Workspace

Query Apple Silicon local model benchmarks.

Pick a model to rank Macs, or pick a Mac to rank models. Throughput, fit, local cost, and evidence stay attached to the same answer.

Bench

Current read

25 Macs can run Qwen3.5-27B under this query.

0 of these rows have direct benchmark coverage. Current top answer: Mac Mini M4 24GB at — with Q5_K_M.

Rows: 191
Models: 32
Macs: 28
Direct: 0

Catalog current through February 27, 2026

Raw data Models Chips

Query

Choose a model, then rank Macs.

Keep capability, quantization, runtime, fit, and evidence in the same workspace instead of spreading them across separate pages.

Model

Capability

Quant target

Runtime

Form factor

Sort

Results

Best Mac matches for this model

0 of these rows have direct benchmark coverage. Current top answer: Mac Mini M4 24GB at — with Q5_K_M.

Viable rows: 25
Direct benchmark-backed: 0

MatchGenPromptQuantRuntimeHeadroomContextLocal costOpen

#1Fit estimateInsufficient data

Mac Mini M4 24GB

M4 · 24GB unified memory · $599 · Desktop

Gen—

Prompt—

QuantQ5_K_M

RuntimeBest available

Headroom2.8 GB

Context11k

Local costHold for speed

Open Reference

Capability

27B dense · tuned for coding

Evidence

Fit is computed; speed is still unmeasured Fit is computed from model size and KV-cache math, but speed still needs direct coverage.

Coverage

No direct benchmark rows yet

Speed is estimated, so this cost read is provisional.

Quant ladder and fit detail

QuantQualityHeadroomContextSpeedRuntimeEvidence

Q5_K_MHigh quality2.8 GB11k—Best availableFit estimate

Q4_K_MBalanced quality6.5 GB27k—Best availableFit estimate

#2Fit estimateInsufficient data

Mac Mini M4 32GB

M4 · 32GB unified memory · $799 · Desktop

Gen—

Prompt—

QuantQ8_0

RuntimeBest available

Headroom3.0 GB

Context12k

Local costHold for speed

Open Reference

Capability

27B dense · tuned for coding

Evidence

Fit is computed; speed is still unmeasured Fit is computed from model size and KV-cache math, but speed still needs direct coverage.

Coverage

No direct benchmark rows yet

Speed is estimated, so this cost read is provisional.

Quant ladder and fit detail

QuantQualityHeadroomContextSpeedRuntimeEvidence

Q8_0Reference quality3.0 GB12k—Best availableFit estimate

Q6_KHigh quality7.9 GB32k—Best availableFit estimate

Q5_K_MHigh quality10.8 GB44k—Best availableFit estimate

Q4_K_MBalanced quality14.5 GB60k—Best availableFit estimate

#3Fit estimateInsufficient data

Mac Mini M4 Pro 24GB

M4-PRO · 24GB unified memory · $1,399 · Desktop

Gen—

Prompt—

QuantQ5_K_M

RuntimeBest available

Headroom2.8 GB

Context11k

Local costHold for speed

Open Reference

Capability

27B dense · tuned for coding

Evidence

Fit is computed; speed is still unmeasured Fit is computed from model size and KV-cache math, but speed still needs direct coverage.

Coverage

No direct benchmark rows yet

Speed is estimated, so this cost read is provisional.

Quant ladder and fit detail

QuantQualityHeadroomContextSpeedRuntimeEvidence

Q5_K_MHigh quality2.8 GB11k—Best availableFit estimate

Q4_K_MBalanced quality6.5 GB27k—Best availableFit estimate

#4Fit estimateInsufficient data

Mac Mini M4 Pro 48GB

M4-PRO · 48GB unified memory · $1,599 · Desktop

Gen—

Prompt—

QuantQ8_0

RuntimeBest available

Headroom19.0 GB

Context78k

Local costHold for speed

Open Reference

Capability

27B dense · tuned for coding

Evidence

Fit is computed; speed is still unmeasured Fit is computed from model size and KV-cache math, but speed still needs direct coverage.

Coverage

No direct benchmark rows yet

Speed is estimated, so this cost read is provisional.

