Models, open and hosted

Context windows, prices, licences, quantizations and hardware needs for 2,176 models. Every fact names its source and the day it was read. Where the sources are silent, the page says Unknown.

2,176 models in the October 9, 2026 count
Free to browse, with a source and a date on every fact

Check what your hardware runs Get set up

Browse the models

Search by name or maker, then narrow the list by licence, context window or the memory you have.

Showing the first 40 of 2,176 models.

Sample records. Names, makers and the 32 GB fits come from the live Can I run it check, October 25, 2026; the other cells show the layout. A fact the sources leave open reads Unknown.
Model	Maker	Parameters	Context window	Licence	On a 32 GB card
Llama-4-Scout-17B-16E-Instruct	Meta	17B active, 16 experts	10,000,000	`Llama 4 Community`	`IQ2_XXS`, 28.1 GiB
Ling-flash-2.0	inclusionAI	100B, 6.1B active	32,768	`MIT`	`IQ2_XXS`, 25.3 GiB
Solar-Open-100B	Upstage	100B	Unknown	Unknown	`IQ2_XXS`, 25.3 GiB
Qwen3-Next-80B-A3B-Instruct	Qwen	80B, 3B active	262,144	`Apache-2.0`	`Q2_K`, 27.2 GiB
Qwen3-Next-80B-A3B-Thinking	Qwen	80B, 3B active	262,144	`Apache-2.0`	`Q2_K`, 26.2 GiB
Hunyuan-A13B-Instruct	Tencent	80B, 13B active	262,144	`Tencent Hunyuan`	`Q2_K`, 27.4 GiB
Qwen2-VL-72B-Instruct	Qwen	72B	32,768	`Qwen`	`IQ2_XXS`, 23.7 GiB

Check other hardware

A page for each model

Each of the 2,176 models has its own page, listed in the models sitemap, with the facts a harness needs before it picks one.

Models / Qwen

Qwen3-Next-80B-A3B-Instruct

By Qwen. 80B parameters with 3B active, a context window of 262,144 tokens and the licence Apache-2.0.

Context window: 262,144 tokens
Licence: Apache-2.0
On a 32 GB card: Q2_K, 27.2 GiB, with up to 16,384 tokens of context
Speed there: 249 to 399 tokens a second, an estimate

Check it against your hardware

FactsMaker, parameters, context window and licence, each linked to its source.
PricesEach provider's input and output price, the context it serves and the day it was read.
Run it on your own hardwareThe memory each quantization needs.
Use it from your harnessThe settings your harness needs to call the model.
Published resultsBenchmark results, linked to where they were published.
Sources of this pageEvery source, with the date it was read.
Paid linksThe page's own statement on paid links.

Every fact has a source and a date

Named sources

Each fact links to the page it came from. Sizes and architecture numbers come from the model's own page, and prices come from each provider's page.

Dated readings

Facts that change, such as prices, show the day they were read. The hardware presets on Can I run it? were read September 24, 2026.

Unknown, in plain words

When the sources leave a fact open, the page says Unknown, so your agent knows to check before relying on it.

Give your agent the same facts

Agent Feeds collections point your agent at the sources behind these pages. Each one is a public address, in Markdown or JSON Feed.

Use Agent Feeds

Compare model cost and latencyMarkdown JSON Feed
Run models on your hardwareMarkdown JSON Feed
Choose a coding modelMarkdown JSON Feed
Assess a model releaseMarkdown JSON Feed

Paid links

Every listing and link on these pages is unpaid. The order and the facts come from the named sources.

Picked a model? Give it the right files.

Pro gives each step the skills, tools and settings it needs, with your own model access.

Get started Compare plans

Pro

$29a month

Includes Agent Feeds + Skills and agent files

The whole library, 1,560,856 files
One downloaded item is the measured unit
A Baltor key for each tool you connect

Model usage stays with your provider. Cancel any time from your account page.