DeepSeek v3 vs DeepSeek V4.1 Flash 1017

DeepSeek V4.1 Flash 1017 leads llmrun Score 134.4 to 89.0; neither fits 24 GB.

Run DeepSeek V4.1 Flash 1017 in the cloud →

Model A · DeepSeek

DeepSeek v3

89.0llmrun Score

684.5B · 40.1B active · 160K context · Dec 2024

ChatFunctions

Model B · DeepSeek

DeepSeek V4.1 Flash 1017

134.4llmrun Score

763.2B · 16B active · 1M context · MIT · Oct 2026

Vision

Which should you choose?

Choose DeepSeek v3 if you want…

Also does chat, function-calling
DeepSeek V4.1 Flash 1017 doesn't list it

Choose DeepSeek V4.1 Flash 1017 if you want…

Higher llmrun Score
134.4 vs 89.0
Stronger at reasoning
137.1 vs 86.9
Stronger at science
134.3 vs 89.5
Ahead on the benchmarks
4 of 4 shared boards
Longer context window
1M vs 160K tokens
Also does vision
DeepSeek v3 doesn't list it

On your hardware

Each model at its default quant (Q4_K_M), a smaller one when it must. Estimates, ±20%.

MemoryDeepSeek v3DeepSeek V4.1 Flash 1017

8 GB

RTX 4060 · 8 GB tier

DeepSeek v3

Doesn't fit

smallest quant needs 189.7 GB

DeepSeek V4.1 Flash 1017

Doesn't fit

smallest quant needs 210.4 GB

12 GB

RTX 3060 12GB · 12 GB tier

DeepSeek v3

Doesn't fit

smallest quant needs 189.7 GB

DeepSeek V4.1 Flash 1017

Doesn't fit

smallest quant needs 210.4 GB

16 GB

RTX 5060 Ti 16GB · 16 GB tier

DeepSeek v3

Doesn't fit

smallest quant needs 189.7 GB

DeepSeek V4.1 Flash 1017

Doesn't fit

smallest quant needs 210.4 GB

24 GBreference

RTX 4090 · 24 GB tier

DeepSeek v3

Doesn't fit

smallest quant needs 189.7 GB

DeepSeek V4.1 Flash 1017

Doesn't fit

smallest quant needs 210.4 GB

32 GB

RTX 5090 · 32 GB tier

DeepSeek v3

Doesn't fit

smallest quant needs 189.7 GB

DeepSeek V4.1 Flash 1017

Doesn't fit

smallest quant needs 210.4 GB

48 GB

RTX A6000 · 48 GB tier

DeepSeek v3

Doesn't fit

smallest quant needs 189.7 GB

DeepSeek V4.1 Flash 1017

Doesn't fit

smallest quant needs 210.4 GB

Mac 32 GB

Mac Mini M4 · Macs

DeepSeek v3

Doesn't fit

smallest quant needs 189.7 GB

DeepSeek V4.1 Flash 1017

Doesn't fit

smallest quant needs 210.4 GB

Mac 64 GB

Mac Studio M4 Max · Macs

DeepSeek v3

Doesn't fit

smallest quant needs 189.7 GB

DeepSeek V4.1 Flash 1017

Doesn't fit

smallest quant needs 210.4 GB

Mac 128 GB

Mac Studio M4 Max · Macs

DeepSeek v3

Doesn't fit

smallest quant needs 189.7 GB

DeepSeek V4.1 Flash 1017

Doesn't fit

smallest quant needs 210.4 GB

Reference GPU: RTX 4090. Tinted cell = the better pick on that machine (fits, more headroom, or at least 10% faster).

