Keep or Upgrade
적합

공개 근거 종합

Mac mini와 Mac Studio 로컬 모델 적합성

최신 Qwen3.8, DeepSeek V4, GLM-5.3, Gemma 4 오픈 웨이트 대표를 위한 계획 도구입니다. 공개 파일 사실과 용량 계산부터 시작하고 런타임과 작업 근거를 분리한 뒤 업그레이드가 타당한지 판단합니다.

마지막 검토: 2026년 8월 28일 · 독립적인 판단 지원이며 Apple의 조언이 아닙니다

짧은 결론

이 페이지에서 계열별 현재 고관심 대표를 비교하세요. 이전 릴리스는 호환성 기록으로만 남기고 주 매트릭스에서는 제외합니다. 모델이 메모리에 들어가도 도구 호출, JSON, 비전, 동시 실행, 지속 동작은 알 수 없을 수 있습니다.

대화형 도구

모델이 Mac에 맞는지 확인하세요.

모델 또는 Mac 구성을 골라 가중치 예산, 메모리 여유, 런타임 근거와 무료 Advisor로의 다음 단계를 확인합니다.

적합성 탐색기 열기
공개 근거 종합이며, 제1자 벤치마크가 아닙니다.

공식 모델 카드, 공개 GGUF/QAT 파일 메타데이터, 런타임 문서, 공개 평가, 명확히 표시한 커뮤니티 보고서를 결합한 계획 도구입니다. Keep or Upgrade가 모든 모델을 실행했다고 주장하지 않으며 Apple이나 모델 제공자도 이 페이지를 보증하지 않습니다.

CSV 다운로드JSON 다운로드 · 이 페이지와 같은 원본 데이터

다운로드는 출처 URL, 근거 수준, 데이터 날짜를 담은 인용 가능한 스냅샷입니다. 정식 설명은 이 Hub에 두며 다운로드 엔드포인트는 사이트맵에 넣지 않습니다.

Mac mini M6 · 16GB153GB/s · US$899 · 기본 시작가

모델 근거 매트릭스

최신 고관심 오픈 웨이트 모델

주 매트릭스는 현재 주요 출시와 Mac 계획에 유용한 대표를 우선합니다. 용량은 가능한 경우 공개 파일 크기를 사용하며 파라미터 수는 계획 추정입니다. 각 행의 출처와 제한을 펼치면 귀속 가능한 커뮤니티 신호와 한계를 볼 수 있습니다.

