OpenModelsHQ

DeepSeek-V4.1-Flash-1017

DeepSeekVisionMoEMITChina Desk

DeepSeek-V4.1-Flash-1017 is an open-weight vision and language model from DeepSeek with 763B parameters (mixture of experts). At 4-bit it needs about 504.5 GB of memory, more than any single device in our directory has.

Overview

Estimated memory at 4-bit
504.5GB

Includes room for an 8k context and runtime overhead. More context needs more memory.

Total parameters
763B
Mixture of experts, all weights counted
Context window
1024K
1,048,576 tokens
Cheapest API, blended / 1M
$0.0815
₹7.69 · Sail Research · 3:1 input to output
All DeepSeek API prices

Key information

Lab
DeepSeek
Released
17 Oct 2026
Parameters
763B · MoE
Type
Vision and language
Context
1,048,576 tokens
License
MIT
HF downloads, 30 days
398,161
View all

Will it run on your hardware?

How we estimate
Memory needed vs devices
DeepSeek-V4.1-Flash-1017 needs about 504.5 GB at 4-bit, 890.6 GB at 8-bit and 1679.8 GB at 16-bit. At 4-bit it fits none of the reference devices. Dashed lines are usable memory on each device; a model fits when its estimate stays within 90% of it. Log scale. 8k context.
Memory needed vs devices: data
Estimate or deviceMemory
Q4504.5 GB
Q8890.6 GB
BF161679.8 GB
RTX 5060 Ti 16 GB16 GB usable
RTX 5090 32 GB32 GB usable
DGX Spark 128 GB108.8 GB usable
Mac Studio M5 Ultra 256 GB192 GB usable

Single-device memory estimate. Pick a device to update the On your device column.

36 GB usable memory

FormatMemory neededOn your deviceExample hardware
4-bit (Q4)≈ 504.5 GBToo big 1401%No single device in our directory has enough memory
8-bit (Q8)≈ 890.6 GBToo big 2474%No single device in our directory has enough memory
16-bit (BF16)≈ 1679.8 GBToo big 4666%No single device in our directory has enough memory

Fits means the estimate stays within 90% of usable memory. Tight is above 90% and up to 100%. Too big is above 100%. These are memory checks, not speed or compatibility tests.

Mac allowances are 66% of total RAM up to 36 GB, and 75% above. AI boxes allow 85% because their memory is shared with the operating system.

Where to run it

On a Mac

No trusted MLX build found yet.

Check your Mac

With GGUF

Choose a file for llama.cpp or a compatible app.

unsloth/DeepSeek-V4.1-Flash-1017-GGUFFind compatible hardware

Through an API

Sail Research (fp4) is the cheapest provider we found.

$0.0727 / $0.108

Input / output per 1M tokens
₹6.85 / ₹10 at ₹94.29 per $1

Blended: $0.0815 (₹7.69) per 1M. Ranked by a 3:1 mix of input and output tokens.

GGUF with llama.cppView repository
llama-cli -hf unsloth/DeepSeek-V4.1-Flash-1017-GGUF \
  -c 8192 -p "Explain how a transformer works."

Example commands built from the repository ids above. They need a working install of llama.cpp. We have not run them; check each repository for available files.

