下载或取得权重
Llama 4 Maverick
open_weights
获取入口：Llama 4 model card (meta-llama/llama-models) ↗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。
7/14 项有公开值 · 当前版本 Llama 4 Maverick · 数据截止 2026-10-20
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。
所有必填事实均通过状态、来源、证据与时效校验。最早核验时间：2026-10-20。
release_datelifecycle_statuscapability_tagsdeployment_optionscontext_window_tokensweights_availabilitymodel_licensetotal_parametersLlama 4 Maverick 是 Meta AI 的 Llama 4 模型，年鉴当前记录 1 个版本，最新公开记录为 Llama 4 Maverick。
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 1M tokens，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 1 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
Dense/MoE、参数、层数、专家与 Tokenizer
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
输入输出、缓存、Batch 与部署成本边界
数据、权限、审计、合规、安全与人工兜底
8 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、07 部署与接入、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。
open_weights
获取入口：Llama 4 model card (meta-llama/llama-models) ↗self_hosted
接入说明：Llama 4 model card (meta-llama/llama-models) ↗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 8 条已审核事实
Model table: Llama 4 Maverick (17Bx128E), 17B activated and 400B total parameters, input modalities multilingual text and image, output modalities multilingual text and code, context length 1M, token count ~22T, knowledge cutoff August 2024. Supported languages include Arabic, English, French, German, Hindi, Indonesian, Italian, Portuguese, Spanish, Tagalog, Thai and Vietnamese. Model Release Date: April 5, 2025. License: a custom commercial license, the Llama 4 Community License Agreement. The long-context benchmark table notes a 128K context window for that specific evaluation. The card states quantized checkpoints are provided for deployment flexibility.
Model table: Llama 4 Maverick (17Bx128E), 17B activated and 400B total parameters, input modalities multilingual text and image, output modalities multilingual text and code, context length 1M, token count ~22T, knowledge cutoff August 2024. Supported languages include Arabic, English, French, German, Hindi, Indonesian, Italian, Portuguese, Spanish, Tagalog, Thai and Vietnamese. Model Release Date: April 5, 2025. License: a custom commercial license, the Llama 4 Community License Agreement. The long-context benchmark table notes a 128K context window for that specific evaluation. The card states quantized checkpoints are provided for deployment flexibility.
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：Set on 2026-10-20 from the card: multimodal input of text and images, with instruction-tuned models intended for assistant-like chat and visual reasoning tasks.
来源：Llama 4 model card (meta-llama/llama-models) ↗Model table: Llama 4 Maverick (17Bx128E), 17B activated and 400B total parameters, input modalities multilingual text and image, output modalities multilingual text and code, context length 1M, token count ~22T, knowledge cutoff August 2024. Supported languages include Arabic, English, French, German, Hindi, Indonesian, Italian, Portuguese, Spanish, Tagalog, Thai and Vietnamese. Model Release Date: April 5, 2025. License: a custom commercial license, the Llama 4 Community License Agreement. The long-context benchmark table notes a 128K context window for that specific evaluation. The card states quantized checkpoints are provided for deployment flexibility.
高置信度，已绑定来源与证据。
条件：Re-verified on 2026-10-20 against the model card table, which lists a 1M context length for Maverick; the 128K figure in the long-context benchmark table describes that evaluation setting only.
来源：Llama 4 model card (meta-llama/llama-models) ↗引用文本与接口返回都只包含已审核公开事实
Meta AI Llama 4 Maverick｜最新版本 Llama 4 Maverick｜已审核事实 8 条｜公开来源 1 个｜数据截止 2026-10-20｜AI 年鉴｜https://ainianjian.com/models/model.meta.llama-4-maverick
公开接口返回该模型的版本、已审核事实 8 条及其来源信息，可直接用于脚本或对比。
https://api.ainianjian.com/api/models/model.meta.llama-4-maverick ↗curl -s 'https://api.ainianjian.com/api/models/model.meta.llama-4-maverick'
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。
Model table: Llama 4 Maverick (17Bx128E), 17B activated and 400B total parameters, input modalities multilingual text and image, output modalities multilingual text and code, context length 1M, token count ~22T, knowledge cutoff August 2024. Supported languages include Arabic, English, French, German, Hindi, Indonesian, Italian, Portuguese, Spanish, Tagalog, Thai and Vietnamese. Model Release Date: April 5, 2025. License: a custom commercial license, the Llama 4 Community License Agreement. The long-context benchmark table notes a 128K context window for that specific evaluation. The card states quantized checkpoints are provided for deployment flexibility.
