Qwen 3.6-35B-A3B runs at 106 tok/s (106.4 as recorded) on a 512 GB M3 Ultra, a community run reported by github.com (MLX, 4-bit, 512-token prompt, 128 generated). It needs about 24 GB of unified memory. It is Alibaba's 35B mixture-of-experts model (3B active), released under the Apache 2.0 licence.
Want to run Qwen 3.6-35B-A3B faster — or step up to bigger models? Find your Mac →This is a community run: a real run, reported by its source and linked below. The index records it as given and never adjusts it.
| Metric | Value |
|---|---|
| Tokens per second | community106 tok/s |
| Provenance | community run · github.com |
| Quantization | 4-bit |
| Engine | MLX |
| Context | 512-token prompt, 128 generated |
| Figure dated | 2026-09-20 |
| Minimum RAM | 24 GB |
| Parameters | 35B MoE (3B active) |
| Chip bandwidth | 819 GB/s |
Each row as the index holds it, with its source. “Mac RAM” is the memory of the Mac a run used, not the model’s requirement (24 GB). The headline figure follows the reference rule: the lower-median measured plain-decode run at 5 bits or fewer (a run whose quant was not stated counts only when no stated-quant run qualifies), else the estimate; a run marked “not used to anchor estimates” is listed here but never becomes the headline.
| Speed | Quant | Engine | Mac RAM | Context | Source | Note |
|---|---|---|---|---|---|---|
| community106.4 tok/s | 4-bit | MLX | 512 GB | 512-token prompt, 128 generated | github.com | Mac Studio, 80-core GPU, 512 GB, macOS 26.5.1. mlx_lm.benchmark -p 512 -g 128 -n 5 (mlx-lm 0.31.3, MLX 0.32.2): plain decode with no HTTP server or chat template, median of the five trial lines in the repo's engine.json; mlx-community/Qwen3.6-35B-A3B-4bit. The repo's headline test streams the same cell through mlx_lm.server (thinking and prompt cache off): 86.3 tok/s. The source discloses a stuck background process on one of the M3 Ultra's 32 CPU cores during these runs, which can only have slowed it. |
| community81.2 tok/s | Q8_0 | llama.cpp | 512 GB | TG128 from empty context | llamaperf.com | llama.cpp b49650adb, flash attention on, F16 KV; anonymous llamaperf submission.; not used to anchor estimates |
Ollama — install it (brew install ollama or the app from ollama.com), then pull the model by its tag, checked on ollama.com on 2026-10-17:
MLX — pick a build from Hugging Face’s MLX listings, then run it with its repo id:
Downloads: Search the Ollama library → · model card →
The 106 tok/s figure above came from MLX at 4-bit; other runtimes and quants run at different speeds.
Prefer a GUI? LM Studio downloads GGUF and MLX builds in one click. Step by step: install Ollama on a Mac.
| Chip | Speed | Quant | Engine | Bandwidth |
|---|---|---|---|---|
| M5 Ultra | community153 tok/s | 4-bit | MLX | 1228 GB/s |
| M5 Max | sourced118 tok/s | Q4_K_M | llama.cpp | 614 GB/s |
| M4 Max | estimated105 tok/s | Q4_K_M | MLX | 546 GB/s |
| M1 Ultra | estimated105 tok/s | Q4_K_M | MLX | 800 GB/s |
| M2 Max | estimated97 tok/s | Q4_K_M | MLX | 400 GB/s |
| M4 Pro | estimated86 tok/s | Q4_K_M | Ollama | 273 GB/s |
| M6 | community64 tok/s | 4-bit | MLX | 170 GB/s |
Alibaba's Apache 2.0 35B MoE with 3B active, the most-run local coding model of 2026 (73.4% SWE-bench Verified). About 20 GB at 4-bit, so it fits a 24 GB Mac. On an M5 Max, Liquid AI's published baseline lists 118 tok/s (llama.cpp Q4_K_M) and a community oMLX run measured ~126 (4-bit, MTP off); a 32 GB Mac mini M6 runs it at ~64 (community).
Apache 2.0 licence · 35B MoE · 262K context · LLMCheck capability 38/50 · see its leaderboard rank →
Qwen 3.6-35B-A3B needs about 24 GB of unified memory. These current Apple Silicon Macs have the headroom to run it comfortably:
As an Amazon Associate, LLMCheck earns from qualifying purchases. These affiliate links cost you nothing extra and help keep the LLMCheck index free.
See how Qwen 3.6-35B-A3B stacks up against other models on your specific Mac hardware.
Open Compare Tool Full Leaderboard