계열 / 단계모델과 버전양자화 / 파일선택 Mac에서의 용량런타임 근거신뢰도세부
Qwen소형Qwen3.8 27BQwen3.8 · 27B denseQ4_0 · GGUF16.1 GB · Public file metadata적합하지 않음-8.06GB 잔여공식 지원mlx2 소셜 신호 · 출처 보기high확인 2026-08-28
출처와 제한
  • 모델 공식 출처모델 family, parameters, and release context · 확인 2026-08-28
  • Public file metadata16.1 GB · exact public file metadata
  • 런타임 출처mlx · 공식 지원
  • 커뮤니티 근거 (2)공개 보고서는 신호일 뿐이며 프로젝트 테스트로 승격되지 않습니다.
  • 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. 제한: 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. 제한: 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.
무료 Advisor에 적용
Qwen주류Qwen3.8-Flash-Next 180B (6B active)Qwen3.8-Flash-Next · 180B total / 6B activeBF16 · Safetensors360.0 GB · Public file metadata적합하지 않음-352GB 잔여커뮤니티 보고mlx1 소셜 신호 · 출처 보기medium확인 2026-08-28
출처와 제한
  • 모델 공식 출처모델 family, parameters, and release context · 확인 2026-08-28
  • Public file metadata360.0 GB · exact public file metadata
  • 런타임 출처mlx · 커뮤니티 보고
  • 커뮤니티 근거 (1)공개 보고서는 신호일 뿐이며 프로젝트 테스트로 승격되지 않습니다.
  • 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. 제한: 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.
무료 Advisor에 적용
Qwen경계Qwen3.8 2.4T-A95BQwen3.8 · 2400B total / 95B activeQ4_K_M · GGUF1440.0 GB · Public model source; exact file not registered적합하지 않음-1432GB 잔여공식 지원mlxmedium확인 2026-08-28
출처와 제한
  • 모델 공식 출처모델 family, parameters, and release context · 확인 2026-08-28
  • Public model source; exact file not registered1440.0 GB · parameter planning estimate
  • 런타임 출처mlx · 공식 지원
  • 이번 새로 고침에서 신뢰할 수 있는 Mac 실제 사용 보고서를 찾지 못했습니다.용량은 공개 메타데이터 기반 계획 추정이며 근거가 없다고 런타임 결론을 강화하지 않습니다.
  • 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.
무료 Advisor에 적용
DeepSeek주류DeepSeek-V4-Flash 284B-A13BDeepSeek V4 · 284B total / 13B activeMLX-4bit · MLX170.4 GB · Public model source; exact file not registered적합하지 않음-162.4GB 잔여커뮤니티 보고mlx7 소셜 신호 · 출처 보기high확인 2026-08-28
출처와 제한
  • 모델 공식 출처모델 family, parameters, and release context · 확인 2026-08-28
  • Public model source; exact file not registered170.4 GB · parameter planning estimate
  • 런타임 출처mlx · 커뮤니티 보고
  • 커뮤니티 근거 (7)공개 보고서는 신호일 뿐이며 프로젝트 테스트로 승격되지 않습니다.
  • 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. 제한: 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. 제한: 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. 제한: 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. 제한: 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. 제한: 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. 제한: 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. 제한: 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.
무료 Advisor에 적용
DeepSeek경계DeepSeek-V4-Pro 1.6T-A49BDeepSeek V4 · 1600B total / 49B activeMLX-4bit · MLX960.0 GB · Public model source; exact file not registered적합하지 않음-952GB 잔여알 수 없음llama.cpphigh확인 2026-08-28
출처와 제한
  • 모델 공식 출처모델 family, parameters, and release context · 확인 2026-08-28
  • Public model source; exact file not registered960.0 GB · parameter planning estimate
  • 런타임 출처llama.cpp · 알 수 없음
  • 이번 새로 고침에서 신뢰할 수 있는 Mac 실제 사용 보고서를 찾지 못했습니다.용량은 공개 메타데이터 기반 계획 추정이며 근거가 없다고 런타임 결론을 강화하지 않습니다.
  • 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.
무료 Advisor에 적용
GLM주류GLM-5.3-Flash 320B-A18BGLM-5.3-Flash · 320B total / 18B activeMLX-4bit · MLX204.0 GB · Public file metadata적합하지 않음-195.99GB 잔여커뮤니티 보고mlx2 소셜 신호 · 출처 보기high확인 2026-08-28
출처와 제한
  • 모델 공식 출처모델 family, parameters, and release context · 확인 2026-08-28
  • Public file metadata204.0 GB · exact public file metadata
  • 런타임 출처mlx · 커뮤니티 보고
  • 커뮤니티 근거 (2)공개 보고서는 신호일 뿐이며 프로젝트 테스트로 승격되지 않습니다.
  • 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. 제한: 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. 제한: 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.
무료 Advisor에 적용
GLM경계GLM-5.3 744B-A40BGLM-5.3 · 744B total / 40B activeMLX-4bit · MLX446.4 GB · Public model source; exact file not registered적합하지 않음-438.4GB 잔여알 수 없음llama.cpphigh확인 2026-08-28
출처와 제한
  • 모델 공식 출처모델 family, parameters, and release context · 확인 2026-08-28
  • Public model source; exact file not registered446.4 GB · parameter planning estimate
  • 런타임 출처llama.cpp · 알 수 없음
  • 이번 새로 고침에서 신뢰할 수 있는 Mac 실제 사용 보고서를 찾지 못했습니다.용량은 공개 메타데이터 기반 계획 추정이며 근거가 없다고 런타임 결론을 강화하지 않습니다.
  • 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.
무료 Advisor에 적용
Google Gemma소형Gemma 4 E2BGemma 4 · 2.3B effective / 5.1B total · PLEQ4_0 · QAT3.35 GB · Public file metadata여유 있음4.65GB 잔여커뮤니티 보고llama.cpp1 소셜 신호 · 출처 보기high확인 2026-08-28
출처와 제한
  • 모델 공식 출처모델 family, parameters, and release context · 확인 2026-08-28
  • Public file metadata3.35 GB · exact public file metadata
  • 런타임 출처llama.cpp · 커뮤니티 보고
  • 커뮤니티 근거 (1)공개 보고서는 신호일 뿐이며 프로젝트 테스트로 승격되지 않습니다.
  • 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. 제한: 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.
무료 Advisor에 적용
Google Gemma주류Gemma 4 12B UnifiedGemma 4 · 11.95B denseQ4_0 · QAT6.98 GB · Public file metadata여유 부족1.02GB 잔여알 수 없음llama.cpphigh확인 2026-08-28
출처와 제한
  • 모델 공식 출처모델 family, parameters, and release context · 확인 2026-08-28
  • Public file metadata6.98 GB · exact public file metadata
  • 런타임 출처llama.cpp · 알 수 없음
  • 이번 새로 고침에서 신뢰할 수 있는 Mac 실제 사용 보고서를 찾지 못했습니다.용량은 공개 메타데이터 기반 계획 추정이며 근거가 없다고 런타임 결론을 강화하지 않습니다.
  • 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.
무료 Advisor에 적용
Google Gemma경계Gemma 4 26B-A4BGemma 4 · 25.2B total / 3.8B activeQ4_0 · QAT14.4 GB · Public file metadata적합하지 않음-6.44GB 잔여커뮤니티 보고llama.cpphigh확인 2026-08-28
출처와 제한
  • 모델 공식 출처모델 family, parameters, and release context · 확인 2026-08-28
  • Public file metadata14.4 GB · exact public file metadata
  • 런타임 출처llama.cpp · 커뮤니티 보고
  • 이번 새로 고침에서 신뢰할 수 있는 Mac 실제 사용 보고서를 찾지 못했습니다.용량은 공개 메타데이터 기반 계획 추정이며 근거가 없다고 런타임 결론을 강화하지 않습니다.
  • 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.
무료 Advisor에 적용
Qwen · 소형

