Public API · v1

Embed a compatibility badge

Tell readers at a glance whether the model in your blog post / README / model card will run on the hardware you chose. Live SVG, no JS, ~1 KB.

Endpoint

GET https://runthismodel.com/api/badge/{modelId}
       ?gpu={gpu name}     // e.g. "RTX 4090", "M3 Max", "RX 7900 XTX"
       &vram={GB}          // optional, overrides our GPU lookup
       &ram={GB}           // system RAM (Apple Silicon: unified memory)

With no params, the badge shows the model’s minimum-VRAM requirement. Pass any combination of gpu, vram, ram and the badge grades S–F against that hardware.

Llama 3.1 8B on RTX 4090

↑ live image — refresh to re-grade

URL:

https://runthismodel.com/api/badge/meta-llama-3.1-8b-instruct?gpu=RTX+4090

Markdown:

[![Llama 3.1 8B on RTX 4090](https://runthismodel.com/api/badge/meta-llama-3.1-8b-instruct?gpu=RTX+4090)](https://runthismodel.com/models/meta-llama-3.1-8b-instruct)

Qwen 2.5 32B on M3 Max

↑ live image — refresh to re-grade

URL:

https://runthismodel.com/api/badge/qwen2.5-32b-instruct?gpu=M3+Max&ram=64

Markdown:

[![Qwen 2.5 32B on M3 Max](https://runthismodel.com/api/badge/qwen2.5-32b-instruct?gpu=M3+Max&ram=64)](https://runthismodel.com/models/qwen2.5-32b-instruct)

FLUX.1 Schnell — minimum spec only

↑ live image — refresh to re-grade

URL:

https://runthismodel.com/api/badge/flux1-schnell-gguf

Markdown:

[![FLUX.1 Schnell minimum spec](https://runthismodel.com/api/badge/flux1-schnell-gguf)](https://runthismodel.com/models/flux1-schnell-gguf)

Phi-3.5 MoE on a 16 GB GPU

↑ live image — refresh to re-grade

URL:

https://runthismodel.com/api/badge/phi-3.5-moe-instruct?vram=16&ram=32

Markdown:

[![Phi-3.5 MoE on 16 GB](https://runthismodel.com/api/badge/phi-3.5-moe-instruct?vram=16&ram=32)](https://runthismodel.com/models/phi-3.5-moe-instruct)

Notes

Color encodes the grade — S/A green, B blue, C yellow, D orange, F red.
For LLM/code/multimodal models, the right side shows estimated tok/s (e.g. B · ~28 tok/s) when hardware is provided.
Mixture-of-Experts models speed-grade against active parameters but VRAM-grade against total parameters — the same math your model detail page uses.
Cache: 1 hour on the browser, 1 day on Cloudflare’s edge. Hardware tweaks update on next refresh.
CORS: Access-Control-Allow-Origin: * — works in any browser, any origin.
Need a different hardware preset or a deep-link target? Open an issue or grab the model id from the URL of any model page.