Nestmodels for Linnet
Models / gpt-oss-20b

GPT-OSS 20B

A 21B-parameter mixture-of-experts decoder, 32 experts per layer with 4 active per token, alternating 128-position sliding-window and full causal attention with learned attention sinks, YaRN rope scaling for a 131072-token context, and expert weights read straight from the checkpoint's MXFP4 blocks and scales.

20.9B parametersgpt_ossApache-2.0text-generationdecoder-onlymixture-of-expertsgrouped-query-attentionattention-sinkssliding-window-attentionmxfp4
Hugging FaceGitHubarXivHomepage Registry entry

Greedy decoding of one chat prompt, through Linnet and through transformers (KV cache, argmax, bf16).

Prompt

Explain in two sentences why the sky is blue.

Linnet

analysisThe user asks: "Explain in two sentences why the sky is blue." They want a concise explanation in exactly two sentences. So we need to produce two sentences. The explanation: Rayleigh scattering of sunlight by air molecules, shorter wavelengths scatter more, so blue dominates. Also mention that the sky appears blue because of scattering of shorter wavelengths. So produce two sentences. Ensure exactly two sentences. Let's craft: "The sky appears blue because sunlight is scattered by the Earth's

transformers (KV cache, argmax, bf16)

analysisThe user asks: "Explain in two sentences why the sky is blue." They want a concise explanation in two sentences. So we need to produce exactly two sentences. The explanation: Rayleigh scattering of sunlight by air molecules, shorter wavelengths scatter more, making sky appear blue. Also mention that the sun's light is white, but blue is scattered. So produce two sentences. Ensure exactly two sentences. Let's craft: "The sky appears blue because sunlight, which contains

The first 24 of 96 generated tokens agree with the reference; after that the two stacks round a near-tie differently and part ways, which bf16 greedy decoding allows.