Opeli has developed some modern models of open language optimized for consumer GPU. IN Blog postOpeli announced “GPT-OSS-120B” and “GPT-OSS-20B”, the first designed to work on one GPU processor by 80 GB, and the second optimized for operation on EDGE devices with only 16 GB of memory.
Both models operate the transformer using an expert mix model, popularized model using Deepseek R1. Despite the concentration of GPT-OS-120B and 20B design on consumer GPU, both support up to 131,072 context lengths, the longest available for local inference. GPT-OS-120B activates the parameters of 5.1 billion on token, and GPT-OS-20B activates 3.6 billion parameters per token. Both models operate alternating dense and locally banded attention patterns and operate a multi -story group with group 8.
Both models operate the architecture of reasoning reasoning with mixed emphasis on reasoning, performance and utility in the real world. Two GPT-OSS models are also the first language models open since GPT-2. Open AI models are similar to Open Source software, providing easier availability of programmers. Opeli decided on the two latest Open Source models to raise adoption in emerging markets and other sectors that may not be able to accept their own models.
The GPT-OS-1220B model allegedly reaches almost identical performance with the outgoing O4-Mini language model based on basic comparative tests, but is able to do all this on one 80 GB graphics processor. GPT-OSS-20B provides similar performance to the outgoing Model of the O3-Mini OpenAI language, and at the same time was able to operate on devices with only 16 GB of memory.
In conducting OPENAI, GPT-OS-120B evaluated O3-Mini and adapted or exceeded O4-Mini in competition coding, general problem solving and tools. However, 120B was also able to exceed O4-Mini in queries related to health and mathematics of competition. GPT-OSS-20B was able to perform exactly the same comparative behavior against O3-Mini.
Two modern OPENAI models are now available under the Open-Source Apache 2.0 license. Opeli cooperates with lots of companies to support its latest models on various platforms, including onnx Runtime, Azure, AWS and Ollam.