Qwen3.8 27B

Qwen3.8 · 27B dense

적합하지 않음
양자화 / 파일
Q4_0 · GGUF16.1 GB · Public file metadata
선택 Mac에서의 용량
-8.06GB 잔여
런타임 근거
공식 지원mlx
신뢰도
high확인 2026-08-28
출처와 제한
  • 모델 공식 출처모델 family, parameters, and release context · 확인 2026-08-28
  • Public file metadata16.1 GB · exact public file metadata
  • 런타임 출처mlx · 공식 지원
  • 커뮤니티 근거 (2)공개 보고서는 신호일 뿐이며 프로젝트 테스트로 승격되지 않습니다.
  • 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. 제한: 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. 제한: 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.
무료 Advisor에 적용
Qwen · 주류

Qwen3.8-Flash-Next 180B (6B active)

Qwen3.8-Flash-Next · 180B total / 6B active

적합하지 않음
양자화 / 파일
BF16 · Safetensors360.0 GB · Public file metadata
선택 Mac에서의 용량
-352GB 잔여
런타임 근거
커뮤니티 보고mlx
신뢰도
medium확인 2026-08-28
출처와 제한
  • 모델 공식 출처모델 family, parameters, and release context · 확인 2026-08-28
  • Public file metadata360.0 GB · exact public file metadata
  • 런타임 출처mlx · 커뮤니티 보고
  • 커뮤니티 근거 (1)공개 보고서는 신호일 뿐이며 프로젝트 테스트로 승격되지 않습니다.
  • 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. 제한: 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.
무료 Advisor에 적용
Qwen · 경계

Qwen3.8 2.4T-A95B

Qwen3.8 · 2400B total / 95B active

적합하지 않음
양자화 / 파일
Q4_K_M · GGUF1440.0 GB · Public model source; exact file not registered
선택 Mac에서의 용량
-1432GB 잔여
런타임 근거
공식 지원mlx
신뢰도
medium확인 2026-08-28
출처와 제한
  • 모델 공식 출처모델 family, parameters, and release context · 확인 2026-08-28
  • Public model source; exact file not registered1440.0 GB · parameter planning estimate
  • 런타임 출처mlx · 공식 지원
  • 이번 새로 고침에서 신뢰할 수 있는 Mac 실제 사용 보고서를 찾지 못했습니다.용량은 공개 메타데이터 기반 계획 추정이며 근거가 없다고 런타임 결론을 강화하지 않습니다.
  • 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.
무료 Advisor에 적용
DeepSeek · 주류

DeepSeek-V4-Flash 284B-A13B

DeepSeek V4 · 284B total / 13B active

적합하지 않음
양자화 / 파일
MLX-4bit · MLX170.4 GB · Public model source; exact file not registered
선택 Mac에서의 용량
-162.4GB 잔여
런타임 근거
커뮤니티 보고mlx
신뢰도
high확인 2026-08-28
출처와 제한
  • 모델 공식 출처모델 family, parameters, and release context · 확인 2026-08-28
  • Public model source; exact file not registered170.4 GB · parameter planning estimate
  • 런타임 출처mlx · 커뮤니티 보고
  • 커뮤니티 근거 (7)공개 보고서는 신호일 뿐이며 프로젝트 테스트로 승격되지 않습니다.
  • 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. 제한: 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. 제한: 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. 제한: 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. 제한: 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. 제한: 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. 제한: 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. 제한: 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.
무료 Advisor에 적용
DeepSeek · 경계