llmrun Score

with its 90% range · bars share one scale

DeepSeek v3 DeepSeek V4.1 Flash 1017

Overall

DeepSeek v3

89.085–93

DeepSeek V4.1 Flash 1017

134.4132–137

Coding

DeepSeek v3

96.991–103

DeepSeek V4.1 Flash 1017

135.1127–143

Agents

DeepSeek v3

no data

DeepSeek V4.1 Flash 1017

135.7131–141

Math

DeepSeek v3

no data

DeepSeek V4.1 Flash 1017

132.0129–135

Science

DeepSeek v3

89.582–97

DeepSeek V4.1 Flash 1017

134.3123–146

Reasoning

DeepSeek v3

86.972–102

DeepSeek V4.1 Flash 1017

137.1133–142

DeepSeek v3 · #76 of 106 open models · #72 of 102 that fit ≤512 GB

DeepSeek V4.1 Flash 1017 · #2 of 106 open models · #1 of 102 that fit ≤512 GB

Benchmarks

4 shared boards: DeepSeek v3 leads 0, DeepSeek V4.1 Flash 1017 leads 4 · Benchmarks imported Nov 2026

BoardDeepSeek v3DeepSeek V4.1 Flash 1017

GPQA Diamond

DeepSeek v3

56.5%#47/100 open

IndependentEpoch AI · Dec 2024

DeepSeek V4.1 Flash 1017

93.2%#1/100 open

IndependentEpoch AI · Oct 2026

AIME 2024/2025

DeepSeek v3

15.8%#58/88 open

IndependentEpoch AI · Dec 2024

DeepSeek V4.1 Flash 1017

99.4%#1/88 open

IndependentEpoch AI · Oct 2026

SciCode

DeepSeek v3

35.8%#52/72 open

Lab-reportedEpoch AI · Dec 2024

DeepSeek V4.1 Flash 1017

55.9%#7/72 open

Lab-reportedEpoch AI · Oct 2026

CritPt

DeepSeek v3

0.0%#54/84 open

Lab-reportedEpoch AI · Dec 2024

DeepSeek V4.1 Flash 1017

24.0%#3/84 open

Lab-reportedEpoch AI · Oct 2026

33 more boards (other shared and single-model boards)

Only DeepSeek v3 (14)

Only DeepSeek V4.1 Flash 1017 (19)

VRAM by quantization

Memory to load the model and 2K context; below it, the total at 32K and 128K context. Lighter side in bold.

