Qwen 3 8B runs at 149 tok/s on a 256 GB M5 Ultra, a published figure reported by vsesmart.cz (llama.cpp, 4-bit, empty context, 128 generated). It needs about 6 GB of unified memory. It is Alibaba's 8B dense model, released under the Apache 2.0 licence.
Want to run Qwen 3 8B faster — or step up to bigger models? Find your Mac →This is a published figure: 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 | sourced149 tok/s |
| Provenance | published figure · vsesmart.cz |
| Quantization | 4-bit |
| Engine | llama.cpp |
| Prompt processing (prefill) | sourced4205 tok/s at 4K context |
| Context | empty context, 128 generated |
| Figure dated | 2026-10-26 |
| Minimum RAM | 6 GB |
| Parameters | 8B |
| Chip bandwidth | 1228 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 (6 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 |
|---|---|---|---|---|---|---|
| sourced176 tok/s | 4-bit | MLX | 256 GB | empty context, 128 generated | vsesmart.cz | Vse Smart review: Mac Studio M5 Ultra (80-core GPU, 256 GB), macOS 27.0.1, measured 24 Oct; MLX 0.32 with mlx-lm, point version not stated. 4-bit MLX; the review does not name the repo. |
| sourced149 tok/s | 4-bit | llama.cpp | 256 GB | empty context, 128 generated | vsesmart.cz | Vse Smart review: Mac Studio M5 Ultra (80-core GPU, 256 GB), macOS 27.0.1, measured 24 Oct; llama.cpp with every layer on the GPU, build not stated. 4-bit; the review does not name the GGUF type. |
Measured figures only, each from its source’s own session; the index does not estimate long-context or prefill speed, so a missing cell is —. Each @ column is plain decode with that many tokens already in the prompt, from the same session as the row’s short-context figure: decode slows as the KV cache it reads grows. Prefill is prompt processing, the tokens per second a runtime reads a prompt at; with long agent contexts it sets how long you wait before the first token.
| Run | Decode (row) | @16K | Prefill |
|---|---|---|---|
| MLX · 4-bit · 256 GB Mac · vsesmart.cz | sourced176 tok/s | sourced121 tok/s | sourced6360 tok/s |
| llama.cpp · 4-bit · 256 GB Mac · vsesmart.cz | sourced149 tok/s | — | sourced4205 tok/s |
Ollama — install it (brew install ollama or the app from ollama.com), then pull the model by its tag, checked on ollama.com on 2026-11-14:
MLX — pick a build from Hugging Face’s MLX listings, then run it with its repo id:
Downloads: Ollama page → · Hugging Face model card →
The 149 tok/s figure above came from llama.cpp 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.
Alibaba's 8B model, ~91 tok/s on M5 Max. Apache 2.0 license, LLMCheck capability 15/50.
Apache 2.0 licence · 8B · 128K context · LLMCheck capability 15/50 · see its leaderboard rank →
Estimated, not measured: the index’s minimum for Qwen 3 8B (6 GB — the weights at the quant that fits, plus ~2 GB) plus the KV cache for that many tokens, computed from its config.json. GQA on all 36 layers (8 kv heads x 128). A 16-bit KV cache is the default in llama.cpp, Ollama, LM Studio and mlx-lm; a q8 cache halves the KV part. Coding agents hold 64K tokens or more. Runtimes also add a buffer that grows with context, so leave a few GB spare. macOS lets the GPU use only part of a Mac’s memory by default (figures for an M5 Ultra Mac, which runs macOS 26 or later): “fits” means within the GPU memory a Mac of that tier can use with the limit raised, a tier tagged raise GPU limit fits only after sudo sysctl iogpu.wired_limit_mb raises it (the tooltip has the exact command), and the last column is the smallest Mac memory size whose default limit holds it (the GPU memory limit).
Qwen 3 8B needs about 6 GB of unified memory. These Apple Silicon Macs have enough to run it.
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 8B stacks up against other models on your specific Mac hardware.
Open Compare Tool Full Leaderboard