公开证据综合
Mac mini 与 Mac Studio 的本地模型适配
面向最新 Qwen3.8、DeepSeek V4、GLM-5.3 和 Gemma 4 开放权重代表的规划工具。先从公开文件事实与容量计算开始，再把运行工具与工作流证据分开判断，最后再决定升级是否合理。
简短结论
用本页按模型家族比较当前高关注代表。旧版本仅保留兼容性记录，不放在主矩阵。模型即使能放进内存，工具调用、JSON、视觉、并发和持续运行仍可能未知。
交互工具
检查你的模型是否适配 Mac。
选择模型或 Mac 配置，查看权重预算、内存余量、运行工具证据，以及带入免费 Advisor 的下一步。
公开证据综合，不是第一方跑分。
该规划工具整合官方模型卡、公开 GGUF/QAT 文件元数据、运行工具文档、已发表评测和清晰标注的社区报告。它不声称 Keep or Upgrade 运行过每个模型，Apple 或模型厂商也未背书此页。
下载内容是包含来源 URL、证据等级和数据日期的可引用快照。请把本 Hub 作为规范解释；下载接口不会进入 sitemap。
Mac mini M6 · 16GB153GB/s · US$899 · 起售价
模型证据矩阵
最新且高关注的开放权重模型
主矩阵优先展示当前旗舰发布和最适合 Mac 规划的代表。容量优先采用公开文件大小；参数数值只是规划估算。展开任意行的“来源与限制”可查看可归因的社区信号及其局限。
| 家族 / 层级 | 模型与版本 | 量化 / 文件 | 所选 Mac 上的容量 | 运行工具证据 | 置信度 | 详情 |
|---|---|---|---|---|---|---|
| Qwen小型 | Qwen3.8 27BQwen3.8 · 27B dense | Q4_0 · GGUF16.1 GB · Public file metadata | 不适配-8.06GB 剩余 | 官方支持mlx2 社区信号 · 查看来源 | high核验 2026-08-28 | 来源与限制
|
| Qwen主流 | Qwen3.8-Flash-Next 180B (6B active)Qwen3.8-Flash-Next · 180B total / 6B active | BF16 · Safetensors360.0 GB · Public file metadata | 不适配-352GB 剩余 | 社区报告mlx1 社区信号 · 查看来源 | medium核验 2026-08-28 | 来源与限制
|
| Qwen边界 | Qwen3.8 2.4T-A95BQwen3.8 · 2400B total / 95B active | Q4_K_M · GGUF1440.0 GB · Public model source; exact file not registered | 不适配-1432GB 剩余 | 官方支持mlx | medium核验 2026-08-28 | 来源与限制
|
| DeepSeek主流 | DeepSeek-V4-Flash 284B-A13BDeepSeek V4 · 284B total / 13B active | MLX-4bit · MLX170.4 GB · Public model source; exact file not registered | 不适配-162.4GB 剩余 | 社区报告mlx7 社区信号 · 查看来源 | high核验 2026-08-28 | 来源与限制
|
| DeepSeek边界 | DeepSeek-V4-Pro 1.6T-A49BDeepSeek V4 · 1600B total / 49B active | MLX-4bit · MLX960.0 GB · Public model source; exact file not registered | 不适配-952GB 剩余 | 未知llama.cpp | high核验 2026-08-28 | 来源与限制
|
| GLM主流 | GLM-5.3-Flash 320B-A18BGLM-5.3-Flash · 320B total / 18B active | MLX-4bit · MLX204.0 GB · Public file metadata | 不适配-195.99GB 剩余 | 社区报告mlx2 社区信号 · 查看来源 | high核验 2026-08-28 | 来源与限制
|
| GLM边界 | GLM-5.3 744B-A40BGLM-5.3 · 744B total / 40B active | MLX-4bit · MLX446.4 GB · Public model source; exact file not registered | 不适配-438.4GB 剩余 | 未知llama.cpp | high核验 2026-08-28 | 来源与限制
|
| Google Gemma小型 | Gemma 4 E2BGemma 4 · 2.3B effective / 5.1B total · PLE | Q4_0 · QAT3.35 GB · Public file metadata | 有余量4.65GB 剩余 | 社区报告llama.cpp1 社区信号 · 查看来源 | high核验 2026-08-28 | 来源与限制
|
| Google Gemma主流 | Gemma 4 12B UnifiedGemma 4 · 11.95B dense | Q4_0 · QAT6.98 GB · Public file metadata | 紧配1.02GB 剩余 | 未知llama.cpp | high核验 2026-08-28 | 来源与限制
|
| Google Gemma边界 | Gemma 4 26B-A4BGemma 4 · 25.2B total / 3.8B active | Q4_0 · QAT14.4 GB · Public file metadata | 不适配-6.44GB 剩余 | 社区报告llama.cpp | high核验 2026-08-28 | 来源与限制
|
Qwen · 小型
不适配Qwen3.8 27B
Qwen3.8 · 27B dense
- 量化 / 文件
- Q4_0 · GGUF16.1 GB · Public file metadata
- 所选 Mac 上的容量
- -8.06GB 剩余
- 运行工具证据
- 官方支持mlx
- 置信度
- high核验 2026-08-28
来源与限制
- 模型官方来源模型 family, parameters, and release context · 核验 2026-08-28
- Public file metadata16.1 GB · exact public file metadata
- 运行工具来源mlx · 官方支持
- 社区证据 (2)公开报告只作为信号，不会把本页升级为项目测试。
- Reddit · Qwen3.8-27B on a 24GB M4 Pro Mac miniMac mini · M4 Pro · 24GB · llama.cpp b10488 · GGUF Q4_K_M / IQ4_XSQ4_K_M 17.77GB; about 11.4 tok/s decode; about 16.6GB resident with q8 KV at 32K context · medium confidenceA first-person report says the 27B model is usable on 24GB, but the memory ceiling is visible once context and KV cache are included. 限制: One user, one build, and one workload; not a Keep or Upgrade test.
