Which model, on which hardware.

Say what you want to run and on what. Every pair below is checked against the memory it needs; speed is measured where we have a benchmark, otherwise estimated and labeled as an estimate.

Most recommended

What developers are shipping with

All 141 models →

Claude Fable 5

Anthropic

1M ctx

$10 in / $50 out per 1M

Claude Fable 5 (batch)

Anthropic

1M ctx

$5 in / $25 out per 1M

Claude Fable 5.1

Anthropic

1M ctx

$10 in / $50 out per 1M

Claude Fable 5.1 (batch)

Anthropic

1M ctx

$5 in / $25 out per 1M

Claude Opus 4.8

Anthropic

1M ctx

$5 in / $25 out per 1M

Browse by need

Start with your workflow or setup

Best LLMs for Agents & Tools

Cloud and local models that handle tool calls, multi-step work, and agentic workflows.

106 models

Best Local LLMs for Hermes Agent

Private local models suited for Hermes-style agent runs — tool-capable, steady, and no per-token meter.

25 models

Local & Open-Weight LLMs

The full private/offline catalog — open-weight and local-first models in one place.

25 models

Best LLMs for a 24GB GPU

Strong local models for a 4090-class setup — high capability without cloud inference.

21 models

Best LLMs for an 8GB GPU

Compact local models for laptop and entry-level GPUs — practical picks that still fit.

13 models