70.4Berget AIA Stockholm-based, Swedish-owned EU-sovereign inference provider serving open-weight models (Llama, Mistral-Small, Gemma, gpt-oss, GLM, plus embedding/reranking models) over an OpenAI-compatible API.Swedensubstantial57%B70GroqA US-headquartered inference provider running open models on custom LPU hardware for very low latency, serving a curated/narrower open catalogue (Llama 3.1 8B, Llama 3.3 70B, gpt-oss 20B/120B, Qwen3, Gemma, Mixtral, DeepSeek R1 Distill Llama 70B).United Statessubstantial56%C60.4DeepInfraA US-headquartered, low-cost inference provider serving ~77-90+ open models (Llama 3.x/4.x, DeepSeek V3/V3.2/V4 variants, Qwen3, Kimi K2, GLM-5, gpt-oss-120B, MiniMax-M2, Nemotron, Gemma) over pay-as-you-go per-token pricing plus by-the-hour dedicated GPUs, and is among the cheapest providers available.United Statespartial49%D45.6RunwareAn inference platform ("one API for all AI") that runs a large catalogue of open-weight Hugging Face models on its own custom stack, the Sonic Inference Engine, across pods in the US and central Europe.United Kingdompartial44%