Open-weight models

Qwen 3 flagship MoE. 235B total / 22B active per token, with built-in 'thinking' and 'non-thinking' modes that trade speed for reasoning dep

Mid-tier Qwen 3 MoE. 30B total / 3B active means 70B-class quality at 7B-class inference speed on a single 24GB card. The sweet spot of the

Coding-specialist Qwen 2.5. Beats GPT-4o on HumanEval and matches Sonnet on many code-edit benchmarks. The default local-coding model on 24G

Dense Qwen 3 32B. Best dense open-weight model in its size class at release; pairs nicely with a single RTX 5090 or 4090.

Qwen 3 at the 8B scale. Direct head-to-head against Llama 3.1 8B on most benchmarks; usually wins on coding and structured output.

14B Qwen 3. Fits on 12GB cards at Q4. Strong default for users with a single mid-range GPU.

The community-default small Qwen prior to Qwen 3. Still widely used because of mature ecosystem support.

14B Qwen 2.5. Sweet spot for 16GB VRAM. Many production deployments still on this version.

Dense 32B Qwen 2.5. Strong daily-driver on 24GB cards prior to Qwen 3 32B.

Qwen team's reasoning-focused experimental release. Visible chain-of-thought in <think> tags. Precursor to Qwen 3's thinking mode.

The flagship of Qwen 2.5. Workstation-tier; needs 48GB+ VRAM for usable inference.

Compact Qwen 3 for edge and laptop deployment. Outperforms many 7B models from prior generations.

Meta's small flagship. Strong general reasoning, 128K context, broad multilingual. The default first try for most local-AI use cases on cons

Meta's 2026 flagship MoE model. 109B total parameters with only 17B active per forward pass and a record 10-million-token context window — u

Late-2024 refresh of the 70B Llama line. Roughly matches Llama 3.1 405B on most benchmarks at one-fifth the parameter count. The default hig

Lightweight 3B for edge and laptop deployment. Runs comfortably on 8GB VRAM at 30+ tok/s on Apple Silicon.

The 70B sibling of Llama 3.1 8B. Strong generalist reasoning with 128K context, popular base for agentic fine-tunes (Hermes 3, Nemotron). Mo

NVIDIA's HelpSteer2-tuned Llama 3.1 70B. Topped Arena Hard at release. The pre-Nemotron-3 NVIDIA reference open weights.

First-party multimodal Llama. Accepts images alongside text for VQA, document understanding, and chart reading. Runs on 12GB+ VRAM.

Meta's high-end Llama 4 sibling — 128 experts MoE built for performance over efficiency. Multilingual strength is its standout. Effectively

NVIDIA's top open reasoning model in the Llama 3.1 lineage. Server-tier; trained for groundbreaking reasoning accuracy on agentic workloads.

Smallest of the Nemotron reasoning trio. NAS-optimized for inference efficiency on RTX hardware.

True edge-tier Llama. Runs on a phone or Raspberry Pi. Useful for classification, simple summarization, and on-device agents.

The 90B vision Llama. Best-in-class first-party multimodal open weight at the time of release. Workstation-class only.

Open reasoning model that closed the gap with frontier proprietary reasoners. Visible chain-of-thought, MIT license, and a family of distill

Reasoning distillation onto Llama 3.3 70B. Best-in-class open-weight reasoner you can actually fit on a workstation.

32B distill — fits on a single 24GB card with reasoning capability. Best price-per-thinking-token combo for prosumers.

DeepSeek's flagship MoE — 671B total / 37B active. Server-tier, but the smaller R1 distills make this lineage approachable.

Smallest practical R1 distill. Reasoning on a 6GB GPU.

14B reasoning distill. Fits on 12GB cards.

MoE coding specialist — 16B total / 2.4B active. Fast on 12GB cards.

Google's flagship dense Gemma 4. Beats some 400B-class proprietary models on benchmarks. Targets the 24GB single-GPU sweet spot.

MoE variant of Gemma 4. Faster per-token than the 31B dense at similar quality on most tasks.

Pre-Gemma-4 flagship. Multimodal (4B+ variants), 128K context, 140 languages. Strong daily driver on 24GB cards.

Edge-class Gemma 4. The 'Effective 4B' branding signals it punches above its parameter count via training-data quality.

12B Gemma 3. Fits on 12GB consumer cards. Multimodal.

4B Gemma 3 for edge. Multimodal.

Mid-size Gemma 2. Strong chat quality with a different training mix from Llama family.

Smallest Gemma 4. Designed for phones and Raspberry-Pi-class hardware.

Smallest text-only Gemma 3 for phones and IoT.

Medical-specialist Gemma fine-tune. Trained on de-identified medical literature and imaging. Research use under HAI-DEF terms.

Coding-specialist Gemma. Decent FIM completion. Now mostly historical with Qwen 2.5 Coder dominating.

Zhipu's GLM-5 currently leads the Open LLM Leaderboard 2026. Strong reasoning and bilingual EN/ZH capability.

NVIDIA's hybrid Mamba-2 + Transformer MoE for on-device agents. 30B total / 3B active. 1M-token context window with reasoning ON/OFF modes a

Moonshot's long-context, agent-oriented MoE. Optimized for stability under tool use and multi-step coding/planning workflows.

Workstation-tier Nemotron 3. 120B total / 12B active. 5× higher throughput than the prior Super, 1M context, designed for multi-agent applic

Fully-open OLMo 2. AI2 publishes the full training data, code, and weights — the most reproducible 32B model.

Re-release of Mistral Small under Apache 2.0. Competitive with Llama 3.3 70B at one-third the size for many tasks.

Joint Mistral/NVIDIA release with native 128K context and a new Tekken tokenizer. Strong multilingual; popular fine-tune base.

Mistral's multimodal entry. 12B parameters, vision + text, Apache 2.0. Good document and chart understanding.

Mistral's coding-specialist. Strong fill-in-the-middle for IDE autocompletion. Personal/research use only.

Mistral's flagship dense model. Open weights but restricted commercial license — research and non-commercial only.

The reference 7B from Mistral. Apache 2.0 with native function calling. Mature ecosystem.

Microsoft's Phi-4 14B trained on synthetic textbook-quality data. Punches above weight on reasoning and math; MIT licensed.

Reasoning-focused fine-tune of Phi-4. Visible chain-of-thought, competitive with much larger models on math and STEM benchmarks.

Compact 3.8B Phi for edge deployment. 128K context. Strong reasoning per parameter.

Multimodal Phi 3.5. Document and chart understanding at edge size. MIT licensed.

NousResearch's Hermes fine-tune of Llama 3.1 8B. Stronger system-prompt adherence, JSON output, role-play, and agent steering than the base

Hermes 3 at 70B. Workstation-tier agent-tuned model.

Eric Hartford's Dolphin fine-tune of Mistral Small 3 — uncensored, function-calling, agent-friendly.

The MoE model that introduced the 8-experts pattern to the open-weight world. 47B params total, 13B active. Still a viable workhorse on 36GB

The bigger Mixtral. 141B total / 39B active. Strong general model, workstation-tier deployment.

Cohere's flagship — RAG-tuned, multilingual. Open weights but non-commercial.

Cohere's mid-tier — RAG and tool use. Non-commercial license.

01.AI's 34B model. Solid bilingual EN/ZH performance, Apache 2.0.

Microsoft's RLHF-heavy fine-tune of Mixtral 8x22B. Briefly the top open chat model on LMSYS at release.