GPT-OSS

With GPT-OSS, OpenAI has released open-weight models designed for real-world deployment: local, private cloud, on-prem, and air-gapped environments.

In this meetup we look into GPT-OSS: what it is, how it compares to other open models, and how engineers can actually deploy and use it in production systems.

### What we’ll cover

• What GPT-OSS is (and what it is not)
• How GPT-OSS compares to closed APIs and other open models (LLaMA, Mistral, Qwen)
• Deployment architectures: local inference, VPC, on-prem GPU clusters
• Tool calling, structured output, and agent-friendly behavior
• Fine-tuning options (full FT, LoRA / QLoRA)
• Performance, cost, and operational tradeoffs
• When open models make sense — and when APIs are still the better choice