DeepSeek-V4-Pro 1.6T-A49B

DeepSeek V4 · 1600B total / 49B active

적합하지 않음
양자화 / 파일
MLX-4bit · MLX960.0 GB · Public model source; exact file not registered
선택 Mac에서의 용량
-952GB 잔여
런타임 근거
알 수 없음llama.cpp
신뢰도
high확인 2026-08-28
출처와 제한
  • 모델 공식 출처모델 family, parameters, and release context · 확인 2026-08-28
  • Public model source; exact file not registered960.0 GB · parameter planning estimate
  • 런타임 출처llama.cpp · 알 수 없음
  • 이번 새로 고침에서 신뢰할 수 있는 Mac 실제 사용 보고서를 찾지 못했습니다.용량은 공개 메타데이터 기반 계획 추정이며 근거가 없다고 런타임 결론을 강화하지 않습니다.
  • 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.
무료 Advisor에 적용
GLM · 주류

GLM-5.3-Flash 320B-A18B

GLM-5.3-Flash · 320B total / 18B active

적합하지 않음
양자화 / 파일
MLX-4bit · MLX204.0 GB · Public file metadata
선택 Mac에서의 용량
-195.99GB 잔여
런타임 근거
커뮤니티 보고mlx
신뢰도
high확인 2026-08-28
출처와 제한
  • 모델 공식 출처모델 family, parameters, and release context · 확인 2026-08-28
  • Public file metadata204.0 GB · exact public file metadata
  • 런타임 출처mlx · 커뮤니티 보고
  • 커뮤니티 근거 (2)공개 보고서는 신호일 뿐이며 프로젝트 테스트로 승격되지 않습니다.
  • 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. 제한: 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. 제한: 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.
무료 Advisor에 적용
GLM · 경계

GLM-5.3 744B-A40B

GLM-5.3 · 744B total / 40B active

적합하지 않음
양자화 / 파일
MLX-4bit · MLX446.4 GB · Public model source; exact file not registered
선택 Mac에서의 용량
-438.4GB 잔여
런타임 근거
알 수 없음llama.cpp
신뢰도
high확인 2026-08-28
출처와 제한
  • 모델 공식 출처모델 family, parameters, and release context · 확인 2026-08-28
  • Public model source; exact file not registered446.4 GB · parameter planning estimate
  • 런타임 출처llama.cpp · 알 수 없음
  • 이번 새로 고침에서 신뢰할 수 있는 Mac 실제 사용 보고서를 찾지 못했습니다.용량은 공개 메타데이터 기반 계획 추정이며 근거가 없다고 런타임 결론을 강화하지 않습니다.
  • 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.
무료 Advisor에 적용
Google Gemma · 소형

Gemma 4 E2B

Gemma 4 · 2.3B effective / 5.1B total · PLE

여유 있음
양자화 / 파일
Q4_0 · QAT3.35 GB · Public file metadata
선택 Mac에서의 용량
4.65GB 잔여
런타임 근거
커뮤니티 보고llama.cpp
신뢰도
high확인 2026-08-28
출처와 제한
  • 모델 공식 출처모델 family, parameters, and release context · 확인 2026-08-28
  • Public file metadata3.35 GB · exact public file metadata
  • 런타임 출처llama.cpp · 커뮤니티 보고
  • 커뮤니티 근거 (1)공개 보고서는 신호일 뿐이며 프로젝트 테스트로 승격되지 않습니다.
  • 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. 제한: 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.
무료 Advisor에 적용
Google Gemma · 주류

Gemma 4 12B Unified

Gemma 4 · 11.95B dense

여유 부족
양자화 / 파일
Q4_0 · QAT6.98 GB · Public file metadata
선택 Mac에서의 용량
1.02GB 잔여
런타임 근거
알 수 없음llama.cpp
신뢰도
high확인 2026-08-28
출처와 제한
  • 모델 공식 출처모델 family, parameters, and release context · 확인 2026-08-28
  • Public file metadata6.98 GB · exact public file metadata
  • 런타임 출처llama.cpp · 알 수 없음
  • 이번 새로 고침에서 신뢰할 수 있는 Mac 실제 사용 보고서를 찾지 못했습니다.용량은 공개 메타데이터 기반 계획 추정이며 근거가 없다고 런타임 결론을 강화하지 않습니다.
  • 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.
무료 Advisor에 적용
Google Gemma · 경계

Gemma 4 26B-A4B

Gemma 4 · 25.2B total / 3.8B active

적합하지 않음
양자화 / 파일
Q4_0 · QAT14.4 GB · Public file metadata
선택 Mac에서의 용량
-6.44GB 잔여
런타임 근거
커뮤니티 보고llama.cpp
신뢰도
high확인 2026-08-28
출처와 제한
  • 모델 공식 출처모델 family, parameters, and release context · 확인 2026-08-28
  • Public file metadata14.4 GB · exact public file metadata
  • 런타임 출처llama.cpp · 커뮤니티 보고
  • 이번 새로 고침에서 신뢰할 수 있는 Mac 실제 사용 보고서를 찾지 못했습니다.용량은 공개 메타데이터 기반 계획 추정이며 근거가 없다고 런타임 결론을 강화하지 않습니다.
  • 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.
무료 Advisor에 적용

