OpenAI raises gpt-oss-20b API prices on OpenRouter
Esta notícia ainda só está disponível em inglês.
OpenAI increased the API pricing for its open-weight model gpt-oss-20b on OpenRouter. The input price rose from $0.018 to $0.03 per million tokens, while the output price increased from $0.09 to $0.13 per million tokens. This represents a 66.7% increase for input tokens and a 44.4% increase for output tokens compared to the previous day's rates.
Model Specifications and Licensing
According to OpenAI, gpt-oss-20b is one of two open-weight models released by the company, alongside gpt-oss-120b. The model features 21B parameters with 3.6B active parameters and is trained on the harmony response format. OpenAI states that the model should only be used with this specific format.
The model is released under a permissive Apache 2.0 license, making it fine-tunable for specialized use cases. OpenAI notes that both models utilize MXFP4 quantization of MoE weights and provide full chain-of-thought access for debugging and trust. Configurable reasoning effort is available across low, medium, and high levels.
Capabilities and Hardware Requirements
OpenAI lists agentic capabilities for the model, including function calling, web browsing, Python code execution, and Structured Outputs. Regarding hardware requirements, OpenAI states that gpt-oss-20b runs within 16GB of memory.
For the larger companion model, gpt-oss-120b, OpenAI claims it runs on a single 80GB GPU like NVIDIA H100 or AMD MI300X. The gpt-oss-20b model has a context window of 131,072 tokens on OpenRouter.
Pontos principais
- The input price for gpt-oss-20b on OpenRouter increased by 66.7% to $0.03 per million tokens.
- The output price for gpt-oss-20b on OpenRouter increased by 44.4% to $0.13 per million tokens.
- OpenAI states that gpt-oss-20b requires only 16GB of memory to run.
Anúncio oficial
Publicado por openai. Leia na fonte.