Silicon Score Apple Silicon local LLM rankings
← Canonical rankings
Canonical Rankings

Best Macs for this model

Qwen3.6-35B-A3B ranked across the Mac lineup at the best practical quantization, using the best available runtime evidence. Model picker is focused on current-market choices.

29 ranked MacsUse the strongest current runtime evidence for each row.28 historical models hiddenStatic paths cover only canonical model pages; sort and quantization stay as query state.
RankMacScoreQuantTok/sRuntimeFitsHeadroomContextEvidencePriceWhy it ranks here
1Mac Studio M3 Ultra 256GB4808bit 48.0 tok/s Fastest evidence path: 8bit · 48.0 tok/s · MLX · EstimatedMLXFits222.3 GB262kEstimated$7,4998bit is the current best practical quantization. 48.0 tok/s is estimated from nearby benchmark coverage. 222.3 GB headroom remains at this quantization.
2Mac Pro M2 Ultra 192GB4168bit 48.0 tok/s Fastest evidence path: 8bit · 48.0 tok/s · MLX · EstimatedMLXFits158.3 GB262kEstimated$6,9998bit is the current best practical quantization. 48.0 tok/s is estimated from nearby benchmark coverage. 158.3 GB headroom remains at this quantization.
3MacBook Pro M5 Max 128GB 16-inch3808bit 55.0 tok/s Fastest evidence path: 8bit · 55.0 tok/s · MLX · EstimatedMLXFits94.3 GB262kEstimated$5,3998bit is the current best practical quantization. 55.0 tok/s is estimated from nearby benchmark coverage. 94.3 GB headroom remains at this quantization.
4Mac Studio M4 Max 128GB3528bit 48.0 tok/s Fastest evidence path: 8bit · 48.0 tok/s · MLX · EstimatedMLXFits94.3 GB262kEstimated$4,4998bit is the current best practical quantization. 48.0 tok/s is estimated from nearby benchmark coverage. 94.3 GB headroom remains at this quantization.
5MacBook Pro M4 Max 128GB 16-inch3528bit 48.0 tok/s Fastest evidence path: 8bit · 48.0 tok/s · MLX · EstimatedMLXFits94.3 GB262kEstimated$5,9998bit is the current best practical quantization. 48.0 tok/s is estimated from nearby benchmark coverage. 94.3 GB headroom remains at this quantization.
6Mac Studio M3 Ultra 96GB3208bit 48.0 tok/s Fastest evidence path: 8bit · 48.0 tok/s · MLX · EstimatedMLXFits62.3 GB262kEstimated$3,9998bit is the current best practical quantization. 48.0 tok/s is estimated from nearby benchmark coverage. 62.3 GB headroom remains at this quantization.
7Mac Studio M4 Max 64GB2888bit 48.0 tok/s Fastest evidence path: 8bit · 48.0 tok/s · MLX · EstimatedMLXFits30.3 GB262kEstimated$2,9998bit is the current best practical quantization. 48.0 tok/s is estimated from nearby benchmark coverage. 30.3 GB headroom remains at this quantization.
8MacBook Pro M4 Max 64GB 16-inch2888bit 48.0 tok/s Fastest evidence path: 8bit · 48.0 tok/s · MLX · EstimatedMLXFits30.3 GB262kEstimated$4,4998bit is the current best practical quantization. 48.0 tok/s is estimated from nearby benchmark coverage. 30.3 GB headroom remains at this quantization.
9Mac Mini M4 Pro 48GB2728bit 48.0 tok/s Fastest evidence path: 8bit · 48.0 tok/s · MLX · EstimatedMLXFits14.3 GB101kEstimated$1,5998bit is the current best practical quantization. 48.0 tok/s is estimated from nearby benchmark coverage. 14.3 GB headroom remains at this quantization.
10MacBook Pro M4 Pro 48GB 14-inch2728bit 48.0 tok/s Fastest evidence path: 8bit · 48.0 tok/s · MLX · EstimatedMLXFits14.3 GB101kEstimated$2,4998bit is the current best practical quantization. 48.0 tok/s is estimated from nearby benchmark coverage. 14.3 GB headroom remains at this quantization.
11MacBook Pro M4 Pro 48GB 16-inch2728bit 48.0 tok/s Fastest evidence path: 8bit · 48.0 tok/s · MLX · EstimatedMLXFits14.3 GB101kEstimated$2,9998bit is the current best practical quantization. 48.0 tok/s is estimated from nearby benchmark coverage. 14.3 GB headroom remains at this quantization.
12Mac Studio M4 Max 36GB260Q6_K 48.0 tok/s Fastest evidence path: Q6_K · 48.0 tok/s · MLX · EstimatedMLXFits8.1 GB44kEstimated$1,999Q6_K is the current best practical quantization. 48.0 tok/s is estimated from nearby benchmark coverage. 8.1 GB headroom remains at this quantization.
13MacBook Pro M4 Max 36GB 14-inch260Q6_K 48.0 tok/s Fastest evidence path: Q6_K · 48.0 tok/s · MLX · EstimatedMLXFits8.1 GB44kEstimated$2,999Q6_K is the current best practical quantization. 48.0 tok/s is estimated from nearby benchmark coverage. 8.1 GB headroom remains at this quantization.
14MacBook Pro M4 Max 36GB 16-inch260Q6_K 48.0 tok/s Fastest evidence path: Q6_K · 48.0 tok/s · MLX · EstimatedMLXFits8.1 GB44kEstimated$3,499Q6_K is the current best practical quantization. 48.0 tok/s is estimated from nearby benchmark coverage. 8.1 GB headroom remains at this quantization.
