M1 Pro 32GB

APPLE · 32GB unified memory · 3 reports

On the M1 Pro 32GB, LFM2.5 24B · 2.5B active at 4-bit, the model most run on it, generates a median of 52.0 tokens per second from one plain run (community reports, as of 29 Oct 2026).

As of 29 Oct 2026, the models most run on the M1 Pro 32GB, with generation and prompt-processing medians from plain runs (one device, one request, no speculative decoding):

LFM2.5 24B · 2.5B active
52.0gen t/s at 4-bit · one run
Qwen3.6 35B · 3B active
25.4gen t/s at 4-bit · one run
PP 38.6 t/s · 1 run

Engines people use on it: llama.cpp 2 (reports)

See reported performance Estimate what fits

How to compare these reports →Filter reports by setup →All M1 Macs compared →

Run models on your M1 Pro 32GB? 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 M1 Pro 32GB

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 32 GB of unified memory.

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

LFM2.5 24B (2.45B active) A2.5B

M1 Pro 32GB · llama.cpp · 65,536 ctx

generation:: 52.0 tokens/s
quant:: Q4_K_M (GGUF)

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.

tool-useagentic

User compares LFM2.5-24B-A2B against Qwen3.6-35B-A3B on 75 multi-step tool-use tasks on an M1 Pro 32GB, with LFM2.5 solving 37/75 (49%) and Qwen 68/75 (91%). Both ran in llama.cpp b10964 with --jinja at 64k context, LFM2.5 at Q4_K_M (14.5 GB) and Qwen at Q4_K_M (22.3 GB), thinking on. LFM2.5 was faster per token (~52 vs 34 tok/s) but used about twice the tokens per solved task, 418 s per correct result vs Qwen's 328. In at least 23 of its 38 failures it reported the task as done. The user notes this is 4-bit llama.cpp rather than Liquid's vLLM setup, and that the two tie on BFCLv4.

Qwen3.6 35B (3B active)

M1 Pro 32GB · llama.cpp · 32,768 ctx

generation:: 25.4 tokens/s
prompt processing (prefill):: 38.6 tokens/s
quant:: Q4_K_S (GGUF)
kv:: q8_0/q4_0

Reported by the source; GPU count, offloading and concurrent requests can change these figures. Prompt-processing speed is input throughput, not time to first token. Check the full setup before comparing.

codingagentictool-uselong-context

User reports Qwen3.6-35B-A3B at 25.4 tok/s decode and 38.6 tok/s prefill on a MacBook Pro M1 Pro 32GB. Setup is llama.cpp build b9466 with Q4_K_S GGUF and q8_0/q4_0 KV cache at 32K context, single sequence, all layers on the Metal GPU. User compares the MoE against dense Qwen3.6-27B on the same host: the dense model decodes at 5.3 tok/s and prefills at 12.9 tok/s, uses 17.9 GB wired versus 22.4 GB, and swaps ~1 GB versus ~7.6 GB. Prefill is the bottleneck; cold TTFT on a 7.5K prompt is 92.8 s versus 0.2 s warm. Tool-calling was clean in 10/10 episodes.

GLM-5.3 320B (18B active) Flash

M1 Pro 32GB · apus · 1,048,576 ctx

generation:: 0.1 tokens/s
quant:: FP8-E4M3 (FP8)

User reports GLM-5.3-Flash at ~0.085 tok/s decode (~11-12 s/token) on a MacBook Pro M1 Pro 32 GB. Setup is the apus engine with the official FP8-E4M3 checkpoint, 2 GiB expert cache, --tiered streaming of routed experts from NVMe, MTP speculative decoding on by default (draft depth 3). Prefill was ~76 s for a 26-token prompt, model load ~7 s, peak footprint ~22.3 GiB, weights ~306 GiB, context up to 1,048,576 tokens. The engine reads ~7.9 GiB of experts per generated token cold; the author notes the wall at 32 GB is memory capacity, not disk.

Get a weekly email of new M1 Pro 32GB 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
LFM2.5 24B (2.45B active) A2.5B M1 Pro 32GB	Q4_K_M llama.cpp	65,536	52.0 tokens/s
Qwen3.6 35B (3B active) M1 Pro 32GB	Q4_K_S llama.cpp	32,768	25.4 tokens/s
GLM-5.3 320B (18B active) Flash M1 Pro 32GB	FP8-E4M3 apus	1,048,576	0.1 tokens/s

Frequently asked

How fast is the M1 Pro 32GB for local LLMs?

On the M1 Pro 32GB, LFM2.5 24B · 2.5B active at 4-bit, the model most run on it, generates a median of 52.0 tokens per second from one plain run (community reports, as of 29 Oct 2026). Also reported: Qwen3.6 35B · 3B active at 4-bit (25.4 t/s, one plain run).

What is the biggest model reported running on the M1 Pro 32GB?

The largest model with a plain generation run on the M1 Pro 32GB is Qwen3.6 35B · 3B active at 4-bit (25.4 t/s, one plain run), as of 29 Oct 2026. Plain means one device, one request, no speculative decoding, so a run split across several cards does not count.