MODEL PROFILE

gpt-oss-20b (batch)

OpenAI · gpt-oss-20b:batch

gpt-oss-20b是OpenAI在Apache 2.0许可下发布的开放权重的21B参数模型。它使用混合专家(MoE)架构,每次转发传递3.6B活动参数,针对低延迟推理和可部署性在消费者或单gpu硬件上进行了优化。该模型以OpenAI的Harmony响应格式进行训练,并支持推理级别配置、微调和代理功能,包括函数调用、工具使用和结构化输出。

Translation queued. Showing the source text.

Text generationreasoningstructured_outputstools
Context window131,072 Token
Input → OutputText → Text
Model weightsOpen weights
Profile updated2026/11/4 08:46:58
OVERVIEW

Model overview

gpt-oss-20b是OpenAI在Apache 2.0许可下发布的开放权重的21B参数模型。它使用混合专家(MoE)架构,每次转发传递3.6B活动参数,针对低延迟推理和可部署性在消费者或单gpu硬件上进行了优化。该模型以OpenAI的Harmony响应格式进行训练,并支持推理级别配置、微调和代理功能,包括函数调用、工具使用和结构化输出。

The summary below is compiled from sources by an AI model, not our own testing. Verify selection decisions against original sources and your workload.

Key features

  • 采用混合专家(MoE)架构,每次前向传播仅激活约3.6B参数,兼顾性能与效率
  • 专为低延迟推理和单GPU/消费级硬件部署优化,支持本地可运行性
  • 原生支持Harmony响应格式、函数调用、工具使用及结构化输出等代理能力

Limitations & considerations

Sources are insufficient to establish specific limitations. Verify task performance, latency, data policies, licensing and version lifecycle.

PRICING & COST

Source prices & cost analysis

Prices vary by channel, region and configuration. Channel quotes are listed independently and are not treated as a single official provider price.

Source / channelInputOutputBilling unitCollected at
OpenRouter · openai/gpt-oss-20b ↗USD 0.03USD 0.13Per million tokens2026/11/2 06:25:30
OpenRouter · openai/gpt-oss-20b:batch ↗USD 0.024USD 0.112Per million tokens2026/11/2 06:25:30
Apilio · default ↗Not verifiedNot verified渠道倍率待核验2026/11/2 06:25:15
Apilio · openai官-优质 ↗Not verifiedNot verified渠道倍率待核验2026/11/2 06:25:15
Apilio · origin ↗Not verifiedNot verified渠道倍率待核验2026/11/2 06:25:15
Apilio · ssvip ↗Not verifiedNot verified渠道倍率待核验2026/11/2 06:25:15

Some channels offer cache discounts. Check the source for cache read, write and additional fees.

Estimate monthly base token costs

—

Estimates exclude cache discounts, tiered pricing, multimodal surcharges, web search, taxes and additional reasoning output.

INDEPENDENT BENCHMARKS

Public benchmarks & run configuration

暂无已采集的可核验公开评分。模型未上榜不代表能力较低。

View full indexed rankings →

Frequently asked questions

该模型是否开源且允许商用?

是的,模型以Apache 2.0许可证发布,允许自由使用、修改和商用。

它能否在单块消费级GPU上运行?

是的,设计目标明确包含在消费者或单GPU硬件上的可部署性与低延迟推理。

是否支持函数调用和结构化输出?

是的,官方描述明确指出支持推理级配置、函数调用、工具使用和结构化输出。

Do open weights permit free commercial use?

Not necessarily. Check the version license, commercial conditions and usage policies. Self-hosting incurs compute and maintenance costs.

Are these real-time billing prices?

These are timestamped source snapshots, not live billing prices. Check the channel for current versions, regions, discounts and billing methods.

Sources & traceability

apilio · gpt-oss-20b ↗2026/11/2 06:25:15

许可证:apache-2.0。模型介绍与第三方评测均以原来源的具体版本和方法为准。