models · deepseek-ai

DeepSeek-V4-Flash

DeepSeek-V4 efficient inference variant. Optimized for low-cost, high-throughput deployment. 284B parameters, 1M context window.

Compare Verify

What this registry has checked

Quarantine. We hold the publisher’s own published hash for these weights and have recorded it against this entry. We have not downloaded and hashed the weights ourselves, so the hash is the publisher’s claim, not our measurement.

Weight file SHA-256	94a4467bac50616db8e2fdadc2bb1d49aa98c11b97d51e492d4d83badf82df36
Method	hf-published-sha256

Re-check this yourself → Source → Full record →

What is recorded

Publisher	deepseek-ai
Origin	Recorded as a covered nation under 10 U.S.C. § 4872(f) — the PRC, Russia, Iran or North Korea. The registry did not record which.
Licence	MIT
Parameters	290,944,616,402 (290.9B)
Context window	1M
Modality	text->text
Architecture	deepseek_v4
Layers	43
Hidden size	4096
Attention	GQA (64 heads, 1 KV)
Native precision	I8
Pulls recorded here	670,019
Downloads reported upstream	1,750,419

Reported scores

GPQA Diamond	89.4
IFBench	79.2
Humanity's Last Exam	34.8

Reported by a third party and recorded here, not re-run by us. Source: artificialanalysis.ai/api/v2.

Measured on real hardware

None of these runs are ours. Each figure is the reporter’s, linked to their own evidence, under their own trust label. The number shown is single-stream decode — one user, concurrency 1 — because aggregate throughput across many users is a different quantity and reads as far faster than anyone will see.

Reported to local-ai-registry

Hardware	Engine	Build measured	Decode, 1 stream	Label	Reported by
Radeon AI PRO R9700 ×4 devices	hipfire	MQ2R	54.3 tok/s @ 2,052	candidate	localmaxxing
RTX PRO 6000 Blackwell ×4 devices	sglang	safetensors FP8	37.6 tok/s @ 8,192	candidate	0xsero
Apple M5 Max 128GB	llama.cpp	GGUF UD_Q2_K_XL	32.2 tok/s @ 8,256	candidate	exo-postgres
Apple M5 Max 128GB	llama.cpp	IQ2_XXS (asym MoE)	26.3 tok/s @ 32,768	candidate	localmaxxing
Apple M5 Max 128GB	llama.cpp	GGUF IQ2XXS	25.5 tok/s @ 8,256	candidate	exo-postgres
Apple M3 Ultra 96GB 80-core GPU	llama.cpp	GGUF IQ2XXS	22.5 tok/s @ 8,256	candidate	exo-postgres
Apple M4 Max 128GB	llama.cpp	GGUF IQ2XXS	20.1 tok/s @ 8,256	candidate	exo-postgres
Apple M4 Max 128GB	llama.cpp	GGUF IQ2XXS	20.1 tok/s @ 8,256	candidate	exo-postgres
Apple M5 Max 128GB	llama.cpp	GGUF UD-Q2_K_XL	13.0 tok/s @ 8,256	candidate	exo-postgres
Apple M5 Max 128GB	llama.cpp	GGUF IQ2XXS	12.7 tok/s @ 8,256	candidate	exo-postgres

From local-ai-registry (MIT), commit 124959d: 82 runs across 13 machines, 0 validated — model revision and runtime pinned, launch accepted — and 82 candidate, their label for useful evidence without a reproducible-launch promise. 16 more not listed.

Measured by local.ai

Hardware	Build measured	Decode @ 8K	Memory @ 8K	Decode mode
MacBook Pro M5 Max 128GB · 40-core GPU	UD-Q2_K_XL	29.9 tok/s	99.1 GB	ordinary
NVIDIA DGX Spark 128GB	UD-Q2_K_XL	17.5 tok/s	98.8 GB	ordinary
Mac Studio M3 Ultra 96GB · 80-core GPU	UD-Q2_K_XL	7.1 tok/s	99.1 GB	ordinary

Measured by local.ai, not by us: 3 runs. Each names its engine pinned by image digest and the exact serve command, in the full record.

Elsewhere in the registry

All models → Browse with filters → What will run on my hardware? →

Registry record as of 2026-08-13. Identity, licence and parameter count are properties of a release and do not move; tier can, and the verify link above re-checks it against the live registry rather than this page.