The Best Open Source LLMs: 120+ Models Compared

Open source LLMs are one of the most important AI trends of 2026.

And for good reason:

Open source models were long significantly weaker than proprietary models. But by spring 2026 they have caught up, especially out of Chinese labs:

DeepSeek V4 Pro, GLM-5.1 from Z.ai, Kimi K2.6 from Moonshot AI, and Qwen3.5 from Alibaba can compete with the best proprietary LLMs like Claude Opus 5, GPT-5.5, or Gemini 3.1 Pro, and even beat them on specific benchmarks like SWE-Bench Pro and HumanEval. GPT-5.6 Sol, OpenAI's default Codex model since July 2026, set the new proprietary reference point for Terminal-Bench 2.1 at 88.8%, with Sol Ultra reaching 91.9%. On SWE-Bench Pro, Sol scores 64.6%, about 15 points behind Claude Mythos 5 and Fable 5. Comparable open-source Terminal-Bench 2.1 scores have not been published yet.

In this article, you'll find a sortable, filterable directory of 120+ open source LLMs, including benchmark scores, licenses, API prices, context windows, and capabilities.

Additionally, I'll show you how to easily and freely use open LLMs on your own computer (without needing to program or use the terminal).

All Open Source LLMs at a Glance

The directory contains every open-weights model from the models.dev catalog plus curated classics, sorted by release date by default. Use "Columns" to reveal more data, such as modalities, knowledge cutoff, max output, or the number of API providers:

1. Key Benchmarks Explained

To objectively compare open source LLMs, I use three central benchmark categories:

MMLU / MMLU-Pro: The Massive Multitask Language Understanding Benchmark tests general knowledge across 57 subjects (STEM, social sciences, humanities). MMLU-Pro is the more challenging variant with less contamination. Top models score 85-90% here.

MATH / GPQA: These benchmarks test mathematical and scientific reasoning. MATH-500 contains challenging math problems, while GPQA (Graduate-Level Google-Proof Q&A) tests expert knowledge in biology, physics, and chemistry. Top models score 70-97% here.

HumanEval / LiveCodeBench: These benchmarks test code generation. HumanEval contains Python programming tasks, LiveCodeBench tests code performance with current, uncontaminated tasks. Top models score 60-90% here.

The table shows up to three benchmark scores per model; the small label under each badge tells you which benchmark it is. Older and niche models don't have every score, in which case you'll see a dash.

SWE-bench Verified shows how close the top open models have come to the proprietary flagships:

Open models are close to GPT-5.5 and Opus 4.8; Fable 5 pulls ahead

Models:V4 ProK2.6Fable 5Opus 4.8GPT-5.53.1 Pro

DeepSeek V4 Pro

Kimi K2.6

Claude Fable 5

Claude Opus 4.8

GPT-5.5

Gemini 3.1 Pro

Sources: DeepSeek, Moonshot AI, Anthropic, Vals AI, Google DeepMind

PNG|SVG

CC BY 4.0

gradually.ai

The gap becomes even clearer on price. The leading open models deliver almost the same coding performance at a fraction of the API cost:

DeepSeek V4 Pro and Kimi K2.6: best price-performance in the scatter

Ideal: strong + cheap

DeepSeek

Moonshot AI

Anthropic

OpenAI

Google

Efficiency frontier (best price-performance)

Sources: official API price lists from DeepSeek, Moonshot AI, Anthropic, OpenAI, and Google

PNG|SVG

CC BY 4.0

gradually.ai

2. Top Models of April 2026

First, a look ahead: on July 16, 2026, Moonshot AI unveiled Kimi K3, a 2.8-trillion-parameter MoE with a 1M-token context window and native vision. That would make K3 by far the largest open-weight model ever, and it already scores 93.5% on GPQA Diamond. For now, K3 is available only via the app and API, priced at $3/$15 per million tokens (also on OpenRouter). The weights are slated to land on Hugging Face by July 27, 2026; Moonshot will state the exact license at release, though the previous Kimi models used Modified MIT. Once the weights are out, it will move into the directory below. Until then, the April standings apply:

DeepSeek V4 Pro is the new leader, released April 24, 2026. The 1.6 trillion parameter MoE activates only 49B per token and tops the field on MMLU-Pro and LiveCodeBench. Same MIT license as the rest of the DeepSeek lineup, and it ships with native 1M-token context. Full scores are in the table below.

