Noometry

DeepSeek, open weights

DeepSeek V4 Flash

DeepSeek V4 Flash by DeepSeek ranks 35th of 354 ranked models on the Noometry Index as of October 2026, with a score of 53.6. Its strongest category is reasoning, where it ranks 30th. API pricing starts at $0.15 per million input tokens and $0.60 per million output tokens, with a 1M-token context window.

Last verified

Noometry rank
#35 of 354
Index score
53.6
Evidence
Confirmed 41 results
Provider
DeepSeek
Released
April 24, 2026
Weights
Open weights
Reasoning
Yes
Context window
1M
Max output
393K
Input price
$0.15 / M
Output price
$0.60 / M
Blended price
$0.26 / M
Output speed
6 tokens/s Kagi
Value
#41 of 219
Knowledge cutoff
May 2025
Input
text, image
Hugging Face
deepseek-ai/DeepSeek-V4-Flash

Category scores

Each category score combines every public result we have in that category.

  1. Coding 47.9
  2. Reasoning 53.7
  3. Math 60.3
  4. Knowledge 55.4
  5. Multilingual 53.0
  6. Instruction Following 74.9
  7. Long Context 43.8
  8. Writing & Preference 63.8
CategoryScoreRankResults
Coding47.9#595
Reasoning53.7#3011
Math60.3#376
Knowledge55.4#483
Multilingual53.0#721
Instruction Following74.9#811
Long Context43.8#851
Writing & Preference63.8#614

Strengths and weaknesses

Categories where DeepSeek V4 Flash places highest and lowest among the models ranked in each, with its score against that category's median.

Strongest categories

CategoryScorevs medianRank
Reasoning53.7+30.1#30 of 350, top 9%
Math60.3+23.7#37 of 327, top 12%
Knowledge55.4+18.1#48 of 314, top 16%

Weakest categories

CategoryScorevs medianRank
Long Context43.8+2.9#85 of 296, top 29%
Instruction Following74.9+3.6#81 of 305, top 27%
Multilingual53.0+5.6#72 of 297, top 25%

Closest competitors

The models ranked just above and below DeepSeek V4 Flash. When scores are this close, price and speed are often the better way to choose.

ModelRankScoreBlended $/MSpeed
DeepSeek V4 Pro#3154.3$0.9916Compare
Gemini 3.5 Flash#3254.2$3.38—Compare
Gemini 3.6 Flash#3354.1$1.50—Compare
GPT-5.2#3454.1$4.8115Compare
GPT-6 Luna#3653.3$0.20—Compare
Grok 4.7#3753.1$3—Compare
DeepSeek V4.1 Flash#3852.8$0.26—Compare
GPT-5.2 Pro#3952.3$57.75—Compare

Sponsored placements are available on pages like this one. Advertise on Noometry

Benchmark results

Every published result we track, with its source. Bold rows are the ones used for ranking; where several exist we prefer independent runs over self-reported numbers.

Coding

BenchmarkScorePositionSettingSourceDate
FrontierCode18.8%#32 of 37, top 87%Epoch AI
LMArena WebDev1431LMArena2026-10-08
LMArena WebDev1582#28 of 113, top 25%highLMArena2026-10-08
SciCode42%highEpoch AI
SciCode44.9%maxEpoch AI
SciCode49.9%#44 of 121, top 37%maxEpoch AI
WeirdML57%highEpoch AI
WeirdML43.8%highEpoch AI
WeirdML63%#25 of 119, top 22%maxEpoch AI
WeirdML45.6%maxEpoch AI
LMArena Coding1440LMArena2026-10-08
LMArena Coding1457#75 of 294, top 26%LMArena2026-10-08
ALE-Bench678.2highEpoch AI
ALE-Bench1,306#25 of 105, top 24%maxEpoch AI
ALE-Bench324.98noneEpoch AI

Reasoning

BenchmarkScorePositionSettingSourceDate
ARC-AGI-256%highEpoch AI
ARC-AGI-246%lowEpoch AI
ARC-AGI-261.4%#25 of 83, top 31%maxEpoch AI
ARC-AGI-22.1%noneEpoch AI
SimpleBench61.1%#21 of 77, top 28%Epoch AI
SimpleBench46.3%Epoch AI
Kagi LLM Benchmark52.2%#61 of 99, top 62%Kagi LLM Benchmark
NYT Connections (extended)89.6%#20 of 91, top 22%Lech Mazur benchmarks
ARC-AGI-187%highEpoch AI
ARC-AGI-184%lowEpoch AI
ARC-AGI-189%#28 of 83, top 34%maxEpoch AI
ARC-AGI-111.8%noneEpoch AI
CritPt3.4%highEpoch AI
CritPt7.1%maxEpoch AI
CritPt16.6%#34 of 134, top 26%maxEpoch AI
Chess Puzzles33%#31 of 129, top 25%maxEpoch AI2026-08-02
LMArena Hard Prompts1431LMArena2026-10-08
LMArena Hard Prompts1444#75 of 297, top 26%LMArena2026-10-08
Mystery Game Puzzles34%#20 of 74, top 28%maxEpoch AI2026-08-05
DTBench90.9%#32 of 151, top 22%maxEpoch AI
DTBench86.4%maxEpoch AI
LMCA35.9%maxEpoch AI
LMCA41.7%#42 of 125, top 34%maxEpoch AI
Epoch Capabilities Index146.07Epoch AI2026-04-24
Epoch Capabilities Index154.49#34 of 213, top 16%Epoch AI2026-07-31