기능 근거

작업 흐름 기능은 별도 질문입니다

컨텍스트 창은 모델 메타데이터입니다. 동시 실행, 도구 호출, JSON 신뢰성, 비전, 지속 동작은 정확한 모델, 런타임, 클라이언트, 메모리 압박, 작업에 따라 달라집니다. 긍정적인 셀도 작업 흐름을 보장하지 않습니다.

모델컨텍스트동시 실행도구 호출JSON비전지속 동작
Qwen3.8 27B소형
공식 지원Public model metadata lists 262,144 tokens. This is not proof that the selected Mac can hold that context with the weights.출처
시스템 추론Concurrency is inferred only from memory budget and runtime architecture. No apples-to-apples public concurrency benchmark is attached.출처
공식 지원The current Qwen material documents tool or agent paths. The exact model template and runtime still need a workflow check.출처
알 수 없음Structured JSON behavior depends on the exact model template, runtime parser, and client; no exact public result is registered here.
알 수 없음Vision support is not assumed from family branding or an unrelated model variant.
알 수 없음No public, comparable sustained Apple Silicon run is registered for this model and Mac configuration.
Qwen3.8-Flash-Next 180B (6B active)주류
공식 지원Public model metadata lists 262,144 tokens. This is not proof that the selected Mac can hold that context with the weights.출처
시스템 추론Concurrency is inferred only from memory budget and runtime architecture. No apples-to-apples public concurrency benchmark is attached.출처
공식 지원The current Qwen material documents tool or agent paths. The exact model template and runtime still need a workflow check.출처
알 수 없음Structured JSON behavior depends on the exact model template, runtime parser, and client; no exact public result is registered here.
공식 지원The current model documentation describes multimodal inputs for this representative; the selected Mac runtime path still needs checking.출처
알 수 없음No public, comparable sustained Apple Silicon run is registered for this model and Mac configuration.
Qwen3.8 2.4T-A95B경계
공식 지원Public model metadata lists 262,144 tokens. This is not proof that the selected Mac can hold that context with the weights.출처
시스템 추론Concurrency is inferred only from memory budget and runtime architecture. No apples-to-apples public concurrency benchmark is attached.출처
공식 지원The current Qwen material documents tool or agent paths. The exact model template and runtime still need a workflow check.출처
알 수 없음Structured JSON behavior depends on the exact model template, runtime parser, and client; no exact public result is registered here.
알 수 없음Vision support is not assumed from family branding or an unrelated model variant.
알 수 없음No public, comparable sustained Apple Silicon run is registered for this model and Mac configuration.
DeepSeek-V4-Flash 284B-A13B주류
공식 지원Public model metadata lists 1,048,576 tokens. This is not proof that the selected Mac can hold that context with the weights.출처
시스템 추론Concurrency is inferred only from memory budget and runtime architecture. No apples-to-apples public concurrency benchmark is attached.출처
공식 지원The official DeepSeek V4 material documents tool use in the model family. The exact local template and runtime still need a workflow check.출처
알 수 없음Structured JSON behavior depends on the exact model template, runtime parser, and client; no exact public result is registered here.
알 수 없음Vision support is not assumed from family branding or an unrelated model variant.
알 수 없음No public, comparable sustained Apple Silicon run is registered for this model and Mac configuration.
DeepSeek-V4-Pro 1.6T-A49B경계
공식 지원Public model metadata lists 1,048,576 tokens. This is not proof that the selected Mac can hold that context with the weights.출처
시스템 추론Concurrency is inferred only from memory budget and runtime architecture. No apples-to-apples public concurrency benchmark is attached.출처
공식 지원The official DeepSeek V4 material documents tool use in the model family. The exact local template and runtime still need a workflow check.출처
알 수 없음Structured JSON behavior depends on the exact model template, runtime parser, and client; no exact public result is registered here.
알 수 없음Vision support is not assumed from family branding or an unrelated model variant.
알 수 없음No public, comparable sustained Apple Silicon run is registered for this model and Mac configuration.
GLM-5.3-Flash 320B-A18B주류
공식 지원Public model metadata lists 1,000,000 tokens. This is not proof that the selected Mac can hold that context with the weights.출처
시스템 추론Concurrency is inferred only from memory budget and runtime architecture. No apples-to-apples public concurrency benchmark is attached.출처
공식 지원The official GLM-5.3 material documents function calling and agent capabilities. It is not a Mac runtime integration benchmark.출처
공개 벤치마크Public GLM evaluations cover agentic or structured tasks, but they do not establish JSON validity in every client.출처
공식 지원The current model documentation describes multimodal inputs for this representative; the selected Mac runtime path still needs checking.출처
알 수 없음No public, comparable sustained Apple Silicon run is registered for this model and Mac configuration.
GLM-5.3 744B-A40B경계
공식 지원Public model metadata lists 1,000,000 tokens. This is not proof that the selected Mac can hold that context with the weights.출처
시스템 추론Concurrency is inferred only from memory budget and runtime architecture. No apples-to-apples public concurrency benchmark is attached.출처
공식 지원The official GLM-5.3 material documents function calling and agent capabilities. It is not a Mac runtime integration benchmark.출처
