下载或取得权重
Mistral Small 4
open_weights
获取入口：Mistral Small 4 119B A6B (official model card) ↗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 Mistral Small 4 · 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_parametersMistral Small 是 Mistral AI 的 Mistral Small 模型，年鉴当前记录 2 个版本，最新公开记录为 Mistral Small 4。
权重为开放权重，代码为未披露，训练数据为未披露，部署方式为API + 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 256K 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。
当前最新版本的 approved Claim 投影 · 17 条历史事实正在复核
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：Mistral Small 4 119B A6B (official model card) ↗self_hosted 119B MoE weights; vLLM tag; NVFP4 quantised checkpoint; 4x NVIDIA HGX H100 minimum
接入说明：Mistral Small 4 119B A6B (official model card) ↗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 · deprecated
模型版本 · active
Mistral announced Small 4 as a unified model combining instruct, reasoning, multimodal and agentic coding capabilities.
来源：Introducing Mistral Small 4 ↗显示最新核验的 8 条 · 共 8 条已审核事实
Mistral Small 4 is a hybrid model that unifies Instruct, Reasoning and Devstral capabilities into one multimodal model. Key features: MoE with 128 experts and 4 active; 119B parameters with 6.5B activated per token; 256k context length; multimodal input accepting text and images with text output; instruct and reasoning functionality with function calls and configurable reasoning effort; vision, multilingual and agentic capabilities with native function calling and JSON output; Apache 2.0 license for commercial and non-commercial use. The card is tagged vLLM and links NVFP4 and eagle speculative-decoding checkpoints. Front matter: license: apache-2.0. The Hugging Face model API reports the repository as public and not gated.
Mistral Small 4 is a hybrid model that unifies Instruct, Reasoning and Devstral capabilities into one multimodal model. Key features: MoE with 128 experts and 4 active; 119B parameters with 6.5B activated per token; 256k context length; multimodal input accepting text and images with text output; instruct and reasoning functionality with function calls and configurable reasoning effort; vision, multilingual and agentic capabilities with native function calling and JSON output; Apache 2.0 license for commercial and non-commercial use. The card is tagged vLLM and links NVFP4 and eagle speculative-decoding checkpoints. Front matter: license: apache-2.0. The Hugging Face model API reports the repository as public and not gated.
2 个版本 · 每项可回到来源与技术解释
查看最近版本差异待审核内容不会直接进入公开事实。
高置信度，已绑定来源与证据。
条件：Set on 2026-10-20 from the model card: agentic capabilities with native function calling and JSON output, image input support, and a togglable reasoning mode.
来源：Mistral Small 4 119B A6B (official model card) ↗Mistral Small 4 is a hybrid model that unifies Instruct, Reasoning and Devstral capabilities into one multimodal model. Key features: MoE with 128 experts and 4 active; 119B parameters with 6.5B activated per token; 256k context length; multimodal input accepting text and images with text output; instruct and reasoning functionality with function calls and configurable reasoning effort; vision, multilingual and agentic capabilities with native function calling and JSON output; Apache 2.0 license for commercial and non-commercial use. The card is tagged vLLM and links NVFP4 and eagle speculative-decoding checkpoints. Front matter: license: apache-2.0. The Hugging Face model API reports the repository as public and not gated.
引用文本与接口返回都只包含已审核公开事实
Mistral AI Mistral Small｜最新版本 Mistral Small 4｜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.mistral.small
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.mistral.small ↗curl -s 'https://api.ainianjian.com/api/models/model.mistral.small'
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。
Mistral Small 4 is a hybrid model that unifies Instruct, Reasoning and Devstral capabilities into one multimodal model. Key features: MoE with 128 experts and 4 active; 119B parameters with 6.5B activated per token; 256k context length; multimodal input accepting text and images with text output; instruct and reasoning functionality with function calls and configurable reasoning effort; vision, multilingual and agentic capabilities with native function calling and JSON output; Apache 2.0 license for commercial and non-commercial use. The card is tagged vLLM and links NVFP4 and eagle speculative-decoding checkpoints. Front matter: license: apache-2.0. The Hugging Face model API reports the repository as public and not gated.
Mistral Small 4 is a hybrid model that unifies Instruct, Reasoning and Devstral capabilities into one multimodal model. Key features: MoE with 128 experts and 4 active; 119B parameters with 6.5B activated per token; 256k context length; multimodal input accepting text and images with text output; instruct and reasoning functionality with function calls and configurable reasoning effort; vision, multilingual and agentic capabilities with native function calling and JSON output; Apache 2.0 license for commercial and non-commercial use. The card is tagged vLLM and links NVFP4 and eagle speculative-decoding checkpoints. Front matter: license: apache-2.0. The Hugging Face model API reports the repository as public and not gated.
