gpt-oss-20b

gpt-oss-20b is useful to evaluate because it combines a specific architecture, license and operating envelope rather than simply adding another row to a model leaderboard.

The 20B model follows the same gpt-oss training philosophy as the 120B release: mostly English text pretraining with strong STEM and coding content, then supervised fine-tuning and reinforcement learning for reasoning and instruction following.

From an infrastructure perspective, the key sizing facts are 20.9B total parameters, 3.6B/token active parameters and 131,072 tokens of context. These values should be read together: weight memory, active compute and KV-cache growth describe different resource constraints.

From a governance perspective, the checkpoint is distributed under Apache License 2.0. OpenWeightModels records that independently from the fact that the weights are downloadable, because “open weight” is an access classification—not a universal statement about commercial, redistribution or derivative-work rights.

gpt-oss-20b uses sparse expert routing rather than activating its complete parameter pool for every token. The distinction between 20.9B total parameters and 3.6B/token active parameters is central to capacity planning: active parameters influence compute per token, while the complete expert set still affects storage, model loading, sharding and network topology.

This is why OpenWeightModels never maps the active-parameter figure directly to a dense-model hardware class. A sparse model can be computationally efficient per token and still require datacenter-class memory and interconnect to keep all experts available.

Publisher-documented architecture details for this checkpoint include the elements below. These details are more useful for deployment planning than a parameter count alone because attention layout, expert routing, modality encoders and context design can affect throughput and memory independently.

OpenAI positioned gpt-oss-20b for memory-constrained and local scenarios and stated that it can run with roughly 16 GB of memory in suitable deployments. Real requirements vary with context, runtime and acceleration.

Real memory use includes model weights, KV cache, activations/buffers, multimodal encoders when applicable, framework overhead and batching. A published “fits on” statement is meaningful only together with precision, context, batch and host/GPU topology.

gpt-oss-20b is most relevant when the application specifically values reasoning and tool-oriented workflows on more accessible local or edge-class hardware. That should be balanced against its hardware class, context behavior, license and modality requirements.

A smaller model can be operationally superior when latency, privacy, device deployment or predictable cost matter more than peak benchmark capability. Conversely, a larger sparse model may justify its complexity when the workload benefits from higher capacity, sophisticated reasoning or agent behavior.

The selection decision should therefore compare at least five dimensions: required task quality, data/control requirements, legal eligibility, serving cost and ecosystem/runtime support. OpenWeightModels exposes these dimensions separately so a model is not selected on benchmark reputation alone.

Last full review: 29 September 2026. Publisher model cards, repositories and license texts are preferred. Runtime claims are attached to the relevant runtime/provider rather than inferred from the base checkpoint.

Where documentation conflicts or a field is ambiguous, the passport uses the more conservative interpretation and explains the discrepancy instead of silently selecting the largest number.