알 수 없음Structured JSON behavior depends on the exact model template, runtime parser, and client; no exact public result is registered here.
알 수 없음Vision support is not assumed from family branding or an unrelated model variant.
알 수 없음No public, comparable sustained Apple Silicon run is registered for this model and Mac configuration.
Gemma 4 E2B소형
공식 지원Public model metadata lists 131,072 tokens. This is not proof that the selected Mac can hold that context with the weights.출처
시스템 추론Concurrency is inferred only from memory budget and runtime architecture. No apples-to-apples public concurrency benchmark is attached.출처
알 수 없음No model-specific public tool-calling result is strong enough for a positive claim in this snapshot.
알 수 없음Structured JSON behavior depends on the exact model template, runtime parser, and client; no exact public result is registered here.
공식 지원The current model documentation describes multimodal inputs for this representative; the selected Mac runtime path still needs checking.출처
알 수 없음No public, comparable sustained Apple Silicon run is registered for this model and Mac configuration.
Gemma 4 12B Unified주류
공식 지원Public model metadata lists 262,144 tokens. This is not proof that the selected Mac can hold that context with the weights.출처
시스템 추론Concurrency is inferred only from memory budget and runtime architecture. No apples-to-apples public concurrency benchmark is attached.출처
알 수 없음No model-specific public tool-calling result is strong enough for a positive claim in this snapshot.
알 수 없음Structured JSON behavior depends on the exact model template, runtime parser, and client; no exact public result is registered here.
알 수 없음Vision support is not assumed from family branding or an unrelated model variant.
알 수 없음No public, comparable sustained Apple Silicon run is registered for this model and Mac configuration.
Gemma 4 26B-A4B경계
공식 지원Public model metadata lists 262,144 tokens. This is not proof that the selected Mac can hold that context with the weights.출처
시스템 추론Concurrency is inferred only from memory budget and runtime architecture. No apples-to-apples public concurrency benchmark is attached.출처
알 수 없음No model-specific public tool-calling result is strong enough for a positive claim in this snapshot.
알 수 없음Structured JSON behavior depends on the exact model template, runtime parser, and client; no exact public result is registered here.
알 수 없음Vision support is not assumed from family branding or an unrelated model variant.
알 수 없음No public, comparable sustained Apple Silicon run is registered for this model and Mac configuration.
Qwen3.8 27B소형
컨텍스트
공식 지원Public model metadata lists 262,144 tokens. This is not proof that the selected Mac can hold that context with the weights.출처
동시 실행
시스템 추론Concurrency is inferred only from memory budget and runtime architecture. No apples-to-apples public concurrency benchmark is attached.출처
도구 호출
공식 지원The current Qwen material documents tool or agent paths. The exact model template and runtime still need a workflow check.출처
JSON
알 수 없음Structured JSON behavior depends on the exact model template, runtime parser, and client; no exact public result is registered here.
비전
알 수 없음Vision support is not assumed from family branding or an unrelated model variant.
지속 동작
알 수 없음No public, comparable sustained Apple Silicon run is registered for this model and Mac configuration.
Qwen3.8-Flash-Next 180B (6B active)주류
컨텍스트
공식 지원Public model metadata lists 262,144 tokens. This is not proof that the selected Mac can hold that context with the weights.출처
동시 실행
시스템 추론Concurrency is inferred only from memory budget and runtime architecture. No apples-to-apples public concurrency benchmark is attached.출처
도구 호출
공식 지원The current Qwen material documents tool or agent paths. The exact model template and runtime still need a workflow check.출처
JSON
알 수 없음Structured JSON behavior depends on the exact model template, runtime parser, and client; no exact public result is registered here.
비전
공식 지원The current model documentation describes multimodal inputs for this representative; the selected Mac runtime path still needs checking.출처
지속 동작
알 수 없음No public, comparable sustained Apple Silicon run is registered for this model and Mac configuration.
Qwen3.8 2.4T-A95B경계
컨텍스트
공식 지원Public model metadata lists 262,144 tokens. This is not proof that the selected Mac can hold that context with the weights.출처
동시 실행
시스템 추론Concurrency is inferred only from memory budget and runtime architecture. No apples-to-apples public concurrency benchmark is attached.출처
도구 호출
공식 지원The current Qwen material documents tool or agent paths. The exact model template and runtime still need a workflow check.출처
JSON
알 수 없음Structured JSON behavior depends on the exact model template, runtime parser, and client; no exact public result is registered here.
비전
알 수 없음Vision support is not assumed from family branding or an unrelated model variant.
지속 동작
알 수 없음No public, comparable sustained Apple Silicon run is registered for this model and Mac configuration.
DeepSeek-V4-Flash 284B-A13B주류
컨텍스트
공식 지원Public model metadata lists 1,048,576 tokens. This is not proof that the selected Mac can hold that context with the weights.출처
동시 실행
시스템 추론Concurrency is inferred only from memory budget and runtime architecture. No apples-to-apples public concurrency benchmark is attached.출처
