M5 16GB

APPLE · 16GB unified memory · 3 reports

On the M5 16GB, Qwen3.8 27B at 3-bit, the model most run on it, generates a median of 9.0 tokens per second from one plain run (community reports, as of 25 Oct 2026).

As of 25 Oct 2026, the models most run on the M5 16GB, with generation and prompt-processing medians from plain runs (one device, one request, no speculative decoding):

Qwen3.8 27B
9.0gen t/s at 3-bit · one run

Engines people use on it: llama.cpp 1 (reports)

See reported performance Estimate what fits

How to compare these reports →Filter reports by setup →All M5 Macs compared →

Run models on your M5 16GB? Add your numbers.

Your coding agent can run a speed test on your machine and send it here with a free account. Or paste a result you've already posted.

Benchmark with your agent Paste a result

Reported performance on the M5 16GB

These reports mix GPU counts and memory setups, so a model listed here may have needed extra cards or CPU offloading. The calculator estimates which models fit in 16 GB of unified memory.

Date ↓Date ↑Token speed ↓Token speed ↑Prefill ↓Prefill ↑Context ↓Context ↑Ranked

Qwen3.5 9B

M5 16GB · QVex-3.5

generation:: 33-37 tokens/s
quant:: qvex-q4 (MLX)

Reported by the source; GPU count, offloading and concurrent requests can change these figures. Prompt-processing speed is shown only when the source reports it; generation speed cannot tell us time to first token. Check the full setup before comparing.

agentic

User reports Qwen3.5-9B running at about 25 tok/s without speculation and 33-37 tok/s with MTP speculative decoding on a fanless MacBook Air M5 16GB. Setup is a custom MLX engine (QVex-3.5) with a self-made qvex-q4 4-bit imatrix quant, using Qwen's own MTP head as the draft model (about 490 MB). The custom quant is 2.7x closer to bf16 than stock MLX 4-bit (KL 0.026 vs 0.069). A 3-bit build was about 13% faster but used 20-45% more energy per token, and MTP costs 10-25% more energy per token. With a 512-token thinking budget the model solved 24/24 tasks; the user calls this a sanity check rather than a benchmark, tested on one machine only.

Qwen3.5 35B (3B active)

M5 16GB

generation:: 17.0 tokens/s

User reports Qwen3.5 35B A3B running at roughly 17 t/s on a MacBook Pro M5 16GB. Setup uses slotstream; no engine, quant or context length stated. The figure is given as approximate.

Qwen3.8 27B

M5 16GB · llama.cpp · 8,192 ctx

generation:: 9.0 tokens/s
quant:: Q3_xxs
kv:: 8

summarization

User asks whether about 9 t/s is normal for their setup, and is new to local LLMs. User also asks for model recommendations for PDF summaries on 16 GB of RAM.

Get a weekly email of new M5 16GB reports and newly released models that fit it.

Email me new reports

Reported configurations

Individual observations from this page, not expected speeds or a ranking. Select a model to open its report, with the source and offloading conditions. Reported context may be a limit rather than actual input length.

Model / hardware	Quant / engine	Reported context	Generation
Qwen3.5 35B (3B active) M5 16GB	Not reported Engine not reported	Not reported	17.0 tokens/s
Qwen3.8 27B M5 16GB	Q3_xxs llama.cpp	8,192	9.0 tokens/s

Frequently asked

How fast is the M5 16GB for local LLMs?

On the M5 16GB, Qwen3.8 27B at 3-bit, the model most run on it, generates a median of 9.0 tokens per second from one plain run (community reports, as of 25 Oct 2026).