Math

BenchmarkScorePositionSettingSourceDate
FrontierMath (Tiers 1-3)57.5%#38 of 81, top 47%maxEpoch AI2026-08-02
FrontierMath Tier 424.4%#38 of 63, top 61%maxEpoch AI2026-08-02
MathArena Final-Answer Competitions76.5%#8 of 29, top 28%maxMathArena
OTIS Mock AIME 2024-202594.4%#38 of 173, top 22%maxEpoch AI2026-08-02
ProofBench56%#23 of 77, top 30%Epoch AI
LMArena Math1425LMArena2026-10-08
LMArena Math1427#92 of 285, top 33%LMArena2026-10-08

Knowledge

BenchmarkScorePositionSettingSourceDate
GPQA Diamond91%#28 of 186, top 16%maxEpoch AI2026-08-02
SimpleQA Verified33.6%#53 of 77, top 69%maxEpoch AI2026-08-27
LMArena Expert1436LMArena2026-10-08
LMArena Expert1441#82 of 273, top 31%LMArena2026-10-08

Multilingual

BenchmarkScorePositionSettingSourceDate
LMArena Non-English1410LMArena2026-10-08
LMArena Non-English1420#72 of 297, top 25%LMArena2026-10-08
LMArena Chinese1464LMArena2026-10-08
LMArena Chinese1468#76 of 285, top 27%LMArena2026-10-08
LMArena French1437LMArena2026-10-08
LMArena French1439#84 of 223, top 38%LMArena2026-10-08
LMArena German1418#78 of 231, top 34%LMArena2026-10-08
LMArena German1415LMArena2026-10-08
LMArena Japanese1406#54 of 211, top 26%LMArena2026-10-08
LMArena Japanese1389LMArena2026-10-08
LMArena Korean1384#75 of 213, top 36%LMArena2026-10-08
LMArena Korean1357LMArena2026-10-08
LMArena Russian1428#70 of 283, top 25%LMArena2026-10-08
LMArena Russian1426LMArena2026-10-08
LMArena Spanish1418LMArena2026-10-08
LMArena Spanish1436#72 of 226, top 32%LMArena2026-10-08

Instruction Following

BenchmarkScorePositionSettingSourceDate
LMArena Instruction Following1421#71 of 298, top 24%LMArena2026-10-08
LMArena Instruction Following1417LMArena2026-10-08

Long Context

BenchmarkScorePositionSettingSourceDate
LMArena Longer Query1434#74 of 291, top 26%LMArena2026-10-08
LMArena Longer Query1424LMArena2026-10-08

Writing & Preference

BenchmarkScorePositionSettingSourceDate
LMArena Text1424LMArena2026-10-08
LMArena Text1432#76 of 297, top 26%LMArena2026-10-08
LMArena Creative Writing1399LMArena2026-10-08
LMArena Creative Writing1403#72 of 295, top 25%LMArena2026-10-08
EQ-Bench Creative Writing1559#48 of 115, top 42%EQ-Bench
EQ-Bench Creative Writing1441EQ-Bench
LMArena Multi-Turn1431LMArena2026-10-08
LMArena Multi-Turn1449#57 of 295, top 20%LMArena2026-10-08

API pricing by provider

RouteInput $/MOutput $/MCached input $/MChecked
alibaba$0.20$0.40$0.042026-10-10
azure$0.19$0.51—2026-10-10
deepinfra$0.06$0.18$0.0152026-10-10
deepseek$0.15$0.60$0.0032026-10-10
openrouter$0.03$1.28$0.032026-10-10
together$0.14$0.28$0.032026-10-10

All DeepSeek API prices → Estimate your cost →

Compare DeepSeek V4 Flash

Other DeepSeek models

Frequently asked questions

How good is DeepSeek V4 Flash?

DeepSeek V4 Flash by DeepSeek ranks 35th of 354 ranked models on the Noometry Index as of October 2026, with a score of 53.6. Its strongest category is reasoning, where it ranks 30th. API pricing starts at $0.15 per million input tokens and $0.60 per million output tokens, with a 1M-token context window.

How much does DeepSeek V4 Flash cost?

DeepSeek V4 Flash costs $0.15 per million input tokens and $0.60 per million output tokens on DeepSeek's own API, with cached input at $0.003.

What is DeepSeek V4 Flash's context window?

DeepSeek V4 Flash accepts up to 1M tokens of input and can write up to 393K tokens in one response.

Is DeepSeek V4 Flash open source?

Yes. DeepSeek V4 Flash's weights are downloadable from Hugging Face (deepseek-ai/DeepSeek-V4-Flash); check the license for commercial terms.

How fast is DeepSeek V4 Flash?

DeepSeek V4 Flash generated about 6 output tokens per second in the Kagi LLM Benchmark's timed runs. Speed varies by provider, load and reasoning effort.

What are DeepSeek V4 Flash's strengths and weaknesses?

Relative to other ranked models, DeepSeek V4 Flash places best in reasoning, math, knowledge and lowest in long context, instruction following, multilingual.

What is DeepSeek V4 Flash best at?

Its best category is reasoning, where it ranks 30th on Noometry.