- Tech media / blog · Qwen3.8-27B 4-bit on a 32GB M1 ProMacBook Pro · M1 Pro · 32GB · MLX-VLM · MLX 4-bit8–8.7 tok/s; weights about 16.05GB; peak memory about 18.5–21.7GB · medium confidenceA detailed local deployment log reports a stable 32GB starting point and explicitly discourages 16GB for this 27B representative. 限制: Community log with different thermals, context, and prompt mix from the product matrix.
- Public file metadata or a parameter estimate is not a first-party benchmark.
- Community artifacts and reports do not imply vendor endorsement or Keep or Upgrade testing.
- Social evidence is directional: build, context, thermals, storage streaming, and prompt mix can change the result.
- Context, concurrency, tool calling, JSON, vision, and sustained behavior remain separate capability checks.
- The primary matrix prioritizes current flagship releases; legacy versions remain compatibility records but are intentionally omitted here.
Qwen · 主流
不适配Qwen3.8-Flash-Next 180B (6B active)
Qwen3.8-Flash-Next · 180B total / 6B active
- 量化 / 文件
- BF16 · Safetensors360.0 GB · Public file metadata
- 所选 Mac 上的容量
- -352GB 剩余
- 运行工具证据
- 社区报告mlx
- 置信度
- medium核验 2026-08-28
来源与限制
- 模型官方来源模型 family, parameters, and release context · 核验 2026-08-28
- Public file metadata360.0 GB · exact public file metadata
- 运行工具来源mlx · 社区报告
- 社区证据 (1)公开报告只作为信号，不会把本页升级为项目测试。
- Reddit · Latest-model comparison thread (non-Apple hardware)4× DGX Spark / other non-Apple systemsUsers discuss Qwen3.8 Flash as an exploration/subagent model; no Mac tuple or reproducible throughput is provided. · low confidenceThis is a freshness and interest signal only, not evidence that Flash Next fits a Mac configuration. 限制: Cross-hardware discussion; excluded from Mac capacity conclusions.
- Public file metadata or a parameter estimate is not a first-party benchmark.
- Community artifacts and reports do not imply vendor endorsement or Keep or Upgrade testing.
- Social evidence is directional: build, context, thermals, storage streaming, and prompt mix can change the result.
- Context, concurrency, tool calling, JSON, vision, and sustained behavior remain separate capability checks.
- The primary matrix prioritizes current flagship releases; legacy versions remain compatibility records but are intentionally omitted here.
Qwen · 边界
不适配Qwen3.8 2.4T-A95B
Qwen3.8 · 2400B total / 95B active
- 量化 / 文件
- Q4_K_M · GGUF1440.0 GB · Public model source; exact file not registered
- 所选 Mac 上的容量
- -1432GB 剩余
- 运行工具证据
- 官方支持mlx
- 置信度
- medium核验 2026-08-28
来源与限制
- 模型官方来源模型 family, parameters, and release context · 核验 2026-08-28
- Public model source; exact file not registered1440.0 GB · parameter planning estimate
- 运行工具来源mlx · 官方支持
- 本次刷新未捕获可信的 Mac 第一手报告。容量仍是公开元数据规划估算；没有证据不会升级 runtime 结论。
- Public file metadata or a parameter estimate is not a first-party benchmark.
- Community artifacts and reports do not imply vendor endorsement or Keep or Upgrade testing.
- Social evidence is directional: build, context, thermals, storage streaming, and prompt mix can change the result.
- Context, concurrency, tool calling, JSON, vision, and sustained behavior remain separate capability checks.
- The primary matrix prioritizes current flagship releases; legacy versions remain compatibility records but are intentionally omitted here.
