Model Catalogue

A place
to test
local models

Curated local and open-source AI models. From community workhorses to experimental frontier architectures.

Test Local ModelsTry models instantly in your browser without installing them.
Check the specsCompare parameters, context sizes, and find your niche.
Exchange knowledgeDiscuss hardware configurations with other open source enthusiasts.

Model of the Month

Spotlight — June 2026

Qwen3.5-27B
OpenNewPopular

Alibaba's Mixture-of-Experts flagship activates only 22B of its 27B parameters per token. It offers frontier-class reasoning with remarkable efficiency. Hybrid thinking mode toggles chain-of-thought on or off at inference time. Outperforms GPT-4o on multiple benchmarks while remaining fully locally runnable on high-end consumer rigs.

235B / 22B activeParameters
128KContext
Apache 2.0License
MoEArchitecture
100+ langsLanguage
Key:
OpenApache / MIT / similar
NewReleased last 90 days
ExperimentalEarly / research stage
PopularCommunity favourite
LocalConsumer hardware

Language & Reasoning Models

View all
H3
OpenPopular
Hermes-3-405B
Nous Research

Based on Hermes-3-Llama-3.1-405B, built for thoughtful discussion and helpful, structured reasoning. Strong multilingual generation, precise instruction-following, and clean formatting across technical and analytical tasks.

405BParams
1MContext
Llama 3.1License
DenseArch
Explore Model
H4
OpenNew
Hermes-4-70B
Nous Research

The newest Hermes generation in an efficient 70B class. Balanced speed and reasoning depth with controlled verbosity, reliable structured output, and strong code generation, debugging, and optimization.

70BParams
1MContext
Llama 3.1License
DenseArch
Explore Model
Φ4
OpenNew
Phi-4
Microsoft Research

Phi-Reasoning delivers long, thorough reasoning responses for deep research and complex problem solving. Optimal for in-depth analysis, academic assistance, and technical explanations. Can analyse PDF documents.

14BParams
CoTMode
MITLicense
PDF ✓Files
Explore Model
Q3
OpenNewLocal
Qwen3.5-27B
Alibaba / Qwen Team

A 27B dense hybrid model pairing Gated Delta Networks with attention for long-context efficiency. Strong at coding, reasoning, and agent tasks across 201 languages and dialects under Apache 2.0.

27BParams
262KContext
Apache 2.0License
201 langsLanguage
Explore Model
NM
OpenLocal
Mistral-NeMo
Mistral AI / NVIDIA

A 12B collaboration between Mistral and NVIDIA, designed to be a drop-in efficient workhorse. State-of-the-art multilingual coverage, large context, and strong instruction following on a single consumer GPU.

12BParams
128KContext
Apache 2.0License
Multiling.Language
Explore Model
TM
NewExperimentalLocal
Trinity-Mini
Open-Source-Lab

A compact, efficiency-first assistant model tuned for fast everyday reasoning, summarization, and chat. Designed to deliver responsive performance on modest hardware while staying lightweight.

MiniClass
128KContext
EfficientProfile
DenseArch
Explore Model

Code & Engineering Models

View all
LG
NewPopular
Laguna-M.1
Poolside

Poolside's leading coding agent model for intricate software development and agentic engineering workflows. Excels at repository-scale reasoning, debugging, refactoring, and autonomous development with strong tool-calling.

128KContext
8K outOutput
fp8Quant
AgentFocus
Explore Model
G5
OpenNew
GLM-5.1
Zhipu AI

An advanced model combining deep engineering productivity with speed and emotional intelligence. Excels at autonomously driving end-to-end project execution, debugging, and log analysis. Broad file processing support.

754BParams
205KContext
Zhipu AIMaker
MultifileFiles
Explore Model
M2
OpenNew
Minimax-M2.7
MiniMax

A fast, next-generation model combining engineering productivity with speed. Stands out at autonomously driving end-to-end project execution and handling complex tasks like log analysis and debugging.

205KContext
MultimodalFiles
AgenticFocus
FastProfile
Explore Model
MV
OpenNew
MiMo-V2-Flash
Xiaomi MiMo

A cheap MoE reasoning model designed for coding and agent performance. Dynamically routes tokens through specialized experts with a hybrid attention architecture for strong scalability and fast inference.

309B / 15BParams
256KContext
MoEArch
MultimodalFiles
Explore Model

Vision & Multimodal

View all
S4
OpenNewLocal
Mistral-Small-4
Mistral AI

A unified iteration consolidating several Mistral models into one system: Magistral's reasoning, Pixtral's multimodal understanding, and Devstral's agentic coding. Handles analysis, software dev, and visual tasks in one workflow.

MultimodalModality
ReasoningMagistral
CodingDevstral
VisionPixtral
Explore Model
N2
New
Nova-Lite-2
Amazon

A cost-effective multimodal foundation model that efficiently processes text, images, and video — prioritizing speed and affordability. Massive 1M-token context with adjustable verbosity and temperature controls.

1MContext
Text+Img+VidModality
AmazonMaker
PDF/JPGFiles
Explore Model

Ready to Go Local?

Start running open models today. No API keys, no subscriptions, no data leaving your machine.

Get StartedView Benchmarks

Account Access

Welcome back

Sign in to your Open-Source-Lab account to track models, talk to the community and use internal tools.

No account yet? Create one

Get Started

Create account

Join the Lab. We'll send a verification code to confirm your email.

Already have an account? Sign in

Verify Email

Check your inbox

We sent a 6-digit verification code to . Enter it below to activate your account.

Demo mode: real email delivery needs a backend. Your code is shown below.
------

Signed In

You're all set

Welcome to Open-Source-Lab. Your account is verified and active.

Benchmarks

Model Benchmarks

Reasoning, knowledge and coding scores for catalogue models alongside reference open-source baselines. Click any column header to sort.

ModelTypeMMLUGPQAHumanEvalMATHAvg

MMLU = general knowledge · GPQA = graduate-level science (Diamond) · HumanEval = code generation · MATH = competition mathematics. Scores are representative composite figures gathered from public reports and provider cards for orientation only — verify against primary sources before citing. Reference baselines are widely-used open-weight models included for comparison.

Community Hub

Model Discussions & Votes

Upvote local open-weights models and share hardware configurations, execution speeds, or setup reviews.

Catalog Models

Qwen3.5-27B

Alibaba / Qwen

0 votes
commenting on Qwen3.5-27B