Model table: Llama 4 Maverick (17Bx128E), 17B activated and 400B total parameters, input modalities multilingual text and image, output modalities multilingual text and code, context length 1M, token count ~22T, knowledge cutoff August 2024. Supported languages include Arabic, English, French, German, Hindi, Indonesian, Italian, Portuguese, Spanish, Tagalog, Thai and Vietnamese. Model Release Date: April 5, 2025. License: a custom commercial license, the Llama 4 Community License Agreement. The long-context benchmark table notes a 128K context window for that specific evaluation. The card states quantized checkpoints are provided for deployment flexibility.
Model table: Llama 4 Maverick (17Bx128E), 17B activated and 400B total parameters, input modalities multilingual text and image, output modalities multilingual text and code, context length 1M, token count ~22T, knowledge cutoff August 2024. Supported languages include Arabic, English, French, German, Hindi, Indonesian, Italian, Portuguese, Spanish, Tagalog, Thai and Vietnamese. Model Release Date: April 5, 2025. License: a custom commercial license, the Llama 4 Community License Agreement. The long-context benchmark table notes a 128K context window for that specific evaluation. The card states quantized checkpoints are provided for deployment flexibility.
Model table: Llama 4 Maverick (17Bx128E), 17B activated and 400B total parameters, input modalities multilingual text and image, output modalities multilingual text and code, context length 1M, token count ~22T, knowledge cutoff August 2024. Supported languages include Arabic, English, French, German, Hindi, Indonesian, Italian, Portuguese, Spanish, Tagalog, Thai and Vietnamese. Model Release Date: April 5, 2025. License: a custom commercial license, the Llama 4 Community License Agreement. The long-context benchmark table notes a 128K context window for that specific evaluation. The card states quantized checkpoints are provided for deployment flexibility.
Model table: Llama 4 Maverick (17Bx128E), 17B activated and 400B total parameters, input modalities multilingual text and image, output modalities multilingual text and code, context length 1M, token count ~22T, knowledge cutoff August 2024. Supported languages include Arabic, English, French, German, Hindi, Indonesian, Italian, Portuguese, Spanish, Tagalog, Thai and Vietnamese. Model Release Date: April 5, 2025. License: a custom commercial license, the Llama 4 Community License Agreement. The long-context benchmark table notes a 128K context window for that specific evaluation. The card states quantized checkpoints are provided for deployment flexibility.
Model table: Llama 4 Maverick (17Bx128E), 17B activated and 400B total parameters, input modalities multilingual text and image, output modalities multilingual text and code, context length 1M, token count ~22T, knowledge cutoff August 2024. Supported languages include Arabic, English, French, German, Hindi, Indonesian, Italian, Portuguese, Spanish, Tagalog, Thai and Vietnamese. Model Release Date: April 5, 2025. License: a custom commercial license, the Llama 4 Community License Agreement. The long-context benchmark table notes a 128K context window for that specific evaluation. The card states quantized checkpoints are provided for deployment flexibility.
Model table: Llama 4 Maverick (17Bx128E), 17B activated and 400B total parameters, input modalities multilingual text and image, output modalities multilingual text and code, context length 1M, token count ~22T, knowledge cutoff August 2024. Supported languages include Arabic, English, French, German, Hindi, Indonesian, Italian, Portuguese, Spanish, Tagalog, Thai and Vietnamese. Model Release Date: April 5, 2025. License: a custom commercial license, the Llama 4 Community License Agreement. The long-context benchmark table notes a 128K context window for that specific evaluation. The card states quantized checkpoints are provided for deployment flexibility.
高置信度，已绑定来源与证据。
条件：Set on 2026-10-20 from the card: released weights with quantized checkpoints provided for deployment flexibility; no managed first-party API price is recorded here.
来源：Llama 4 model card (meta-llama/llama-models) ↗Model table: Llama 4 Maverick (17Bx128E), 17B activated and 400B total parameters, input modalities multilingual text and image, output modalities multilingual text and code, context length 1M, token count ~22T, knowledge cutoff August 2024. Supported languages include Arabic, English, French, German, Hindi, Indonesian, Italian, Portuguese, Spanish, Tagalog, Thai and Vietnamese. Model Release Date: April 5, 2025. License: a custom commercial license, the Llama 4 Community License Agreement. The long-context benchmark table notes a 128K context window for that specific evaluation. The card states quantized checkpoints are provided for deployment flexibility.