Mistral Small 4 is a hybrid model that unifies Instruct, Reasoning and Devstral capabilities into one multimodal model. Key features: MoE with 128 experts and 4 active; 119B parameters with 6.5B activated per token; 256k context length; multimodal input accepting text and images with text output; instruct and reasoning functionality with function calls and configurable reasoning effort; vision, multilingual and agentic capabilities with native function calling and JSON output; Apache 2.0 license for commercial and non-commercial use. The card is tagged vLLM and links NVFP4 and eagle speculative-decoding checkpoints. Front matter: license: apache-2.0. The Hugging Face model API reports the repository as public and not gated.
Mistral Small 4 is a hybrid model that unifies Instruct, Reasoning and Devstral capabilities into one multimodal model. Key features: MoE with 128 experts and 4 active; 119B parameters with 6.5B activated per token; 256k context length; multimodal input accepting text and images with text output; instruct and reasoning functionality with function calls and configurable reasoning effort; vision, multilingual and agentic capabilities with native function calling and JSON output; Apache 2.0 license for commercial and non-commercial use. The card is tagged vLLM and links NVFP4 and eagle speculative-decoding checkpoints. Front matter: license: apache-2.0. The Hugging Face model API reports the repository as public and not gated.
Mistral Small 4 is a hybrid model that unifies Instruct, Reasoning and Devstral capabilities into one multimodal model. Key features: MoE with 128 experts and 4 active; 119B parameters with 6.5B activated per token; 256k context length; multimodal input accepting text and images with text output; instruct and reasoning functionality with function calls and configurable reasoning effort; vision, multilingual and agentic capabilities with native function calling and JSON output; Apache 2.0 license for commercial and non-commercial use. The card is tagged vLLM and links NVFP4 and eagle speculative-decoding checkpoints. Front matter: license: apache-2.0. The Hugging Face model API reports the repository as public and not gated.
Mistral Small 4 is a hybrid model that unifies Instruct, Reasoning and Devstral capabilities into one multimodal model. Key features: MoE with 128 experts and 4 active; 119B parameters with 6.5B activated per token; 256k context length; multimodal input accepting text and images with text output; instruct and reasoning functionality with function calls and configurable reasoning effort; vision, multilingual and agentic capabilities with native function calling and JSON output; Apache 2.0 license for commercial and non-commercial use. The card is tagged vLLM and links NVFP4 and eagle speculative-decoding checkpoints. Front matter: license: apache-2.0. The Hugging Face model API reports the repository as public and not gated.
高置信度，已绑定来源与证据。
条件：Re-verified on 2026-10-20 from the model card: 256k context length.
来源：Mistral Small 4 119B A6B (official model card) ↗Mistral Small 4 is a hybrid model that unifies Instruct, Reasoning and Devstral capabilities into one multimodal model. Key features: MoE with 128 experts and 4 active; 119B parameters with 6.5B activated per token; 256k context length; multimodal input accepting text and images with text output; instruct and reasoning functionality with function calls and configurable reasoning effort; vision, multilingual and agentic capabilities with native function calling and JSON output; Apache 2.0 license for commercial and non-commercial use. The card is tagged vLLM and links NVFP4 and eagle speculative-decoding checkpoints. Front matter: license: apache-2.0. The Hugging Face model API reports the repository as public and not gated.
高置信度，已绑定来源与证据。
条件：Re-verified on 2026-10-20; the card is tagged vLLM and provides NVFP4 and eagle-head checkpoints. The minimum 4x H100 footprint was recorded earlier from the release announcement and is not restated on the model card.
来源：Mistral Small 4 119B A6B (official model card) ↗Mistral Small 4 is a hybrid model that unifies Instruct, Reasoning and Devstral capabilities into one multimodal model. Key features: MoE with 128 experts and 4 active; 119B parameters with 6.5B activated per token; 256k context length; multimodal input accepting text and images with text output; instruct and reasoning functionality with function calls and configurable reasoning effort; vision, multilingual and agentic capabilities with native function calling and JSON output; Apache 2.0 license for commercial and non-commercial use. The card is tagged vLLM and links NVFP4 and eagle speculative-decoding checkpoints. Front matter: license: apache-2.0. The Hugging Face model API reports the repository as public and not gated.