DeepSeek · 主流
不适配DeepSeek-V4-Flash 284B-A13B
DeepSeek V4 · 284B total / 13B active
- 量化 / 文件
- MLX-4bit · MLX170.4 GB · Public model source; exact file not registered
- 所选 Mac 上的容量
- -162.4GB 剩余
- 运行工具证据
- 社区报告mlx
- 置信度
- high核验 2026-08-28
来源与限制
- 模型官方来源模型 family, parameters, and release context · 核验 2026-08-28
- Public model source; exact file not registered170.4 GB · parameter planning estimate
- 运行工具来源mlx · 社区报告
- 社区证据 (7)公开报告只作为信号，不会把本页升级为项目测试。
- Reddit · DeepSeek V4 Flash on an M5 Max 128GBMac · Apple M5 Max · 128GB · ds4 / SSD-streaming path · DwarfStar IQ2XXSAbout 81GB model working set and about 31.06 tok/s in the reported run · medium confidenceA direct Apple Silicon report shows a practical 128GB path for the Flash representative. 限制: Community benchmark; exact context, thermal state, and build can change the result.
- Tech media / blog · DeepSeek V4 Flash on an M3 Max 128GBMacBook Pro · M3 Max · 128GB · ds4 experimental forkAbout 21 tok/s and about 81GB resident in the reported run · medium confidenceAn independent technical write-up corroborates that 128GB-class Apple Silicon can run a local Flash path. 限制: The article is tied to an experimental fork and does not validate mainline Ollama or llama.cpp.
- Reddit · DeepSeek V4 Flash on 64GB Macs with SSD streaming64GB Apple Silicon Macs · 64GB · ds4 SSD streaming · IQ2XXS / 2-bit classReports cluster around roughly 10–15 tok/s; technically possible, but not a comfortable in-memory workflow · medium confidence64GB appears technically viable only with aggressive quantization and storage streaming, so it is kept as a constrained edge case. 限制: Mixed community reports and different Mac generations; do not read this as a 64GB recommendation.
- GitHub · MLX DeepSeek V4 residency growth and resource-limit crashM4 Max 128GB and other Apple Silicon reports · 128GB · MLX-LMReproduction reports deterministic resource-limit failure around 11,300 generated tokens · high confidenceThe issue is important negative evidence: a model can fit and still fail on long generations because of runtime memory behavior. 限制: Issue state and fixes can change; check the runtime version and referenced patch before relying on MLX.
- GitHub · MLX cache/residency failure on an M3 Ultra 512GBMac Studio · M3 Ultra 512GB · 512GB · MLX-LMReports failure around 11,456–11,488 generated tokens in a production-style run · high confidenceEven very large unified-memory systems are not automatically stable for long-context MLX runs. 限制: This is a runtime bug report, not a capacity limit; a patched release may change the result.
- GitHub · DeepSeek V4 Flash Q8 garbled output on Apple NEONApple Silicon CPU / NEON · llama.cpp · Q8A specific repack path produced garbled output · high confidenceThis narrows the risk to a quantization/repack/runtime combination instead of treating every llama.cpp path as equivalent. 限制: Do not generalize one broken Q8 repack to all DeepSeek V4 Flash formats.
- GitHub · ds4 Apple Silicon throughput tableMac Studio · M3 Ultra 512GB · 512GB · ds4 Metal/CUDA engine · Q4 classRepository table reports short/long generation figures around 78.95 / 35.50 tok/s · medium confidenceA purpose-built engine provides a useful upper-bound signal for a large-memory Mac Studio, but it is not a mainstream runtime guarantee. 限制: Single graph worker, no batching, and engine-specific optimizations; compare only directionally.
- Public file metadata or a parameter estimate is not a first-party benchmark.
- Community artifacts and reports do not imply vendor endorsement or Keep or Upgrade testing.
- Social evidence is directional: build, context, thermals, storage streaming, and prompt mix can change the result.
- Context, concurrency, tool calling, JSON, vision, and sustained behavior remain separate capability checks.
- The primary matrix prioritizes current flagship releases; legacy versions remain compatibility records but are intentionally omitted here.
DeepSeek · 边界
不适配DeepSeek-V4-Pro 1.6T-A49B
DeepSeek V4 · 1600B total / 49B active
- 量化 / 文件
- MLX-4bit · MLX960.0 GB · Public model source; exact file not registered
- 所选 Mac 上的容量
- -952GB 剩余
- 运行工具证据
- 未知llama.cpp
- 置信度
- high核验 2026-08-28
来源与限制
- 模型官方来源模型 family, parameters, and release context · 核验 2026-08-28
- Public model source; exact file not registered960.0 GB · parameter planning estimate
- 运行工具来源llama.cpp · 未知
- 本次刷新未捕获可信的 Mac 第一手报告。容量仍是公开元数据规划估算；没有证据不会升级 runtime 结论。
- Public file metadata or a parameter estimate is not a first-party benchmark.
- Community artifacts and reports do not imply vendor endorsement or Keep or Upgrade testing.