Hosted API 38

USD and INR per million tokensAll DeepSeek API prices
Precision
Sail ResearchCheapest$0.0815₹7.69$0.0727₹6.85$0.108₹101,048,576FP4
Particle AI$0.085₹8.01$0.03₹2.83$0.25₹241,048,576Not listed
Ionstream$0.1162₹11$0.0775₹7.31$0.2325₹221,048,576Not listed
DeepInfra$0.1176₹11$0.0784₹7.39$0.2352₹221,048,576FP8
Wafer$0.1535₹14$0.038₹3.58$0.50₹471,048,576MXFP4
Baidu$0.1719₹16$0.0977₹9.21$0.3944₹371,048,576FP8
StepFun$0.1837₹17$0.105₹9.90$0.42₹401,000,000FP8
StreamLake$0.1943₹18$0.111₹10$0.444₹421,024,000FP8
PrimeIntellect$0.2078₹20$0.0437₹4.12$0.70₹661,048,576Not listed
River$0.21₹20$0.12₹11$0.48₹451,048,576Not listed
InferenceNet$0.2112₹20$0.08₹7.54$0.605₹571,040,000FP8
Relace$0.215₹20$0.12₹11$0.50₹471,048,576FP4
Morph$0.2238₹21$0.104₹9.81$0.583₹551,048,576FP8
Cloudflare$0.225₹21$0.10₹9.43$0.60₹571,048,576Not listed
Cohere$0.2375₹22$0.15₹14$0.50₹471,048,576FP8
OpenInference$0.2525₹24$0.07₹6.60$0.80₹751,048,576FP4
DeepSeek$0.2625₹25$0.15₹14$0.60₹571,048,576Not listed
Novita$0.2888₹27$0.165₹16$0.66₹621,048,576FP8
DekaLLM$0.305₹29$0.18₹17$0.68₹641,048,576Not listed
DigitalOcean$0.315₹30$0.18₹17$0.72₹681,048,576Not listed
GMICloud$0.315₹30$0.18₹17$0.72₹681,048,575FP8
AtlasCloud$0.3412₹32$0.195₹18$0.78₹741,048,576FP8
SiliconFlow$0.3625₹34$0.30₹28$0.55₹521,048,576FP8
Alibaba$0.3675₹35$0.21₹20$0.84₹791,000,000Not listed
Fireworks$0.3675₹35$0.21₹20$0.84₹791,048,576Not listed
Reka$0.395₹37$0.16₹15$1.10₹1041,048,576Not listed
Makora$0.4125₹39$0.15₹14$1.20₹1131,048,576FP8
PrimaLabs$0.42₹40$0.24₹23$0.96₹911,048,576Not listed
Modal$0.4375₹41$0.25₹24$1.00₹941,048,576Not listed
Crusoe$0.5175₹49$0.29₹27$1.20₹1131,048,576FP8
BaseTen$0.525₹50$0.30₹28$1.20₹1131,048,576Not listed
CoreWeave$0.525₹50$0.30₹28$1.20₹1131,048,576FP8
Mistral$0.525₹50$0.30₹28$1.20₹1131,048,576FP8
Near AI$0.525₹50$0.30₹28$1.20₹1131,048,576Not listed
Nebius$0.525₹50$0.30₹28$1.20₹1131,048,576FP8
Parasail$0.525₹50$0.30₹28$1.20₹1131,048,576FP8
Together$0.525₹50$0.30₹28$1.20₹1131,048,576Not listed
Venice$0.525₹50$0.30₹28$1.20₹1131,000,000FP8
View on OpenRouterEndpoint sourceSorted by the blended price: 3 parts input to 1 part output. Context limits differ by provider.

INR at ₹94.29 per $1. Prices fetched 14 Nov 2026.

More about this model

Frequently asked questions

How much memory does DeepSeek-V4.1-Flash-1017 need? At an 8k context, about 504.5 GB at 4-bit, 890.6 GB at 8-bit and 1679.8 GB at 16-bit (BF16). A longer context needs more.

Can I run DeepSeek-V4.1-Flash-1017 on a Mac? No Mac in our directory has enough usable memory for the 4-bit estimate of 504.5 GB. We have not found a trusted MLX build yet, but a GGUF build is available.

Does the full context fit? DeepSeek-V4.1-Flash-1017 supports up to 1,048,576 tokens of context. Our memory estimates use 8k, so long prompts need more memory than shown.

What does it cost through an API? The cheapest provider we found is Sail Research, at $0.07 per 1M input tokens and $0.11 per 1M output tokens (₹6.85 and ₹10 at 94.29 rupees per dollar). Ranked by a 3:1 mix of input and output tokens.

License in plain English

MIT. This license permits commercial use, modification and redistribution, as long as the copyright and license notice are kept. This summary is not legal advice.

Read MIT

Read about this license

OMHQ RigBench

Not yet tested. We show no speed or quality score without a measurement.

Revision history and sources

Last updated . Sources:Hugging Face model card, GGUF repository, OpenRouter.