Benchmark result

granite-4.2-3b on NVIDIA GeForce RTX 3060 Laptop GPU — 38.9 tok/s

Measured with Bionic manual on October 26, 2026.

ShareRedditX

Support llm-bench.io

llm-bench.io is free. If it helped you, I'd appreciate your support to cover the infrastructure costs.

Support me on Ko-fiBuy me a coffee

What the model built

Open full screen →

The coding scenario asks for a playable game in a single HTML file. This is exactly what the model returned, unedited; the frame only adds a line that tells this page how tall it is.

Reproduce this run

toolBionic manual
modelgranite-4.2-3b
context16,384
clientv0.4.96+102

Set those in Bionic, then:

llm-benchmark benchmark --model "granite-4.2-3b" --tool "Bionic"

The client prompts for context length and thinking mode, and for the KV cache dtype on Unsloth Studio. Sampling is left at the model default — the values above are what the tool reported using, not overrides the client sent.