- Social evidence is directional: build, context, thermals, storage streaming, and prompt mix can change the result.
- Context, concurrency, tool calling, JSON, vision, and sustained behavior remain separate capability checks.
- The primary matrix prioritizes current flagship releases; legacy versions remain compatibility records but are intentionally omitted here.
GLM · 主流
不适配GLM-5.3-Flash 320B-A18B
GLM-5.3-Flash · 320B total / 18B active
- 量化 / 文件
- MLX-4bit · MLX204.0 GB · Public file metadata
- 所选 Mac 上的容量
- -195.99GB 剩余
- 运行工具证据
- 社区报告mlx
- 置信度
- high核验 2026-08-28
来源与限制
- 模型官方来源模型 family, parameters, and release context · 核验 2026-08-28
- Public file metadata204.0 GB · exact public file metadata
- 运行工具来源mlx · 社区报告
- 社区证据 (2)公开报告只作为信号，不会把本页升级为项目测试。
- Hugging Face · Community GLM-5.3-Flash MLX quantization ladderMLX · 2bit-lite / 2 / 3 / 4 / 6-bit MLXPublic metadata lists about 102.43GB to 295.60GB per variant; the 4-bit root mirror is about 203.99GB · medium confidenceThe newer repository makes the precision trade-off explicit and includes a 2bit-lite path, but it does not prove loadability or speed on a particular Mac. 限制: Community conversion; full-repository size includes duplicate root and variant folders, and no trustworthy Apple first-person benchmark is attached.
- Reddit · GLM-5.3 Flash comparison thread (non-Apple hardware)Dual DGX Spark / other non-Apple systemsOne report describes roughly 22 tok/s on dual Spark and a slower, more verbose interaction style · low confidenceThis indicates active community comparison but cannot be translated into a Mac recommendation. 限制: Cross-hardware, anecdotal, and workload-specific.
- Public file metadata or a parameter estimate is not a first-party benchmark.
- Community artifacts and reports do not imply vendor endorsement or Keep or Upgrade testing.
- Social evidence is directional: build, context, thermals, storage streaming, and prompt mix can change the result.
- Context, concurrency, tool calling, JSON, vision, and sustained behavior remain separate capability checks.
- The primary matrix prioritizes current flagship releases; legacy versions remain compatibility records but are intentionally omitted here.
GLM · 边界
不适配GLM-5.3 744B-A40B
GLM-5.3 · 744B total / 40B active
- 量化 / 文件
- MLX-4bit · MLX446.4 GB · Public model source; exact file not registered
- 所选 Mac 上的容量
- -438.4GB 剩余
- 运行工具证据
- 未知llama.cpp
- 置信度
- high核验 2026-08-28
来源与限制
- 模型官方来源模型 family, parameters, and release context · 核验 2026-08-28
- Public model source; exact file not registered446.4 GB · parameter planning estimate
- 运行工具来源llama.cpp · 未知
- 本次刷新未捕获可信的 Mac 第一手报告。容量仍是公开元数据规划估算；没有证据不会升级 runtime 结论。
- Public file metadata or a parameter estimate is not a first-party benchmark.
- Community artifacts and reports do not imply vendor endorsement or Keep or Upgrade testing.
- Social evidence is directional: build, context, thermals, storage streaming, and prompt mix can change the result.
- Context, concurrency, tool calling, JSON, vision, and sustained behavior remain separate capability checks.
- The primary matrix prioritizes current flagship releases; legacy versions remain compatibility records but are intentionally omitted here.
Google Gemma · 小型
有余量Gemma 4 E2B
Gemma 4 · 2.3B effective / 5.1B total · PLE
- 量化 / 文件
- Q4_0 · QAT3.35 GB · Public file metadata
- 所选 Mac 上的容量
- 4.65GB 剩余
- 运行工具证据
- 社区报告llama.cpp
- 置信度
- high核验 2026-08-28
来源与限制
- 模型官方来源模型 family, parameters, and release context · 核验 2026-08-28
- Public file metadata3.35 GB · exact public file metadata
- 运行工具来源llama.cpp · 社区报告
- 社区证据 (1)公开报告只作为信号，不会把本页升级为项目测试。
- Tech media / blog · Apple Silicon local-AI comparison including Gemma 4Mac Studio / M4 Max 128GB test systems · 128GB · llama.cppGemma 4 appears in the comparison; the article warns that bandwidth is not the sole predictor and does not publish this exact matrix tuple · low confidenceUseful context for Apple Silicon behavior, but not an exact Gemma 4 E2B fit or performance claim. 限制: Different model/quantization details and media test methodology from this product matrix.
- Public file metadata or a parameter estimate is not a first-party benchmark.
- Community artifacts and reports do not imply vendor endorsement or Keep or Upgrade testing.
- Social evidence is directional: build, context, thermals, storage streaming, and prompt mix can change the result.
- Context, concurrency, tool calling, JSON, vision, and sustained behavior remain separate capability checks.
