下载或取得权重
Qwen3-30B-A3B
Qwen describes this model as open-weighted
获取入口：Qwen3: Think Deeper, Act Faster ↗选择要核对的信息，直接进入对应章节；各项事实均保留版本、条件和来源。
5/14 项有公开值 · 当前版本 Qwen3-30B-A3B · 数据截止 2026-09-17
页面只展示已审核 Claim；待审核数据不会被当作公开事实。只展示已审核事实；未公开、冲突和过期状态不会被补猜。
只判断选型资料是否足够，不评价模型能力高低。
context_window_tokensweights_availabilitymodel_licensetotal_parameterslifecycle_statusrelease_datecapability_tagsdeployment_optionsQwen3-30B-A3B 是 阿里千问 / Qwen 的 Qwen3 模型，年鉴当前记录 1 个版本，最新公开记录为 Qwen3-30B-A3B。
权重为开放权重，代码为未披露，训练数据为partially_disclosed，部署方式为可自行部署；这些维度分别判断，不能合并成单一“开源/闭源”结论。
当前公开快照的上下文窗口为 暂无已审核记录，架构为 未公开；这里只使用通过审核且来源允许公开采集的 Claim。
7 个维度暂无可公开投影的审核记录
补充来源或线索会进入投稿队列，经编辑核验后才会成为公开事实。当前来源记录 2 个。
技术事实、业务价值、企业治理和证据质量分区呈现
厂商、家族、API ID、日期与生命周期
预训练、SFT、偏好优化与安全对齐
教师/学生、基础模型、轨迹和限制
权重、代码、数据、许可证、部署与 API
数据、权限、审计、合规、安全与人工兜底
11 条已审核事实 · 全部来自公开来源闸门通过的记录
尚无已审核公开记录的维度：02 能力与任务、03 架构参数、07 部署与接入、08 成本与 TCO、09 性能与评测、11 业务场景、12 落地与提效。查看资料缺口 ↑
关系只来自年鉴数据库中已登记的归属字段。
下载、API、客户端和更新内容只展示已登记的来源事实；未登记不作推断。
Qwen describes this model as open-weighted
获取入口：Qwen3: Think Deeper, Act Faster ↗Official documentation and model card list Transformers, vLLM, SGLang and local deployment tools.
接入说明：Qwen documentation ↗工具支持、权重和许可证边界分开判断
README 的 Qwen3 模型表与 qwen3 模板说明该系列适配；本档案不替代模型权重、许可证、显存和依赖核验。
保留发布日期精度、生命周期节点和原始来源，未公开日期不会被推断。
模型版本 · active
显示最新核验的 8 条 · 共 11 条已审核事实
Qwen documents 119 languages and dialects for Qwen3. For Qwen3-2507, the maximum assistant message length is 81,920 tokens for thinking models and 16,384 tokens for instruct models; Qwen3-2507 supports 256K context and can be extended to 1M tokens.
Qwen documents 119 languages and dialects for Qwen3. For Qwen3-2507, the maximum assistant message length is 81,920 tokens for thinking models and 16,384 tokens for instruct models; Qwen3-2507 supports 256K context and can be extended to 1M tokens.
1 个版本 · 每项可回到来源与技术解释
高置信度，已绑定来源与证据。
条件：Explicit public claim projection from agent; this is a documented capability tag, not a benchmark or quality conclusion.
来源：Qwen documentation ↗Qwen documents 119 languages and dialects for Qwen3. For Qwen3-2507, the maximum assistant message length is 81,920 tokens for thinking models and 16,384 tokens for instruct models; Qwen3-2507 supports 256K context and can be extended to 1M tokens.
高置信度，已绑定来源与证据。
来源：Qwen documentation ↗Qwen documents 119 languages and dialects for Qwen3. For Qwen3-2507, the maximum assistant message length is 81,920 tokens for thinking models and 16,384 tokens for instruct models; Qwen3-2507 supports 256K context and can be extended to 1M tokens.
引用文本与接口返回都只包含已审核公开事实
阿里千问 / Qwen Qwen3-30B-A3B｜最新版本 Qwen3-30B-A3B｜已审核事实 11 条｜公开来源 2 个｜数据截止 2026-09-17｜AI 年鉴｜https://ainianjian.com/models/model.alibaba.qwen.qwen3-30b-a3b
公开接口返回该模型的版本、已审核事实 11 条及其来源信息，可直接用于脚本或对比。
https://api.ainianjian.com/api/models/model.alibaba.qwen.qwen3-30b-a3b ↗curl -s 'https://api.ainianjian.com/api/models/model.alibaba.qwen.qwen3-30b-a3b'
技术规格使用当前版本已审核且来源允许公开采集的事实；模型级开放信息与跨版本地区记录分别标注范围。请按每项适用范围核对，未登记字段显示「未公开」，不使用估算值填充。
Qwen states that Qwen3 was released on April 29, 2025, open-weights two MoE models and six dense models under Apache 2.0. The model tables list layers, attention heads, experts, and 128K context for the release family.