도구 호출
공식 지원The official DeepSeek V4 material documents tool use in the model family. The exact local template and runtime still need a workflow check.출처
JSON
알 수 없음Structured JSON behavior depends on the exact model template, runtime parser, and client; no exact public result is registered here.
비전
알 수 없음Vision support is not assumed from family branding or an unrelated model variant.
지속 동작
알 수 없음No public, comparable sustained Apple Silicon run is registered for this model and Mac configuration.
DeepSeek-V4-Pro 1.6T-A49B경계
컨텍스트
공식 지원Public model metadata lists 1,048,576 tokens. This is not proof that the selected Mac can hold that context with the weights.출처
동시 실행
시스템 추론Concurrency is inferred only from memory budget and runtime architecture. No apples-to-apples public concurrency benchmark is attached.출처
도구 호출
공식 지원The official DeepSeek V4 material documents tool use in the model family. The exact local template and runtime still need a workflow check.출처
JSON
알 수 없음Structured JSON behavior depends on the exact model template, runtime parser, and client; no exact public result is registered here.
비전
알 수 없음Vision support is not assumed from family branding or an unrelated model variant.
지속 동작
알 수 없음No public, comparable sustained Apple Silicon run is registered for this model and Mac configuration.
GLM-5.3-Flash 320B-A18B주류
컨텍스트
공식 지원Public model metadata lists 1,000,000 tokens. This is not proof that the selected Mac can hold that context with the weights.출처
동시 실행
시스템 추론Concurrency is inferred only from memory budget and runtime architecture. No apples-to-apples public concurrency benchmark is attached.출처
도구 호출
공식 지원The official GLM-5.3 material documents function calling and agent capabilities. It is not a Mac runtime integration benchmark.출처
JSON
공개 벤치마크Public GLM evaluations cover agentic or structured tasks, but they do not establish JSON validity in every client.출처
비전
공식 지원The current model documentation describes multimodal inputs for this representative; the selected Mac runtime path still needs checking.출처
지속 동작
알 수 없음No public, comparable sustained Apple Silicon run is registered for this model and Mac configuration.
GLM-5.3 744B-A40B경계
컨텍스트
공식 지원Public model metadata lists 1,000,000 tokens. This is not proof that the selected Mac can hold that context with the weights.출처
동시 실행
시스템 추론Concurrency is inferred only from memory budget and runtime architecture. No apples-to-apples public concurrency benchmark is attached.출처
도구 호출
공식 지원The official GLM-5.3 material documents function calling and agent capabilities. It is not a Mac runtime integration benchmark.출처
JSON
알 수 없음Structured JSON behavior depends on the exact model template, runtime parser, and client; no exact public result is registered here.
비전
알 수 없음Vision support is not assumed from family branding or an unrelated model variant.
지속 동작
알 수 없음No public, comparable sustained Apple Silicon run is registered for this model and Mac configuration.
Gemma 4 E2B소형
컨텍스트
공식 지원Public model metadata lists 131,072 tokens. This is not proof that the selected Mac can hold that context with the weights.출처
동시 실행
시스템 추론Concurrency is inferred only from memory budget and runtime architecture. No apples-to-apples public concurrency benchmark is attached.출처
도구 호출
알 수 없음No model-specific public tool-calling result is strong enough for a positive claim in this snapshot.
JSON
알 수 없음Structured JSON behavior depends on the exact model template, runtime parser, and client; no exact public result is registered here.
비전
공식 지원The current model documentation describes multimodal inputs for this representative; the selected Mac runtime path still needs checking.출처
지속 동작
알 수 없음No public, comparable sustained Apple Silicon run is registered for this model and Mac configuration.
Gemma 4 12B Unified주류
컨텍스트
공식 지원Public model metadata lists 262,144 tokens. This is not proof that the selected Mac can hold that context with the weights.출처
동시 실행
시스템 추론Concurrency is inferred only from memory budget and runtime architecture. No apples-to-apples public concurrency benchmark is attached.출처
도구 호출
알 수 없음No model-specific public tool-calling result is strong enough for a positive claim in this snapshot.
JSON
알 수 없음Structured JSON behavior depends on the exact model template, runtime parser, and client; no exact public result is registered here.
비전
알 수 없음Vision support is not assumed from family branding or an unrelated model variant.
지속 동작
알 수 없음No public, comparable sustained Apple Silicon run is registered for this model and Mac configuration.
Gemma 4 26B-A4B경계
컨텍스트
공식 지원Public model metadata lists 262,144 tokens. This is not proof that the selected Mac can hold that context with the weights.출처
동시 실행
시스템 추론Concurrency is inferred only from memory budget and runtime architecture. No apples-to-apples public concurrency benchmark is attached.출처
도구 호출
알 수 없음No model-specific public tool-calling result is strong enough for a positive claim in this snapshot.
JSON
알 수 없음Structured JSON behavior depends on the exact model template, runtime parser, and client; no exact public result is registered here.
비전
알 수 없음Vision support is not assumed from family branding or an unrelated model variant.
지속 동작
알 수 없음No public, comparable sustained Apple Silicon run is registered for this model and Mac configuration.