- The primary matrix prioritizes current flagship releases; legacy versions remain compatibility records but are intentionally omitted here.
Google Gemma · 主流
紧配Gemma 4 12B Unified
Gemma 4 · 11.95B dense
- 量化 / 文件
- Q4_0 · QAT6.98 GB · Public file metadata
- 所选 Mac 上的容量
- 1.02GB 剩余
- 运行工具证据
- 未知llama.cpp
- 置信度
- high核验 2026-08-28
来源与限制
- 模型官方来源模型 family, parameters, and release context · 核验 2026-08-28
- Public file metadata6.98 GB · exact public file metadata
- 运行工具来源llama.cpp · 未知
- 本次刷新未捕获可信的 Mac 第一手报告。容量仍是公开元数据规划估算；没有证据不会升级 runtime 结论。
- Public file metadata or a parameter estimate is not a first-party benchmark.
- Community artifacts and reports do not imply vendor endorsement or Keep or Upgrade testing.
- Social evidence is directional: build, context, thermals, storage streaming, and prompt mix can change the result.
- Context, concurrency, tool calling, JSON, vision, and sustained behavior remain separate capability checks.
- The primary matrix prioritizes current flagship releases; legacy versions remain compatibility records but are intentionally omitted here.
Google Gemma · 边界
不适配Gemma 4 26B-A4B
Gemma 4 · 25.2B total / 3.8B active
- 量化 / 文件
- Q4_0 · QAT14.4 GB · Public file metadata
- 所选 Mac 上的容量
- -6.44GB 剩余
- 运行工具证据
- 社区报告llama.cpp
- 置信度
- high核验 2026-08-28
来源与限制
- 模型官方来源模型 family, parameters, and release context · 核验 2026-08-28
- Public file metadata14.4 GB · exact public file metadata
- 运行工具来源llama.cpp · 社区报告
- 本次刷新未捕获可信的 Mac 第一手报告。容量仍是公开元数据规划估算；没有证据不会升级 runtime 结论。
- Public file metadata or a parameter estimate is not a first-party benchmark.
- Community artifacts and reports do not imply vendor endorsement or Keep or Upgrade testing.
- Social evidence is directional: build, context, thermals, storage streaming, and prompt mix can change the result.
- Context, concurrency, tool calling, JSON, vision, and sustained behavior remain separate capability checks.
- The primary matrix prioritizes current flagship releases; legacy versions remain compatibility records but are intentionally omitted here.
能力证据
工作流能力是另一道问题
上下文窗口是模型元数据。并发、工具调用、JSON 可靠性、视觉和持续行为取决于准确模型、运行工具、客户端、内存压力和任务；正面单元格也不是你工作流的保证。
| 模型 | 上下文 | 并发 | 工具调用 | JSON | 视觉 | 持续运行 |
|---|---|---|---|---|---|---|
| Qwen3.8 27B小型 | 官方支持Public model metadata lists 262,144 tokens. This is not proof that the selected Mac can hold that context with the weights.来源 | 系统推断Concurrency is inferred only from memory budget and runtime architecture. No apples-to-apples public concurrency benchmark is attached.来源 | 官方支持The current Qwen material documents tool or agent paths. The exact model template and runtime still need a workflow check.来源 | 未知Structured JSON behavior depends on the exact model template, runtime parser, and client; no exact public result is registered here. | 未知Vision support is not assumed from family branding or an unrelated model variant. | 未知No public, comparable sustained Apple Silicon run is registered for this model and Mac configuration. |
| Qwen3.8-Flash-Next 180B (6B active)主流 | 官方支持Public model metadata lists 262,144 tokens. This is not proof that the selected Mac can hold that context with the weights.来源 | 系统推断Concurrency is inferred only from memory budget and runtime architecture. No apples-to-apples public concurrency benchmark is attached.来源 | 官方支持The current Qwen material documents tool or agent paths. The exact model template and runtime still need a workflow check.来源 | 未知Structured JSON behavior depends on the exact model template, runtime parser, and client; no exact public result is registered here. | 官方支持The current model documentation describes multimodal inputs for this representative; the selected Mac runtime path still needs checking.来源 | 未知No public, comparable sustained Apple Silicon run is registered for this model and Mac configuration. |
| Qwen3.8 2.4T-A95B边界 | 官方支持Public model metadata lists 262,144 tokens. This is not proof that the selected Mac can hold that context with the weights.来源 | 系统推断Concurrency is inferred only from memory budget and runtime architecture. No apples-to-apples public concurrency benchmark is attached.来源 | 官方支持The current Qwen material documents tool or agent paths. The exact model template and runtime still need a workflow check.来源 | 未知Structured JSON behavior depends on the exact model template, runtime parser, and client; no exact public result is registered here. | 未知Vision support is not assumed from family branding or an unrelated model variant. | 未知No public, comparable sustained Apple Silicon run is registered for this model and Mac configuration. |
| DeepSeek-V4-Flash 284B-A13B主流 | 官方支持Public model metadata lists 1,048,576 tokens. This is not proof that the selected Mac can hold that context with the weights.来源 | 系统推断Concurrency is inferred only from memory budget and runtime architecture. No apples-to-apples public concurrency benchmark is attached.来源 | 官方支持The official DeepSeek V4 material documents tool use in the model family. The exact local template and runtime still need a workflow check.来源 | 未知Structured JSON behavior depends on the exact model template, runtime parser, and client; no exact public result is registered here. | 未知Vision support is not assumed from family branding or an unrelated model variant. | 未知No public, comparable sustained Apple Silicon run is registered for this model and Mac configuration. |