Quant ladder and fit detail

QuantQualityHeadroomContextSpeedRuntimeEvidence

Q8_0Reference quality19.0 GB78k—Best availableFit estimate

Q6_KHigh quality23.9 GB98k—Best availableFit estimate

Q5_K_MHigh quality26.8 GB110k—Best availableFit estimate

Q4_K_MBalanced quality30.5 GB125k—Best availableFit estimate

#5Fit estimateInsufficient data

Mac Pro M2 Ultra 192GB

M2-ULTRA · 192GB unified memory · $6,999 · Desktop

Gen—

Prompt—

QuantQ8_0

RuntimeBest available

Headroom163.0 GB

Context262k

Local costHold for speed

Open Reference

Capability

27B dense · tuned for coding

Evidence

Fit is computed; speed is still unmeasured Fit is computed from model size and KV-cache math, but speed still needs direct coverage.

Coverage

No direct benchmark rows yet

Speed is estimated, so this cost read is provisional.

Quant ladder and fit detail

QuantQualityHeadroomContextSpeedRuntimeEvidence

Q8_0Reference quality163.0 GB262k—Best availableFit estimate

Q6_KHigh quality167.9 GB262k—Best availableFit estimate

Q5_K_MHigh quality170.8 GB262k—Best availableFit estimate

Q4_K_MBalanced quality174.5 GB262k—Best availableFit estimate

#6Fit estimateInsufficient data

Mac Studio M3 Ultra 256GB

M3-ULTRA · 256GB unified memory · $7,499 · Desktop

Gen—

Prompt—

QuantQ8_0

RuntimeBest available

Headroom227.0 GB

Context262k

Local costHold for speed

Open Reference

Capability

27B dense · tuned for coding

Evidence

Fit is computed; speed is still unmeasured Fit is computed from model size and KV-cache math, but speed still needs direct coverage.

Coverage

No direct benchmark rows yet

Speed is estimated, so this cost read is provisional.

Quant ladder and fit detail

QuantQualityHeadroomContextSpeedRuntimeEvidence

Q8_0Reference quality227.0 GB262k—Best availableFit estimate

Q6_KHigh quality231.9 GB262k—Best availableFit estimate

Q5_K_MHigh quality234.8 GB262k—Best availableFit estimate

Q4_K_MBalanced quality238.5 GB262k—Best availableFit estimate

#7Fit estimateInsufficient data

Mac Studio M3 Ultra 96GB

M3-ULTRA · 96GB unified memory · $3,999 · Desktop

Gen—

Prompt—

QuantQ8_0

RuntimeBest available

Headroom67.0 GB

Context262k

Local costHold for speed

Open Reference

Capability

27B dense · tuned for coding

Evidence

Fit is computed; speed is still unmeasured Fit is computed from model size and KV-cache math, but speed still needs direct coverage.

Coverage

No direct benchmark rows yet

Speed is estimated, so this cost read is provisional.

Quant ladder and fit detail

QuantQualityHeadroomContextSpeedRuntimeEvidence

Q8_0Reference quality67.0 GB262k—Best availableFit estimate

Q6_KHigh quality71.9 GB262k—Best availableFit estimate

Q5_K_MHigh quality74.8 GB262k—Best availableFit estimate

Q4_K_MBalanced quality78.5 GB262k—Best availableFit estimate

#8Fit estimateInsufficient data

Mac Studio M4 Max 128GB

M4-MAX · 128GB unified memory · $4,499 · Desktop

Gen—

Prompt—

QuantQ8_0

RuntimeBest available

Headroom99.0 GB

Context262k

Local costHold for speed

Open Reference

Capability

27B dense · tuned for coding

Evidence

Fit is computed; speed is still unmeasured Fit is computed from model size and KV-cache math, but speed still needs direct coverage.

Coverage

No direct benchmark rows yet

Speed is estimated, so this cost read is provisional.

Quant ladder and fit detail

QuantQualityHeadroomContextSpeedRuntimeEvidence

Q8_0Reference quality99.0 GB262k—Best availableFit estimate

Q6_KHigh quality103.9 GB262k—Best availableFit estimate

Q5_K_MHigh quality106.8 GB262k—Best availableFit estimate

Q4_K_MBalanced quality110.5 GB262k—Best availableFit estimate