Qwen states that Qwen3 uses approximately 36 trillion pre-training tokens across 119 languages and dialects. The pre-training process has three stages; post-training has four stages: long chain-of-thought cold start, reasoning-based reinforcement learning, thinking mode fusion, and general reinforcement learning.
Qwen states that Qwen3 uses approximately 36 trillion pre-training tokens across 119 languages and dialects. The pre-training process has three stages; post-training has four stages: long chain-of-thought cold start, reasoning-based reinforcement learning, thinking mode fusion, and general reinforcement learning.
Qwen states that Qwen3 was released on April 29, 2025, open-weights two MoE models and six dense models under Apache 2.0. The model tables list layers, attention heads, experts, and 128K context for the release family.
Qwen states that Qwen3 was released on April 29, 2025, open-weights two MoE models and six dense models under Apache 2.0. The model tables list layers, attention heads, experts, and 128K context for the release family.
Qwen documents 119 languages and dialects for Qwen3. For Qwen3-2507, the maximum assistant message length is 81,920 tokens for thinking models and 16,384 tokens for instruct models; Qwen3-2507 supports 256K context and can be extended to 1M tokens.
Qwen states that Qwen3 was released on April 29, 2025, open-weights two MoE models and six dense models under Apache 2.0. The model tables list layers, attention heads, experts, and 128K context for the release family.
高置信度，已绑定来源与证据。
来源：Qwen3: Think Deeper, Act Faster ↗Qwen states that Qwen3 uses approximately 36 trillion pre-training tokens across 119 languages and dialects. The pre-training process has three stages; post-training has four stages: long chain-of-thought cold start, reasoning-based reinforcement learning, thinking mode fusion, and general reinforcement learning.
高置信度，已绑定来源与证据。
来源：Qwen3: Think Deeper, Act Faster ↗Qwen states that Qwen3 uses approximately 36 trillion pre-training tokens across 119 languages and dialects. The pre-training process has three stages; post-training has four stages: long chain-of-thought cold start, reasoning-based reinforcement learning, thinking mode fusion, and general reinforcement learning.
高置信度，已绑定来源与证据。
来源：Qwen3: Think Deeper, Act Faster ↗Qwen states that Qwen3 was released on April 29, 2025, open-weights two MoE models and six dense models under Apache 2.0. The model tables list layers, attention heads, experts, and 128K context for the release family.
高置信度，已绑定来源与证据。
来源：Qwen3: Think Deeper, Act Faster ↗Qwen states that Qwen3 was released on April 29, 2025, open-weights two MoE models and six dense models under Apache 2.0. The model tables list layers, attention heads, experts, and 128K context for the release family.
高置信度，已绑定来源与证据。
来源：Qwen documentation ↗Qwen documents 119 languages and dialects for Qwen3. For Qwen3-2507, the maximum assistant message length is 81,920 tokens for thinking models and 16,384 tokens for instruct models; Qwen3-2507 supports 256K context and can be extended to 1M tokens.
高置信度，已绑定来源与证据。
来源：Qwen3: Think Deeper, Act Faster ↗Qwen states that Qwen3 uses approximately 36 trillion pre-training tokens across 119 languages and dialects. The pre-training process has three stages; post-training has four stages: long chain-of-thought cold start, reasoning-based reinforcement learning, thinking mode fusion, and general reinforcement learning.
高置信度，已绑定来源与证据。
来源：Qwen3: Think Deeper, Act Faster ↗Qwen states that Qwen3 uses approximately 36 trillion pre-training tokens across 119 languages and dialects. The pre-training process has three stages; post-training has four stages: long chain-of-thought cold start, reasoning-based reinforcement learning, thinking mode fusion, and general reinforcement learning.
高置信度，已绑定来源与证据。
来源：Qwen3: Think Deeper, Act Faster ↗Qwen states that Qwen3 was released on April 29, 2025, open-weights two MoE models and six dense models under Apache 2.0. The model tables list layers, attention heads, experts, and 128K context for the release family.