Using the new free GPT-OSS open-weight models (from OpenAI) offline via Ollama?

[Hedda]

Just read the news and wondering if anyone tested the gpt-oss-20b and/or gpt-oss-120b GPT-OSS LLM models (from OpenAI) via Ollama?

• https://openai.com/index/introducing-gpt-oss/

  • https://openai.com/index/gpt-oss-model-card/

    • https://github.com/openai/gpt-oss

FYI, I understand that gpt-oss-20b model require 16GB RAM (GPU/APU) while gpt-oss-120b model require 80GB RAM (GPU/APU).

• gpt-oss-20b — for lower latency, and local or specialized use cases (21B parameters with 3.6B active parameters)

• gpt-oss-120b — for production, general purpose, high reasoning use cases that fit into a single H100 GPU (117B parameters with 5.1B active parameters)

PS: These new LLMs are apparently a game changer for off-line reasoning AI tasks. Check out this quick explaination/review by MattWolfe:

• https://www.youtube.com/watch?v=LEd_b2vTbAM&ab_channel=MattWolfe

[mainframecn]

I had the gpt-oss-20gb "working" with local ollama, but the ai chat was just spitting output where formatting was unusable. Haven't tried a payed online service