Run DeepSeek V4.1 Flash on M3 Ultra

DeepSeek V4.1 Flash runs at 20 tok/s (20.23 as recorded) on a 512 GB M3 Ultra, a community run reported by github.com (MLX, oQ4e, 4K context). It needs about 434 GB of unified memory. It is DeepSeek's 552B mixture-of-experts model (16B active), released under the MIT licence.

Runtime caveat

Runs only in oMLX and Rapid-MLX (their own deepseek_v41 code); upstream llama.cpp support is still open (#28696, plus a competing draft #30817), and stock mlx-lm refuses its Engram checkpoints.

Speed
community20
tok/s
First Token
—
seconds
RAM Needed
434
GB minimum
Engine
MLX
of this figure
Want to run DeepSeek V4.1 Flash 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 secondcommunity20 tok/s
Provenancecommunity run · github.com
QuantizationoQ4e
EngineMLX
Context4K
Figure dated2026-09-11
Minimum RAM434 GB
Parameters552B MoE (16B active)
Chip bandwidth819 GB/s

Every figure for DeepSeek V4.1 Flash on the M3 Ultra

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 (434 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
community38.7 tok/soQ4eMLX512 GB16K promptgithub.comoMLX 0.7.2 release notes (fused decode kernels, PR #4722): the 1017 -mtp checkpoint at oQ4e with MTP off, 32.8 -> 38.7 tok/s, measured at a 16K-token prompt. A long-prompt figure is a different workload from the index's plain-decode rows, so it is shown but does not anchor; the model's estimates still follow the 11 Sep 4K run (20.23).; not used to anchor estimates
community20.23 tok/soQ4eMLX512 GB4Kgithub.comBase V4.1 checkpoint (the 1017 snapshot has the same architecture). oMLX 0.7.0.dev2, -mtp checkpoint run with MTP off, Engram tables in RAM. 19.67 tok/s at 64K.
community6.53 tok/s4/8-bit mixedMLX512 GB8Khuggingface.coPipeNetwork custom V4.1 runtime on mlx 0.32.2 / mlx-lm 0.31.3 (immature loader); 427.6 GB.; not used to anchor estimates
Provenance key
sourced
Published figure — vendor or press review
estimated
community
Linked individual run — blog, forum or submission

Setup: run DeepSeek V4.1 Flash on a Mac

Ollama — the index has not verified an Ollama tag for DeepSeek V4.1 Flash, 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 → · model card →

The 20 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.

DeepSeek V4.1 Flash on other Apple Silicon chips

ChipSpeedQuantEngineBandwidth
M5 Ultraestimated29 tok/soQ4eMLX1228 GB/s
M5 Maxestimated12 tok/s6.6 bpw—614 GB/s

About DeepSeek V4.1 Flash

DeepSeek's MIT-licensed 552B causal encoder-decoder MoE (16B active per decoded token) with image input and 1M context; the 17 Oct 2026 '1017' snapshot lifts DeepSWE v1.1 from 74.2 to 77.0 (Mini-SWE-V2 harness). It runs on a 512 GB Mac Studio through oMLX or Rapid-MLX: the oQ4e MLX build is 432 GB, and oMLX 0.7.2's fused kernels reach 38.7 tok/s on an M3 Ultra (oQ4e, MTP off, 16K prompt; community), against ~20 in an earlier dev build. A 256 GB Mac runs the 1017 checkpoint only below 3-bit or with experts offloaded to SSD. Upstream llama.cpp support is still open (#28696, plus a competing draft #30817), and Ollama is cloud-only.

MIT licence · 552B MoE · 1000K context · LLMCheck capability 50/50 · see its leaderboard rank →

System requirements

Macs with the M3 Ultra

🛒 Get a Mac that runs DeepSeek V4.1 Flash

DeepSeek V4.1 Flash needs about 434 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.

Compare More Models

See how DeepSeek V4.1 Flash stacks up against other models on your specific Mac hardware.

Open Compare Tool Full Leaderboard