15Mac Mini M4 32GB2586bit 48.0 tok/s Fastest evidence path: 6bit · 48.0 tok/s · MLX · EstimatedMLXFits6.4 GB30kEstimated$7996bit is the current best practical quantization. 48.0 tok/s is estimated from nearby benchmark coverage. 6.4 GB headroom remains at this quantization.
16MacBook Air M4 32GB 13-inch2586bit 48.0 tok/s Fastest evidence path: 6bit · 48.0 tok/s · MLX · EstimatedMLXFits6.4 GB30kEstimated$1,4996bit is the current best practical quantization. 48.0 tok/s is estimated from nearby benchmark coverage. 6.4 GB headroom remains at this quantization.
17MacBook Air M4 32GB 15-inch2586bit 48.0 tok/s Fastest evidence path: 6bit · 48.0 tok/s · MLX · EstimatedMLXFits6.4 GB30kEstimated$1,6996bit is the current best practical quantization. 48.0 tok/s is estimated from nearby benchmark coverage. 6.4 GB headroom remains at this quantization.
18Mac Studio M4 Max 48GB2488bit 42.0 tok/s Fastest evidence path: 8bit · 42.0 tok/s · MLX · EstimatedMLXFits14.3 GB101kEstimated$2,4998bit is the current best practical quantization. 42.0 tok/s is estimated from nearby benchmark coverage. 14.3 GB headroom remains at this quantization.
19MacBook Pro M4 Max 48GB 14-inch2488bit 42.0 tok/s Fastest evidence path: 8bit · 42.0 tok/s · MLX · EstimatedMLXFits14.3 GB101kEstimated$3,4998bit is the current best practical quantization. 42.0 tok/s is estimated from nearby benchmark coverage. 14.3 GB headroom remains at this quantization.
20MacBook Pro M4 Max 48GB 16-inch2488bit 42.0 tok/s Fastest evidence path: 8bit · 42.0 tok/s · MLX · EstimatedMLXFits14.3 GB101kEstimated$3,9998bit is the current best practical quantization. 42.0 tok/s is estimated from nearby benchmark coverage. 14.3 GB headroom remains at this quantization.
21Mac Mini M4 16GB2193bit 48.0 tok/s Fastest evidence path: 3bit · 48.0 tok/s · MLX · EstimatedMLXFits2.7 GB11kEstimated$4993bit is the current best practical quantization. 48.0 tok/s is estimated from nearby benchmark coverage. 2.7 GB headroom remains at this quantization.
22MacBook Air M4 16GB 13-inch2193bit 48.0 tok/s Fastest evidence path: 3bit · 48.0 tok/s · MLX · EstimatedMLXFits2.7 GB11kEstimated$1,0993bit is the current best practical quantization. 48.0 tok/s is estimated from nearby benchmark coverage. 2.7 GB headroom remains at this quantization.
23MacBook Air M4 16GB 15-inch2193bit 48.0 tok/s Fastest evidence path: 3bit · 48.0 tok/s · MLX · EstimatedMLXFits2.7 GB11kEstimated$1,2993bit is the current best practical quantization. 48.0 tok/s is estimated from nearby benchmark coverage. 2.7 GB headroom remains at this quantization.
24Mac Mini M4 24GB194Q4_K_M 32.0 tok/s Fastest evidence path: Q4_K_M · 32.0 tok/s · Ollama · Trusted referenceOllamaFits4.2 GB16kTrusted reference$599Q4_K_M is the current best practical quantization. 32.0 tok/s is backed by direct benchmark coverage. 4.2 GB headroom remains at this quantization.
25MacBook Air M4 24GB 13-inch194Q4_K_M 32.0 tok/s Fastest evidence path: Q4_K_M · 32.0 tok/s · Ollama · Trusted referenceOllamaFits4.2 GB16kTrusted reference$1,299Q4_K_M is the current best practical quantization. 32.0 tok/s is backed by direct benchmark coverage. 4.2 GB headroom remains at this quantization.
26Mac Mini M4 Pro 24GB194Q4_K_M 32.0 tok/s Fastest evidence path: Q4_K_M · 32.0 tok/s · Ollama · Trusted referenceOllamaFits4.2 GB16kTrusted reference$1,399Q4_K_M is the current best practical quantization. 32.0 tok/s is backed by direct benchmark coverage. 4.2 GB headroom remains at this quantization.
27MacBook Air M4 24GB 15-inch194Q4_K_M 32.0 tok/s Fastest evidence path: Q4_K_M · 32.0 tok/s · Ollama · Trusted referenceOllamaFits4.2 GB16kTrusted reference$1,499Q4_K_M is the current best practical quantization. 32.0 tok/s is backed by direct benchmark coverage. 4.2 GB headroom remains at this quantization.
28MacBook Pro M4 Pro 24GB 14-inch194Q4_K_M 32.0 tok/s Fastest evidence path: Q4_K_M · 32.0 tok/s · Ollama · Trusted referenceOllamaFits4.2 GB16kTrusted reference$1,999Q4_K_M is the current best practical quantization. 32.0 tok/s is backed by direct benchmark coverage. 4.2 GB headroom remains at this quantization.
29MacBook Pro M4 Pro 24GB 16-inch194Q4_K_M 32.0 tok/s Fastest evidence path: Q4_K_M · 32.0 tok/s · Ollama · Trusted referenceOllamaFits4.2 GB16kTrusted reference$2,499Q4_K_M is the current best practical quantization. 32.0 tok/s is backed by direct benchmark coverage. 4.2 GB headroom remains at this quantization.