高置信度，已绑定来源与证据。
条件：Set on 2026-10-20 because the official Meta model card remains published in the meta-llama repository; Meta does not publish a formal lifecycle table for this release, so this reflects current availability rather than a vendor-stated status.
来源：Llama 4 model card (meta-llama/llama-models) ↗Model table: Llama 4 Maverick (17Bx128E), 17B activated and 400B total parameters, input modalities multilingual text and image, output modalities multilingual text and code, context length 1M, token count ~22T, knowledge cutoff August 2024. Supported languages include Arabic, English, French, German, Hindi, Indonesian, Italian, Portuguese, Spanish, Tagalog, Thai and Vietnamese. Model Release Date: April 5, 2025. License: a custom commercial license, the Llama 4 Community License Agreement. The long-context benchmark table notes a 128K context window for that specific evaluation. The card states quantized checkpoints are provided for deployment flexibility.
高置信度，已绑定来源与证据。
条件：Re-verified on 2026-10-20; the card states a custom commercial license, the Llama 4 Community License Agreement, linked in the repository.
来源：Llama 4 model card (meta-llama/llama-models) ↗Model table: Llama 4 Maverick (17Bx128E), 17B activated and 400B total parameters, input modalities multilingual text and image, output modalities multilingual text and code, context length 1M, token count ~22T, knowledge cutoff August 2024. Supported languages include Arabic, English, French, German, Hindi, Indonesian, Italian, Portuguese, Spanish, Tagalog, Thai and Vietnamese. Model Release Date: April 5, 2025. License: a custom commercial license, the Llama 4 Community License Agreement. The long-context benchmark table notes a 128K context window for that specific evaluation. The card states quantized checkpoints are provided for deployment flexibility.
高置信度，已绑定来源与证据。
条件：Re-verified on 2026-10-20; the card states Model Release Date: April 5, 2025.
来源：Llama 4 model card (meta-llama/llama-models) ↗Model table: Llama 4 Maverick (17Bx128E), 17B activated and 400B total parameters, input modalities multilingual text and image, output modalities multilingual text and code, context length 1M, token count ~22T, knowledge cutoff August 2024. Supported languages include Arabic, English, French, German, Hindi, Indonesian, Italian, Portuguese, Spanish, Tagalog, Thai and Vietnamese. Model Release Date: April 5, 2025. License: a custom commercial license, the Llama 4 Community License Agreement. The long-context benchmark table notes a 128K context window for that specific evaluation. The card states quantized checkpoints are provided for deployment flexibility.
高置信度，已绑定来源与证据。
条件：Re-verified on 2026-10-20; the model table lists 400B total parameters with 17B activated.
来源：Llama 4 model card (meta-llama/llama-models) ↗Model table: Llama 4 Maverick (17Bx128E), 17B activated and 400B total parameters, input modalities multilingual text and image, output modalities multilingual text and code, context length 1M, token count ~22T, knowledge cutoff August 2024. Supported languages include Arabic, English, French, German, Hindi, Indonesian, Italian, Portuguese, Spanish, Tagalog, Thai and Vietnamese. Model Release Date: April 5, 2025. License: a custom commercial license, the Llama 4 Community License Agreement. The long-context benchmark table notes a 128K context window for that specific evaluation. The card states quantized checkpoints are provided for deployment flexibility.
高置信度，已绑定来源与证据。
条件：Re-verified on 2026-10-20; the card documents released Llama 4 Maverick weights with quantized checkpoints and a community license.
来源：Llama 4 model card (meta-llama/llama-models) ↗Model table: Llama 4 Maverick (17Bx128E), 17B activated and 400B total parameters, input modalities multilingual text and image, output modalities multilingual text and code, context length 1M, token count ~22T, knowledge cutoff August 2024. Supported languages include Arabic, English, French, German, Hindi, Indonesian, Italian, Portuguese, Spanish, Tagalog, Thai and Vietnamese. Model Release Date: April 5, 2025. License: a custom commercial license, the Llama 4 Community License Agreement. The long-context benchmark table notes a 128K context window for that specific evaluation. The card states quantized checkpoints are provided for deployment flexibility.