The best open AI models, by what actually matters

Honest, objective shortlists — ranked by what fits your GPU, what your license allows, context length, run-cost, and task, not by a quality score we made up.

By your hardware & cost

best open LLMs you can run on 8 GB of VRAM — computed VRAM-to-run (weights + KV cache at default quant) <= 8 GB, ordered by paramsB descending
best open LLMs you can run on 12 GB of VRAM — computed VRAM-to-run <= 12 GB, ordered by paramsB descending
best open LLMs you can run on 16 GB of VRAM — computed VRAM-to-run <= 16 GB, ordered by paramsB descending
best open LLMs you can run on 24 GB of VRAM — computed VRAM-to-run <= 24 GB, ordered by paramsB descending
best open LLMs you can actually run on a laptop — computed VRAM-to-run <= 6 GB (conservative laptop budget — leaves room for the OS), ordered by paramsB descending
cheapest open LLMs to run via your own API key — models with a REAL (non-estimated) API price, ordered by blended $/1M tokens ascending
cheapest open LLMs to run on your own hardware — all runnable LLMs, ordered by computed VRAM-to-run ascending

By license

best open LLMs you can use commercially — commercialUse === true, ordered by popularity descending
best Apache-2.0 and MIT licensed open LLMs — license matches Apache-2.0 or MIT, ordered by popularity descending

By capability

best small open LLMs (under 4B parameters) — paramsB < 4 (tiny tier), ordered by popularity descending
best long-context open LLMs (128K+ tokens) — contextWindow >= 131072, ordered by contextWindow descending
most recognized open LLMs for coding — models tagged 'coding' or named as a coder (e.g. *-Coder), ordered by popularity descending
most recognized open vision and multimodal LLMs — models tagged 'vision' or named as a vision-language model (e.g. *-VL), ordered by popularity descending

Media generation

most recognized open text-to-image models you can run yourself — media modality 'image', ordered by notability descending
most recognized open video-generation models you can run yourself — media modality 'video', ordered by notability descending
most recognized open text-to-speech and voice-cloning models — media modality 'audio' filtered to text-to-speech + voice-cloning tasks, ordered by notability descending
most recognized open speech-to-text models (Whisper and alternatives) — media modality 'audio' filtered to speech-to-text task, ordered by notability descending

Cost calculator → · Glossary →