Silicon Score › Models › Qwen3.6-35B-A3B

Qwen3.6-35B-A3B — ranking first, raw rows below

Start with the ranked Mac table above. Use the rest of this page to inspect raw Apple Silicon coverage and model metadata.

Quantizations observed: Q4_K - Medium

4Benchmark rows
4Chip tiers covered
55.0Fastest avg tok/s (M5 Max (128 GB))
—Minimum RAM observed

Quick take

Fastest published result is 55.0 tok/s on M5 Max (128 GB) at Q4_K - Medium. Published runtimes include MLX, Ollama. Start with Rankings for the decision, then use the raw rows below to audit the evidence.

Based on 4 external benchmarks; no lab runs yet.

Published runtimes: MLX, Ollama.

Need the best Mac for this model? Use BuyNeed a setup-first answer? Use RunChecking whether it fits? Use FitBrowse Macs by exact hardwareNeed the full audit trail? Use BenchComparing against rented GPUs? Use AI Datacenter Index

Catalog record

35BTotal params
3BActive params
262,144Context window
2026-04-15Release date

What this model is, and what Apple Silicon users are actually seeing

Official model cards tell you what the model is for and which software stacks it targets. Field reality below shows how much Apple Silicon evidence we have so far.

Official brief

Agentic Coding: the model now handles frontend workflows and repository-level reasoning with greater fluency and precision.

Official source  ·  Raw model card

agentscodingreasoningvisual-understanding

Runtime support mentioned

vLLMSGLangTransformersKTransformers

Official specs

  • Type: Causal Language Model with Vision Encoder.
  • Scale: 35B in total and 3B activated.
  • Context: 262,144 natively and extensible up to 1,010,000 tokens.
  • Total parameters: 35B in total and 3B activated.
  • Max input: 262,144 natively and extensible up to 1,010,000 tokens.

Official takeaways

  • Agentic Coding: the model now handles frontend workflows and repository-level reasoning with greater fluency and precision.
  • Thinking Preservation: we've introduced a new option to retain reasoning context from historical messages, streamlining iterative development and reducing overhead.

Official model cards describe intent, capabilities, and supported stacks. They do not prove Apple Silicon speed by themselves.

Field reality on Apple Silicon

Qwen3.6-35B-A3B: 40 Apple Silicon field reports; best reported generation ~203.1 tok/s; best reported prompt processing ~3691 tok/s; reported RAM use ~19.2-79GB; seen on M2 MAX 96GB, MacBook Pro M4 PRO 48GB, M5 MAX 128GB; via oMLX, MLX.

