Run Ling-3.0-tiny on M5 Pro

Ling-3.0-tiny runs at 154 tok/s (153.8 as recorded) on a 48 GB M5 Pro, a community run reported by omlx.ai (MLX, oQ4e, 1k context). It needs about 6 GB of unified memory. It is Ant Group's 7.9B mixture-of-experts model (1.3B active), released under the MIT licence.

Speed
community154
tok/s
First Token
community0.223
seconds
RAM Needed
6
GB minimum
Engine
MLX
of this figure
Want to run Ling-3.0-tiny faster — or step up to bigger models? Find your Mac →

The figure

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.

MetricValue
Tokens per secondcommunity154 tok/s
Provenancecommunity run · omlx.ai
QuantizationoQ4e
EngineMLX
Time to first token0.223s
Prompt processing (prefill)community5183 tok/s at 4K context
Context1k
Figure dated2026-11-06
Minimum RAM6 GB
Parameters7.9B MoE (1.3B active)
Chip bandwidth307 GB/s

Every figure for Ling-3.0-tiny on the M5 Pro

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.

SpeedQuantEngineMac RAMContextSourceNote
community153.8 tok/soQ4eMLX48 GB1komlx.ai16-core GPU (307 GB/s, the same as the 20-core part). oMLX v0.7.3 with a TurboQuant KV cache (KV-cache quantization, still plain decode), MTP/DFlash off; Ling-3.0-tiny-oQ4e; 5.3 GB peak. 157.2 tok/s at 4k.
Provenance key
sourced
Published figure — vendor or press review
estimated
community
Linked individual run — blog, forum or submission

Speed by context depth: Ling-3.0-tiny on the M5 Pro

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.

RunDecode (row)@4KPrefillTTFT
MLX · oQ4e · 48 GB Mac · omlx.aicommunity153.8 tok/s
1k
community157.2 tok/scommunity5183 tok/s
at 4K
community0.223 s

Setup: run Ling-3.0-tiny on a Mac

Ollama — the index has not verified an Ollama tag for Ling-3.0-tiny, so it doesn’t print one (a guessed tag fails with “pull model manifest: file does not exist”). Search the Ollama library →

MLX — pick a build from Hugging Face’s MLX listings, then run it with its repo id:

pip install -U mlx-lm
mlx_lm.generate --model <repo-id> --prompt "Hello"

Downloads: Hugging Face repo → · Hugging Face model card →

The 154 tok/s figure above came from MLX at oQ4e; 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.

Ling-3.0-tiny on other Apple Silicon chips

ChipSpeedQuantEngineBandwidth
M5 Maxcommunity174 tok/soQ4eMLX614 GB/s

About Ling-3.0-tiny

Ant Group's MIT-licensed 7.9B MoE with 1.3B active (10 Aug 2026), 128K context. Its only benchmark evidence is vendor-quoted Artificial Analysis indices (Intelligence 25, Agentic 16), so the index holds its capability score below Gemma 4 E4B's. IQ3_XXS is 3.5 GB and Q4_K_M 4.9 GB, so it fits an 8 GB Mac; llama.cpp b10470+ and mlx-lm 0.32 run it, and a community oMLX run measured ~174 tok/s on an M5 Max (4-bit).

MIT licence · 7.9B MoE · 128K context · LLMCheck capability 15/50 · see its leaderboard rank →

Memory by context length

Estimated, not measured: the index’s minimum for Ling-3.0-tiny (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. MLA on 6 of 24 layers (every 4th, per modeling_bailing_moe_v3.py); the rest are Kimi Delta Attention (linear, constant state). 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 Pro 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).

ContextKV cache (16-bit)RAM neededRAM, q8 KVSmallest tier that fitsWithin the default GPU limit
8K~0.1 GB~6.0 GB~6.0 GB16 GB Macsame
32K~0.2 GB~6.2 GB~6.1 GB16 GB Macsame
64K (coding-agent context)~0.5 GB~6.4 GB~6.2 GB16 GB Macsame
128K~0.9 GB~6.9 GB~6.4 GB16 GB Macsame

System requirements

Macs with the M5 Pro

🛒 Get a Mac that runs Ling-3.0-tiny

Ling-3.0-tiny 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.

Compare More Models

See how Ling-3.0-tiny stacks up against other models on your specific Mac hardware.

Open Compare Tool Full Leaderboard