| DeepSeek-V4-Pro 1.6T-A49B边界 | 官方支持Public model metadata lists 1,048,576 tokens. This is not proof that the selected Mac can hold that context with the weights.来源 | 系统推断Concurrency is inferred only from memory budget and runtime architecture. No apples-to-apples public concurrency benchmark is attached.来源 | 官方支持The official DeepSeek V4 material documents tool use in the model family. The exact local template and runtime still need a workflow check.来源 | 未知Structured JSON behavior depends on the exact model template, runtime parser, and client; no exact public result is registered here. | 未知Vision support is not assumed from family branding or an unrelated model variant. | 未知No public, comparable sustained Apple Silicon run is registered for this model and Mac configuration. |
| GLM-5.3-Flash 320B-A18B主流 | 官方支持Public model metadata lists 1,000,000 tokens. This is not proof that the selected Mac can hold that context with the weights.来源 | 系统推断Concurrency is inferred only from memory budget and runtime architecture. No apples-to-apples public concurrency benchmark is attached.来源 | 官方支持The official GLM-5.3 material documents function calling and agent capabilities. It is not a Mac runtime integration benchmark.来源 | 公开基准Public GLM evaluations cover agentic or structured tasks, but they do not establish JSON validity in every client.来源 | 官方支持The current model documentation describes multimodal inputs for this representative; the selected Mac runtime path still needs checking.来源 | 未知No public, comparable sustained Apple Silicon run is registered for this model and Mac configuration. |
| GLM-5.3 744B-A40B边界 | 官方支持Public model metadata lists 1,000,000 tokens. This is not proof that the selected Mac can hold that context with the weights.来源 | 系统推断Concurrency is inferred only from memory budget and runtime architecture. No apples-to-apples public concurrency benchmark is attached.来源 | 官方支持The official GLM-5.3 material documents function calling and agent capabilities. It is not a Mac runtime integration benchmark.来源 | 未知Structured JSON behavior depends on the exact model template, runtime parser, and client; no exact public result is registered here. | 未知Vision support is not assumed from family branding or an unrelated model variant. | 未知No public, comparable sustained Apple Silicon run is registered for this model and Mac configuration. |
| Gemma 4 E2B小型 | 官方支持Public model metadata lists 131,072 tokens. This is not proof that the selected Mac can hold that context with the weights.来源 | 系统推断Concurrency is inferred only from memory budget and runtime architecture. No apples-to-apples public concurrency benchmark is attached.来源 | 未知No model-specific public tool-calling result is strong enough for a positive claim in this snapshot. | 未知Structured JSON behavior depends on the exact model template, runtime parser, and client; no exact public result is registered here. | 官方支持The current model documentation describes multimodal inputs for this representative; the selected Mac runtime path still needs checking.来源 | 未知No public, comparable sustained Apple Silicon run is registered for this model and Mac configuration. |
| Gemma 4 12B Unified主流 | 官方支持Public model metadata lists 262,144 tokens. This is not proof that the selected Mac can hold that context with the weights.来源 | 系统推断Concurrency is inferred only from memory budget and runtime architecture. No apples-to-apples public concurrency benchmark is attached.来源 | 未知No model-specific public tool-calling result is strong enough for a positive claim in this snapshot. | 未知Structured JSON behavior depends on the exact model template, runtime parser, and client; no exact public result is registered here. | 未知Vision support is not assumed from family branding or an unrelated model variant. | 未知No public, comparable sustained Apple Silicon run is registered for this model and Mac configuration. |
| Gemma 4 26B-A4B边界 | 官方支持Public model metadata lists 262,144 tokens. This is not proof that the selected Mac can hold that context with the weights.来源 | 系统推断Concurrency is inferred only from memory budget and runtime architecture. No apples-to-apples public concurrency benchmark is attached.来源 | 未知No model-specific public tool-calling result is strong enough for a positive claim in this snapshot. | 未知Structured JSON behavior depends on the exact model template, runtime parser, and client; no exact public result is registered here. | 未知Vision support is not assumed from family branding or an unrelated model variant. | 未知No public, comparable sustained Apple Silicon run is registered for this model and Mac configuration. |
Qwen3.8 27B小型
- 上下文
- 官方支持Public model metadata lists 262,144 tokens. This is not proof that the selected Mac can hold that context with the weights.来源
- 并发
- 系统推断Concurrency is inferred only from memory budget and runtime architecture. No apples-to-apples public concurrency benchmark is attached.来源
- 工具调用
- 官方支持The current Qwen material documents tool or agent paths. The exact model template and runtime still need a workflow check.来源
- JSON
- 未知Structured JSON behavior depends on the exact model template, runtime parser, and client; no exact public result is registered here.
