Skip to main content

DeepSeek: DeepSeek V4 Flash

DeepSeek: DeepSeek V4 Flash is deepseek's model. On the un-leaked meo benchmark it ranks #7 of 22 with 56.2% accuracy and an Effective Value (๐•) of 0.0001 โ€” earning a correct answer for about $0.0037 in 370s each.

as of 2026-06-08

meo rank
#7
of 22
Accuracy
56.2%
overall
๐• index (top=100)
0.30
raw ๐• 0.0001 ยท N=10
$ / correct
$0.0037
370s each

Per-domain accuracy

Tape machine
100%
Theory of mind
100%
Regex automaton
100%
CSP (unsatisfiable)
94%
Framework bias
67%
Long arithmetic
33%
Watson-Glaser
25%
Logic / math / CS
5%
Base-rate bias
5%
State tracking
0%

Efficiency

Cost per correct
$0.0037
Seconds per correct
370
Tokens per correct
15,115
Correct answers
61 / 151
Total benchmark cost
$0.22
Avg tokens / item
6,106

Aggregated third-party benchmarks

Artificial Analysis (artificialanalysis.ai). Redistribution requires an AA commercial license.

Model details

Lab
deepseek
Released
2026-04-24
Weights
Open
Context
1.0M
Max output
131K
Knowledge cutoff
โ€”
Price in
$0.10/M
Price out
$0.20/M
Modality
text->text

Sources: openrouter ยท artificial_analysis ยท meo

FAQ

How accurate is DeepSeek: DeepSeek V4 Flash?

DeepSeek: DeepSeek V4 Flash scores 56.2% overall on the meo benchmark (rank #7 of 22), an un-leaked private holdout graded objective-first.

What is DeepSeek: DeepSeek V4 Flash's Effective Value?

Its Effective Value (๐•) is 0.0001 at N=10, ฯ‰=1, ฮด=1.5 โ€” a fused metric of accuracy, speed, cost, and error-cascade. ๐• rankings shift with task depth.

How much does DeepSeek: DeepSeek V4 Flash cost per correct answer?

About $0.0037 per correct answer (15,115 tokens, 370 seconds each) at benchmark-time pricing via OpenRouter.

Is DeepSeek: DeepSeek V4 Flash open weights?

Yes โ€” DeepSeek: DeepSeek V4 Flash is released with open weights.