ULTIMATE AGENTGPT-OSS 120B
Text, streaming, supported reasoning controls and safe native tools. Shared limits apply.
MODEL CATALOG
Choose the model.
Know its boundaries.
Llama 3.2 3B
A small local text model. CPU latency and answer quality vary.
Qwen3 1.7B / Qwen3.5 9B
Local model choices. Product-level search does not require model-native tool calling. Screenshot vision is not advertised as verified.
Ministral 3 14B
Local inference can be slow and memory intensive. Use bounded requests.
Prepared providers are not available models.
Modal GPU, Gemini and paid OpenAI models are shown only after their credentials, authorization, budget protections and real test requests pass. Modal hosts Qwen; Qwen remains the model creator.
Compare authorized models ↗