高置信度，已绑定来源与证据。
条件：Set on 2026-10-20 because the mistralai Hugging Face repository for Small 4 remains public, not gated and not disabled; Mistral does not publish a formal lifecycle table for these weights, so this reflects current availability rather than a vendor-stated status.
来源：Mistral Small 4 119B A6B (official model card) ↗Mistral Small 4 is a hybrid model that unifies Instruct, Reasoning and Devstral capabilities into one multimodal model. Key features: MoE with 128 experts and 4 active; 119B parameters with 6.5B activated per token; 256k context length; multimodal input accepting text and images with text output; instruct and reasoning functionality with function calls and configurable reasoning effort; vision, multilingual and agentic capabilities with native function calling and JSON output; Apache 2.0 license for commercial and non-commercial use. The card is tagged vLLM and links NVFP4 and eagle speculative-decoding checkpoints. Front matter: license: apache-2.0. The Hugging Face model API reports the repository as public and not gated.
高置信度，已绑定来源与证据。
条件：Re-verified on 2026-10-20; the card states Apache 2.0 for commercial and non-commercial use and front matter declares apache-2.0.
来源：Mistral Small 4 119B A6B (official model card) ↗Mistral Small 4 is a hybrid model that unifies Instruct, Reasoning and Devstral capabilities into one multimodal model. Key features: MoE with 128 experts and 4 active; 119B parameters with 6.5B activated per token; 256k context length; multimodal input accepting text and images with text output; instruct and reasoning functionality with function calls and configurable reasoning effort; vision, multilingual and agentic capabilities with native function calling and JSON output; Apache 2.0 license for commercial and non-commercial use. The card is tagged vLLM and links NVFP4 and eagle speculative-decoding checkpoints. Front matter: license: apache-2.0. The Hugging Face model API reports the repository as public and not gated.
高置信度，已绑定来源与证据。
条件：Set on 2026-10-20 from the official repository name suffix 2603; the model card does not state a day, so only month precision is asserted here.
来源：Mistral Small 4 119B A6B (official model card) ↗Mistral Small 4 is a hybrid model that unifies Instruct, Reasoning and Devstral capabilities into one multimodal model. Key features: MoE with 128 experts and 4 active; 119B parameters with 6.5B activated per token; 256k context length; multimodal input accepting text and images with text output; instruct and reasoning functionality with function calls and configurable reasoning effort; vision, multilingual and agentic capabilities with native function calling and JSON output; Apache 2.0 license for commercial and non-commercial use. The card is tagged vLLM and links NVFP4 and eagle speculative-decoding checkpoints. Front matter: license: apache-2.0. The Hugging Face model API reports the repository as public and not gated.
高置信度，已绑定来源与证据。
条件：Re-verified on 2026-10-20 from the model card: 119B parameters with 6.5B activated per token.
来源：Mistral Small 4 119B A6B (official model card) ↗Mistral Small 4 is a hybrid model that unifies Instruct, Reasoning and Devstral capabilities into one multimodal model. Key features: MoE with 128 experts and 4 active; 119B parameters with 6.5B activated per token; 256k context length; multimodal input accepting text and images with text output; instruct and reasoning functionality with function calls and configurable reasoning effort; vision, multilingual and agentic capabilities with native function calling and JSON output; Apache 2.0 license for commercial and non-commercial use. The card is tagged vLLM and links NVFP4 and eagle speculative-decoding checkpoints. Front matter: license: apache-2.0. The Hugging Face model API reports the repository as public and not gated.
高置信度，已绑定来源与证据。
条件：Re-verified on 2026-10-20 against the official Mistral AI Hugging Face repository, which publishes the Small 4 weights, NVFP4 quantisation and an eagle head and is not gated.
来源：Mistral Small 4 119B A6B (official model card) ↗Mistral Small 4 is a hybrid model that unifies Instruct, Reasoning and Devstral capabilities into one multimodal model. Key features: MoE with 128 experts and 4 active; 119B parameters with 6.5B activated per token; 256k context length; multimodal input accepting text and images with text output; instruct and reasoning functionality with function calls and configurable reasoning effort; vision, multilingual and agentic capabilities with native function calling and JSON output; Apache 2.0 license for commercial and non-commercial use. The card is tagged vLLM and links NVFP4 and eagle speculative-decoding checkpoints. Front matter: license: apache-2.0. The Hugging Face model API reports the repository as public and not gated.