- 视觉
- 未知Vision support is not assumed from family branding or an unrelated model variant.
- 持续运行
- 未知No public, comparable sustained Apple Silicon run is registered for this model and Mac configuration.
Qwen3.8-Flash-Next 180B (6B active)主流
- 上下文
- 官方支持Public model metadata lists 262,144 tokens. This is not proof that the selected Mac can hold that context with the weights.来源
- 并发
- 系统推断Concurrency is inferred only from memory budget and runtime architecture. No apples-to-apples public concurrency benchmark is attached.来源
- 工具调用
- 官方支持The current Qwen material documents tool or agent paths. The exact model template and runtime still need a workflow check.来源
- JSON
- 未知Structured JSON behavior depends on the exact model template, runtime parser, and client; no exact public result is registered here.
- 视觉
- 官方支持The current model documentation describes multimodal inputs for this representative; the selected Mac runtime path still needs checking.来源
- 持续运行
- 未知No public, comparable sustained Apple Silicon run is registered for this model and Mac configuration.
Qwen3.8 2.4T-A95B边界
- 上下文
- 官方支持Public model metadata lists 262,144 tokens. This is not proof that the selected Mac can hold that context with the weights.来源
- 并发
- 系统推断Concurrency is inferred only from memory budget and runtime architecture. No apples-to-apples public concurrency benchmark is attached.来源
- 工具调用
- 官方支持The current Qwen material documents tool or agent paths. The exact model template and runtime still need a workflow check.来源
- JSON
- 未知Structured JSON behavior depends on the exact model template, runtime parser, and client; no exact public result is registered here.
- 视觉
- 未知Vision support is not assumed from family branding or an unrelated model variant.
- 持续运行
- 未知No public, comparable sustained Apple Silicon run is registered for this model and Mac configuration.
DeepSeek-V4-Flash 284B-A13B主流
- 上下文
- 官方支持Public model metadata lists 1,048,576 tokens. This is not proof that the selected Mac can hold that context with the weights.来源
- 并发
- 系统推断Concurrency is inferred only from memory budget and runtime architecture. No apples-to-apples public concurrency benchmark is attached.来源
- 工具调用
- 官方支持The official DeepSeek V4 material documents tool use in the model family. The exact local template and runtime still need a workflow check.来源
- JSON
- 未知Structured JSON behavior depends on the exact model template, runtime parser, and client; no exact public result is registered here.
- 视觉
- 未知Vision support is not assumed from family branding or an unrelated model variant.
- 持续运行
- 未知No public, comparable sustained Apple Silicon run is registered for this model and Mac configuration.
DeepSeek-V4-Pro 1.6T-A49B边界
- 上下文
- 官方支持Public model metadata lists 1,048,576 tokens. This is not proof that the selected Mac can hold that context with the weights.来源
- 并发
- 系统推断Concurrency is inferred only from memory budget and runtime architecture. No apples-to-apples public concurrency benchmark is attached.来源
- 工具调用
- 官方支持The official DeepSeek V4 material documents tool use in the model family. The exact local template and runtime still need a workflow check.来源
- JSON
- 未知Structured JSON behavior depends on the exact model template, runtime parser, and client; no exact public result is registered here.
- 视觉
- 未知Vision support is not assumed from family branding or an unrelated model variant.
- 持续运行
- 未知No public, comparable sustained Apple Silicon run is registered for this model and Mac configuration.
GLM-5.3-Flash 320B-A18B主流
- 上下文
- 官方支持Public model metadata lists 1,000,000 tokens. This is not proof that the selected Mac can hold that context with the weights.来源
- 并发
- 系统推断Concurrency is inferred only from memory budget and runtime architecture. No apples-to-apples public concurrency benchmark is attached.来源
- 工具调用
- 官方支持The official GLM-5.3 material documents function calling and agent capabilities. It is not a Mac runtime integration benchmark.来源
- JSON
- 公开基准Public GLM evaluations cover agentic or structured tasks, but they do not establish JSON validity in every client.来源
- 视觉
- 官方支持The current model documentation describes multimodal inputs for this representative; the selected Mac runtime path still needs checking.来源
- 持续运行
- 未知No public, comparable sustained Apple Silicon run is registered for this model and Mac configuration.
