Browse models

Open models you can download and run. Pick your memory to see what fits.

Model results

Here are the closest that fit 16 GB.

Ready

An FP8 edition of Qwen3’s 8B chat model, with thinking and direct-answer modes.

Large
About 13 GB memoryChat & write9.44 GBapache-2.0585,833 catalog downloads
Ready

Answers a written prompt using information from a picture.

Large
About 13 GB memoryUnderstand images9.32 GBapache-2.08,496,525 catalog downloads
Ready

Generates text by continuing a written prompt.

Large
About 11 GB memoryChat & write8.04 GBapache-2.0664,826 catalog downloads

A 4B model dedicated to step-by-step reasoning, with native 256K-token context.

Large
About 11 GB memoryChat & write8.04 GBapache-2.0591,768 catalog downloads
Ready

Scores how closely a passage matches a search question.

Large
About 11 GB memoryRank search results8.04 GBapache-2.0649,962 catalog downloads

Scores how closely a passage matches a search question.

Large
About 11 GB memoryRank search results8.04 GBapache-2.01,755,665 catalog downloads

Turns text into number lists for search and meaning comparisons.

Large
About 11 GB memorySearch documents8.04 GBapache-2.02,446,340 catalog downloads
Ready

A 3.8B instruction model for reasoning and function calling, with 128K-token context.

Large
About 10 GB memoryChat & write7.67 GBmit619,945 catalog downloads
Ready

A 3.8B English instruction model for reasoning and code, with a 4K-token context window.

Large
About 10 GB memoryChat & write7.64 GBmit583,425 catalog downloads
Ready

A 3.8B English instruction model that activates 1.1B parameters per token.

Large
About 10 GB memoryChat & write7.51 GBmit831,711 catalog downloads
deepseek-ai

DeepSeek-OCR-2

Ready

A model for working with images and text together.

Large
About 9 GB memoryUnderstand images6.78 GBapache-2.03,202,247 catalog downloads

A 3B multilingual chat model for writing and structured JSON output.

Large
About 9 GB memoryChat & write6.17 GBqwen-research-non-commercial5,284,682 catalog downloads
HuggingFaceTB

SmolLM3-3B

Ready

A 3B chat model with optional reasoning and support for six languages.

Large
About 8 GB memoryChat & write6.15 GBapache-2.0663,378 catalog downloads

A 4-bit edition of a 7B multilingual chat model for writing and structured JSON output.

Medium
About 8 GB memoryChat & write5.55 GBapache-2.0604,743 catalog downloads

A 7.8B English and Korean chat model in AWQ, with a 32K-token context window.

Medium
About 7 GB memoryChat & write5.3 GBNo license stated613,538 catalog downloads

An FP8 edition of Qwen3’s 4B direct-answer model, with native 256K-token context.

Medium
About 7 GB memoryChat & write5.19 GBapache-2.0873,213 catalog downloads