QuantDeepSeek v3DeepSeek V4.1 Flash 1017

Q2_K

DeepSeek v3

295 GB~290.9 GB file

32K 294 GB · 128K 300 GB

DeepSeek V4.1 Flash 1017

325 GB264.5 GB file

32K 325 GB · 128K 326 GB

Q3_K_M

DeepSeek v3

338 GB319.2 GB file

32K 336 GB · 128K 343 GB

DeepSeek V4.1 Flash 1017

372 GB347.3 GB file

32K 373 GB · 128K 374 GB

Q4_K_M

DeepSeek v3

415 GB404.4 GB file

32K 413 GB · 128K 420 GB

DeepSeek V4.1 Flash 1017

458 GB444.7 GB file

32K 459 GB · 128K 460 GB

Q5_K_M

DeepSeek v3

492 GB475.4 GB file

32K 490 GB · 128K 497 GB

DeepSeek V4.1 Flash 1017

544 GB~543.8 GB file

32K 545 GB · 128K 546 GB

Q6_K

DeepSeek v3

569 GB550.8 GB file

32K 567 GB · 128K 574 GB

DeepSeek V4.1 Flash 1017

630 GB~629.6 GB file

32K 630 GB · 128K 632 GB

Q8_0

DeepSeek v3

688 GB713.3 GB file

32K 687 GB · 128K 694 GB

DeepSeek V4.1 Flash 1017

764 GB~763.2 GB file

32K 764 GB · 128K 765 GB

10 more quants

IQ2_XXS

DeepSeek v3

192 GB174.4 GB file

32K 191 GB · 128K 198 GB

DeepSeek V4.1 Flash 1017

210 GB300.5 GB file

32K 211 GB · 128K 212 GB

IQ2_XS

DeepSeek v3

209 GB195.1 GB file

32K 208 GB · 128K 215 GB

DeepSeek V4.1 Flash 1017

229 GB~229.0 GB file

32K 230 GB · 128K 231 GB

IQ2_S

DeepSeek v3

218 GB197.0 GB file

32K 217 GB · 128K 223 GB

DeepSeek V4.1 Flash 1017

not listed

IQ2_M

DeepSeek v3

235 GB217.4 GB file

32K 234 GB · 128K 241 GB

DeepSeek V4.1 Flash 1017

not listed

IQ3_XXS

DeepSeek v3

269 GB257.9 GB file

32K 268 GB · 128K 275 GB

DeepSeek V4.1 Flash 1017

296 GB394.7 GB file

32K 296 GB · 128K 298 GB

IQ3_XS

DeepSeek v3

286 GB272.8 GB file

32K 285 GB · 128K 292 GB

DeepSeek V4.1 Flash 1017

not listed

IQ3_S

DeepSeek v3

295 GB289.1 GB file

32K 294 GB · 128K 300 GB

DeepSeek V4.1 Flash 1017

325 GB423.6 GB file

32K 325 GB · 128K 326 GB

IQ3_M

DeepSeek v3

312 GB292.1 GB file

32K 311 GB · 128K 318 GB

DeepSeek V4.1 Flash 1017

not listed

IQ4_NL

DeepSeek v3

389 GB378.1 GB file

32K 388 GB · 128K 395 GB

DeepSeek V4.1 Flash 1017

not listed

BF16

DeepSeek v3

1373 GB1342.3 GB file

32K 1372 GB · 128K 1378 GB

DeepSeek V4.1 Flash 1017

1527 GB~1526.4 GB file

32K 1527 GB · 128K 1528 GB

How to run each

DeepSeek v3

Ollama · Q4_K_M

$ollama run hf.co/unsloth/DeepSeek-V3-GGUF:Q4_K_M

All ways to run DeepSeek v3 →

DeepSeek V4.1 Flash 1017

Ollama · Q4_K_M

$ollama run hf.co/mrfakename/DeepSeek-V4.1-Flash-1017-GGUF:Q4_K_M

All ways to run DeepSeek V4.1 Flash 1017 →

Specifications

Specifications of DeepSeek v3 and DeepSeek V4.1 Flash 1017
Spec	DeepSeek v3	DeepSeek V4.1 Flash 1017
Publisher	DeepSeek	DeepSeek
Parameters	684.5B	763.2B
Active parameters	40B per token	16B per token
Architecture	DeepseekV3ForCausalLM	DeepseekV41ForCausalLM
Context	160K tokens	1M tokens
Modalities	Chat, Function-calling	Vision
License	—	MIT
Released	Dec 2024	Oct 2026

Frequently Asked Questions

Which is better, DeepSeek v3 or DeepSeek V4.1 Flash 1017?

DeepSeek V4.1 Flash 1017 leads llmrun Score 134.4 to 89.0; neither fits 24 GB. Choose DeepSeek v3 for also does chat, function-calling; choose DeepSeek V4.1 Flash 1017 for higher llmrun Score and stronger at reasoning.

Which needs less VRAM, DeepSeek v3 or DeepSeek V4.1 Flash 1017?

At Q4_K_M, DeepSeek v3 requires 414.6 GB while DeepSeek V4.1 Flash 1017 needs 458.3 GB, making DeepSeek v3 the easier model to run on your local setup.

Can I run DeepSeek v3 or DeepSeek V4.1 Flash 1017 on a 16 GB GPU?

DeepSeek v3 does not fit (smallest quant needs 189.7 GB); DeepSeek V4.1 Flash 1017 does not fit (smallest quant needs 210.4 GB).

Which has a longer context window, DeepSeek v3 or DeepSeek V4.1 Flash 1017?

DeepSeek v3 supports 163,840 tokens and DeepSeek V4.1 Flash 1017 supports 1,048,576 tokens.