GLM-5.3 744B-A40B边界
- 上下文
- 官方支持Public model metadata lists 1,000,000 tokens. This is not proof that the selected Mac can hold that context with the weights.来源
- 并发
- 系统推断Concurrency is inferred only from memory budget and runtime architecture. No apples-to-apples public concurrency benchmark is attached.来源
- 工具调用
- 官方支持The official GLM-5.3 material documents function calling and agent capabilities. It is not a Mac runtime integration benchmark.来源
- JSON
- 未知Structured JSON behavior depends on the exact model template, runtime parser, and client; no exact public result is registered here.
- 视觉
- 未知Vision support is not assumed from family branding or an unrelated model variant.
- 持续运行
- 未知No public, comparable sustained Apple Silicon run is registered for this model and Mac configuration.
Gemma 4 E2B小型
- 上下文
- 官方支持Public model metadata lists 131,072 tokens. This is not proof that the selected Mac can hold that context with the weights.来源
- 并发
- 系统推断Concurrency is inferred only from memory budget and runtime architecture. No apples-to-apples public concurrency benchmark is attached.来源
- 工具调用
- 未知No model-specific public tool-calling result is strong enough for a positive claim in this snapshot.
- JSON
- 未知Structured JSON behavior depends on the exact model template, runtime parser, and client; no exact public result is registered here.
- 视觉
- 官方支持The current model documentation describes multimodal inputs for this representative; the selected Mac runtime path still needs checking.来源
- 持续运行
- 未知No public, comparable sustained Apple Silicon run is registered for this model and Mac configuration.
Gemma 4 12B Unified主流
- 上下文
- 官方支持Public model metadata lists 262,144 tokens. This is not proof that the selected Mac can hold that context with the weights.来源
- 并发
- 系统推断Concurrency is inferred only from memory budget and runtime architecture. No apples-to-apples public concurrency benchmark is attached.来源
- 工具调用
- 未知No model-specific public tool-calling result is strong enough for a positive claim in this snapshot.
- JSON
- 未知Structured JSON behavior depends on the exact model template, runtime parser, and client; no exact public result is registered here.
- 视觉
- 未知Vision support is not assumed from family branding or an unrelated model variant.
- 持续运行
- 未知No public, comparable sustained Apple Silicon run is registered for this model and Mac configuration.
Gemma 4 26B-A4B边界
- 上下文
- 官方支持Public model metadata lists 262,144 tokens. This is not proof that the selected Mac can hold that context with the weights.来源
- 并发
- 系统推断Concurrency is inferred only from memory budget and runtime architecture. No apples-to-apples public concurrency benchmark is attached.来源
- 工具调用
- 未知No model-specific public tool-calling result is strong enough for a positive claim in this snapshot.
- JSON
- 未知Structured JSON behavior depends on the exact model template, runtime parser, and client; no exact public result is registered here.
- 视觉
- 未知Vision support is not assumed from family branding or an unrelated model variant.
- 持续运行
- 未知No public, comparable sustained Apple Silicon run is registered for this model and Mac configuration.
如何阅读标签
- 官方支持 官方模型/runtime 文档描述了能力路径；具体客户端行为仍需核验。
- 公开基准 存在已发表评测，但不是 Mac 吞吐或工作流基准。
- 社区报告 公开报告或社区文件只是信号，不是项目测试或厂商保证。
- 系统推断 基于容量/runtime 架构的有界推断，不是实测结果。
- 未知 没有捕获到该主张的可靠公开证据。
- 项目已验证 只有项目冒烟台账中精确的模型、量化、runtime 与机器组合才算本项目已验证。
带入你的真实工作流
把公开证据变成个人决策。
免费 Advisor 会加入当前 Mac、上下文、并发、预算、兼容性和离线要求；公开证据不足时，它会明确写“需要核验”。
常见问题
购买前先看清证据边界。
这是 Keep or Upgrade 跑的基准测试吗？
不是。这是公开证据综合和规划工具；只有精确的项目 smoke 元组才会标为项目已验证。
Mac 容量如何计算？
有精确文件或分片总和时用所选公开文件大小；否则使用清楚标注的规划估算。模型预算会为 macOS、运行工具开销和上下文保留内存。
MoE 的活跃参数是否代表需要更少内存？
不是。存储的完整权重才是容量门槛；活跃参数只能解释每 token 的计算量。
官方上下文窗口能在每台 Mac 上运行吗？
不能。上下文窗口只是模型元数据；权重、运行工具、系统和其他应用之后，所选 Mac 仍要有足够内存。
可以直接按表格购买 Mac 吗？
先用矩阵缩小规划范围，再把当前 Mac、上下文、并发、客户端、预算、兼容性和离线要求带入免费 Advisor。公开证据不是购买保证。
带入真实工作流
把公开证据变成你的购买判断。
免费 Advisor 会加入当前 Mac、上下文、并发、预算、兼容性与离线要求；公开证据不足时会明确写需要核验。
Keep or Upgrade 是独立的决策支持工具；Apple 不赞助或背书本页。