라벨 읽는 법

  • 공식 지원 공식 모델/런타임 문서가 기능 경로를 설명하지만 정확한 클라이언트 동작은 확인이 필요합니다.
  • 공개 벤치마크 공개 평가가 있지만 Mac 처리량이나 작업 흐름 벤치마크는 아닙니다.
  • 커뮤니티 보고 공개 보고서나 커뮤니티 결과물은 신호일 뿐 프로젝트 테스트나 공급자 보증이 아닙니다.
  • 시스템 추론 용량/런타임 아키텍처에 기반한 제한적 추론이며 관측 결과가 아닙니다.
  • 알 수 없음 이 주장에 대한 신뢰할 수 있는 공개 근거가 없습니다.
  • 프로젝트 검증됨 프로젝트 스모크 원장에 있는 정확한 모델·양자화·런타임·머신 조합만 검증됨으로 표시됩니다.

실제 작업 가져오기

공개 근거를 개인 결정으로 바꾸세요.

무료 Advisor는 현재 Mac, 컨텍스트, 동시 실행, 예산, 호환성, 오프라인 요구를 추가하고 근거가 부족하면 “검증 필요”라고 명시합니다.

무료 Advisor 실행

자주 묻는 질문

구매 전에 근거의 경계를 확인하세요.

이것은 Keep or Upgrade가 실행한 벤치마크인가요?

아닙니다. 공개 근거 종합과 계획 도구입니다. 정확한 프로젝트 smoke 조합만 프로젝트 검증됨으로 표시됩니다.

Mac 용량은 어떻게 계산하나요?

정확한 파일 또는 샤드 합계가 있으면 선택한 공개 파일 크기를 쓰고, 그렇지 않으면 명시한 계획 추정을 씁니다. 모델 예산은 macOS, 런타임 오버헤드, 컨텍스트용 메모리를 남깁니다.

MoE 활성 파라미터는 더 적은 메모리를 뜻하나요?

아닙니다. 저장된 전체 가중치가 용량 관문입니다. 활성 파라미터는 토큰당 계산량만 설명합니다.

공식 컨텍스트 창은 모든 Mac에서 실행되나요?

아닙니다. 컨텍스트 창은 모델 메타데이터일 뿐입니다. 가중치, 런타임, 시스템, 다른 앱 뒤에도 선택 Mac에 충분한 메모리가 있어야 합니다.

표만 보고 Mac을 구매해도 되나요?

이 매트릭스로 계획 범위를 좁힌 뒤 현재 Mac, 컨텍스트, 동시 실행, 클라이언트, 예산, 호환성, 오프라인 요구를 무료 Advisor에 넣으세요. 공개 근거는 구매 보증이 아닙니다.

실제 작업 가져오기

공개 근거를 내 구매 판단으로 바꾸세요.

무료 Advisor는 현재 Mac, 컨텍스트, 동시 실행, 예산, 호환성, 오프라인 요구를 추가하며 공개 근거가 부족하면 검증 필요라고 명시합니다.

무료 Advisor 시작

Keep or Upgrade는 독립적인 의사결정 지원 도구이며 Apple의 후원을 받지 않습니다.