4Benchmark rows
40Field reports
15Practitioner signals
Sparse BenchmarksEvidence status

What practitioners keep saying

  • The oMLX page reports Qwen3.6-35B-A3B 6bit on an M3 Max 40-core 64GB Mac with oMLX v0.3.8 and macOS 26.4.1.
  • The context table reports 1024 tokens with 1117 tok/s prompt processing, 79.2 tok/s generation, 0.917s TTFT, and peak memory at 29.0GB.
  • The context table reports 4096 tokens with 1428 tok/s prompt processing, 77.0 tok/s generation, 2.869s TTFT, and peak memory at 29.8GB.

Apple Silicon field sources

  • oMLX community benchmarks

    2026-05-05 · Mac Studio M3 Ultra 256GB · oMLX v0.3.8

    A same-day oMLX row shows Qwen3.6-35B-A3B 6-bit running on an M3 Ultra 256GB Mac Studio, filling a tier between laptop Max/Pro rows and the existing 512GB M3 Ultra long-context signal.

  • oMLX community benchmarks

    2026-05-05 · M5 Max 40-core GPU, 128GB unified memory · oMLX

    A same-day oMLX sweep shows Qwen3.6-35B-A3B-oQ4 running fast on M5 Max 128GB at 1k and 4k context, adding a current exact-row reproduction target for the owned high-memory Mac.

  • oMLX community benchmarks

    2026-05-05 · Mac Studio M3 Ultra 512GB · oMLX v0.3.8

    A same-day oMLX report shows Qwen3.6-35B-A3B 8-bit running on an M3 Ultra 512GB Mac Studio at both 4k and 128k contexts, adding an extreme-memory Apple Silicon reproduction target for sparse Qwen3.6.

  • oMLX community benchmarks

    2026-05-05 · M1 Max 24-core GPU, 64GB unified memory · oMLX

    A same-day oMLX sweep shows Qwen3.6-35B-A3B running efficiently on an M1 Max 64GB setup, making older high-memory Max machines more relevant for sparse Qwen3.6 than dense-model fit alone suggests.

  • oMLX community benchmarks

    2026-05-05 · M5 10-core GPU, 32GB unified memory · oMLX

    Same-day oMLX rows show Qwen3.6-35B-A3B-UD fitting at 16k and 32k context on an M5 32GB setup, but single-digit decode makes it a long-context fit caution rather than a throughput recommendation.

10 more Apple Silicon field sources tracked in the research queue.

Runtime mentions in the field

LM StudioMLXoMLX

Hardware mentioned in reports

32GB48GB64GB96GB128GBM1 MaxM3 UltraM4

What would improve confidence

  • Reproduce Field Performance Signal
  • Upgrade To First Party Measurement

Current published coverage

Published chip coverage includes M5 Max (128 GB), M5 Max (64 GB), M4 Max (48 GB), M4 Pro (24 GB). Fastest published row is 55.0 tok/s on M5 Max (128 GB) at Q4_K - Medium.

M5 Max (128 GB)M5 Max (64 GB)M4 Max (48 GB)M4 Pro (24 GB)

Related Qwen3.6-35B-A3B models with published pages: Qwen3.6-27B

Raw benchmark rows for Qwen3.6-35B-A3B

Rows stay below the ranking because this page is answer-first. Use them to inspect exact chips, quantizations, runtimes, and sources.

ChipQuantRAM req.ContextAvg tok/sPrompt tok/sRuntimeSource
M5 Max (128 GB)Q4_K - Medium——55.0 tok/s—MLXref
M5 Max (64 GB)Q4_K - Medium——48.0 tok/s—Ollamaref
M4 Max (48 GB)Q4_K - Medium——42.0 tok/s—MLXref
M4 Pro (24 GB)Q4_K - Medium——32.0 tok/s—Ollamaref

Best Macs for Qwen3.6-35B-A3B

Ordered by fastest published tok/s on the chip family in each Mac. Click through for the full machine page.

MacBook Pro M5 Max 128GB 16-inch — 55.0 tok/sMacBook Pro M4 Max 36GB 14-inch — 42.0 tok/sMacBook Pro M4 Max 48GB 14-inch — 42.0 tok/sMacBook Pro M4 Max 36GB 16-inch — 42.0 tok/sMacBook Pro M4 Max 48GB 16-inch — 42.0 tok/sMacBook Pro M4 Max 64GB 16-inch — 42.0 tok/s

Chips with published results for Qwen3.6-35B-A3B

M5 Max (128 GB)M5 Max (64 GB)M4 Max (48 GB)M4 Pro (24 GB)

Data

benchmarks.json — full dataset  ·  models.json — model summaries  ·  benchmarks.csv — CSV export

See all models →