DeepSeek-V4.1 Flash

DeepSeek's new multimodal Causal Encoder-Decoder MoE has a 552B backbone, activates 8B parameters for input and 16B for output, and supports a one-million-token context. The pinned archive contains about 763B tensors once its 196B Engram memory, vision stack, and auxiliary weights are counted, so even 4-bit local inference is a cluster-scale proposition. Its native FP4 cache is unusually compact at 890 bytes per token.

DeepSeekMixture of expertsagenticreasoning2 machines can run it

Total parameters

763.205B

Official spec sets memory need

Active parameters

16B

Official spec sets decode speed

Context

Official spec tokens

4-bit weights

416 GB

Calculated before KV cache

This is a sparse mixture of experts. All 763.205B of weights must sit in memory, but only 16B are read per generated token — so it needs the memory of a 763.205B model and generates at roughly the speed of a 16B one. That is why large unified-memory machines suit it and fast 32 GB GPUs do not.

Model specification

Publisher

DeepSeek Official spec

Architecture

Sparse mixture of experts Official spec

Total parameters

763.205B Official spec

Active parameters

16B per token Official spec

Context length

1,048,576 tokens Official spec

Attention ⓘ

CED · 1 KV heads × 512 dim × 40 layers Official spec

Licence

MIT Official spec source

Specification confidence ⓘ

High · verified 2026-09-10 Official spec source

Released

10 Sept 2026 Official spec

Official source

Hugging Face

Quantizations and memory

Quantization	Format	Bits/weight	Weights	Quality kept
`MLX 4-bit`	mlx	4.5	415.8 GB	98.0%
`Q4_K_M`default	gguf	4.85	443.8 GB	98.5%
`Q8_0`	gguf	8.5	770.3 GB	99.9%

Weight sizes are computed from the parameter count and bits per weight plus a format-specific overhead for the layers that stay at higher precision — not read from a specific published file. Quality retention is an assumption, not a measured evaluation.

Coding & quality benchmarksCompare coding results →

No published benchmark results have been imported for this model yet. This is missing evidence, not a score of zero.

We show source metrics rather than deriving one opaque quality number. Different benchmarks measure genuinely different things, and collapsing them into a single score would hide exactly the disagreements worth seeing.

Recommendations

Cheapest that can run it

NVIDIA DGX Station GB300 (748 GB)

~70–101 t/s · $90,082

Lowest purchase price among configurations where the model fits at some quantization in our catalogue. Speed is not considered.

Cheapest above 20 t/s

NVIDIA DGX Station GB300 (748 GB)

~70–101 t/s · $90,082

20 tokens/second is roughly the point at which generation keeps pace with reading. Below it, interactive use feels like waiting.

Cheapest above 40 t/s

NVIDIA DGX Station GB300 (748 GB)

~70–101 t/s · $90,082

40 tokens/second is the threshold most people describe as comfortable for coding agents, where output arrives faster than you can review it.

Fastest with real measurements

Nothing in the database qualifies.

Highest throughput among configurations with an actual published measurement rather than our estimate.

Best throughput per purchase-price unit

NVIDIA DGX H200 (8x H200, 1,128 GB)

~451–649 t/s · $359,430

Highest decode tokens/second per 1,000 units of the displayed purchase currency. Ignores running costs and resale — see the economics section for the full picture.

Hardware that runs DeepSeek-V4.1 Flash0 measured, 2 estimated

Quantization

Minimum decode: 0 t/s

2 of 2 rows

Hardware↕	Quant	Memory↕	Decode▼	Prefill↕	Context	Price↕	Fit	Confidence↕
NVIDIA DGX H200 (8x H200, 1,128 GB) NVIDIA · 1,128 GB · 38,400 GB/s	`Q4_K_M`	461.0 GB	~451–649 t/s	~24210–50290 t/s	1024K	$359,430	Comfortable	Estimated
NVIDIA DGX Station GB300 (748 GB) NVIDIA · 748 GB · 7,100 GB/s	`Q4_K_M`	458.2 GB	~70–101 t/s	~3240–6720 t/s	1024K	$90,082	Comfortable	Estimated

Hosted alternatives for this exact model

These endpoints serve the same open weights, so comparing them against local ownership is a like-for-like economic question rather than a quality trade-off.