Kimi K2.6 from Moonshot AI is the second-best open weight overall, leading on HumanEval, GPQA Diamond, and AIME 2026, with a 256K context window and native video input. It ships under the Modified MIT license with 1T parameters.

GLM-5.1 from Z.ai (formerly Zhipu) tops SWE-Bench Pro with 58.4%, beating even GPT-5.4 and Claude Opus 4.6. The 754B-parameter MoE was trained entirely on Huawei Ascend chips and ships under the MIT license. The reasoning sibling, GLM-5, holds the highest knowledge scores in the open-source space, including 96% on MMLU.

Kimi K2.5 still posts the highest HumanEval score on any leaderboard at 99.0, and leads on MATH-500 too. It's the best open weight purely for code generation when latency matters less than peak quality.

DeepSeek V4 Flash is the cost-efficient sibling of V4 Pro (parameter and activation details are in the table below) and the most practical choice when you want frontier-class quality on a single high-end GPU.

The previous generation is still very usable: GPT-OSS-120B (OpenAI's first open-weight model since GPT-2), DeepSeek R1, Qwen3-235B-A22B-Thinking, and Llama 4 Maverick all remain strong, just no longer state-of-the-art.

Here are the five new top models side by side:

3. LLM Licenses Explained

Here's an overview of the most commonly used licenses for open source LLMs.

MIT License

A very permissive open source license, similar to Apache 2.0. It allows unrestricted use, modification, and distribution of the LLM, including in proprietary programs, as long as the copyright notice is retained. DeepSeek V3 uses MIT with some restrictions for military use.

Llama 2 Community / Llama 3 Community

Meta released Llama 2 and Llama 3 under these licenses. They allow free use of the LLMs for research and commercial applications with up to 700 million monthly active users. The source code and model weights are freely available.

Qwen License / Qianwen LICENSE

Qwen models are released under various licenses. While smaller models are often licensed under Apache 2.0, larger models like Qwen2.5-72B have special license terms that allow commercial use with certain restrictions.

Apache 2.0

A very permissive open source license with minimal restrictions. It allows use, modification, and distribution of the LLM, including in proprietary programs, as long as the copyright notice is retained. It contains no copyleft clause.

CC BY-NC-4.0

A Creative Commons license that allows editing and sharing the LLM in any form, but not for commercial purposes. The author's name must be credited.

CC BY-NC-SA-4.0

Similar to CC BY-NC-4.0, but with the additional Share-Alike condition. This means forks or modified versions of an LLM must be distributed under the same conditions.

Non-Commercial

Here, using the LLM for commercial purposes is prohibited. However, what exactly counts as "commercial" is not always clearly defined or delimited.

Usually, "non-commercial" models are only released for research purposes or private use.

4. Using Open Source LLMs Locally on Your Own Computer

Using open source LLMs locally on your own computer is easier than you might think:

1. Download LM Studio

Download LM Studio from the website. It's free and available for Mac, Windows, and Linux:

2. Install and Open LM Studio

Next, install LM Studio on your computer and open it.

3. Download Your Desired Open Source LLMs

Now you need to download the open source LLMs you want to use in LM Studio.

Many popular LLMs are already on the home screen. To download an LLM, simply click the blue download button:

To find specific open source LLMs, you can also use the search function:

4. Important: Check System Requirements Before Downloading

Before downloading an LLM, you should check the system requirements.

Llama 3, for example, requires more than 8 GB RAM and 4.92 GB of free storage:

5. Chat with the Open Source LLM

After downloading an open source LLM, you can use it directly in LM Studio.

Simply click on the speech bubble icon (?) in the left sidebar.

The user interface and settings options are reminiscent of the OpenAI Playground: