Compare Models

DeepSeek V4 Flash 0731

DeepSeek V4 Flash 0731 is an open-weight model from DeepSeek, released July 31, 2026. Its best result is #41 of 123 on Vibe Code Bench v1.1.

Release Date: Jul 31, 2026

Developer DeepSeek  🇨🇳
Context Window 1M
Max Output Tokens 384k
Token Costs (in/out) $0.30/1.20
Weights Open
Input Modalities

Accuracy

55.07 %

Avg. Cost (In/Out)

$ 0.30 / $ 1.20

Latency

38 min 20 s

Vals Index
BenchmarksAccuracyRankings
Code Migration

38.63%

±4.38
33/79
CyberBench v1.1

72.08%

±5.41
17/47
EMB

56.98%

±3.39
44/76
Finance Agent (v2)

49.52%

±0.47
48/81
Legal Research Bench

30.29%

±3.19
52/89
MedCode

41.41%

±2.15
53/107
MedScribe

80.36%

±1.97
58/109
ProofBench v1.1

56.00%

±4.99
27/52
Tax Agent Bench

28.07%

±3.16
34/80
TaxEval v2

70.69%

±0.90
90/145
Vibe Code Bench v1.1

74.74%

±3.44
41/123
BioMysteryBench

64.44%

±1.28
23/29
GPQA Diamond

89.90%

±1.69
27/138
IOI

32.72%

±2.35
41/46
LiveCodeBench

87.26%

±0.97
17/143
LegalBench

77.71%

±0.50
111/151
MMLU Pro

86.21%

±0.34
47/138
ProgramBench

0.00%

±0.00
39/64
SkillsBench

50.67%

±4.43
22/35
SWE-bench

88.80%

±1.41
11/88
Terminal-Bench 4.0

18.69%

±3.54
35/59
Terminal-Bench-Science

4.29%

±2.44
28/43
Proprietary BenchmarksAcademic BenchmarksIndustry Partners
Vals
Default Provider : DeepSeek
Temperature: 1
Top P: Default
Top K: Default
Max Output Tokens: 384,000
Reasoning Effort: high

Compared with

DeepSeek V4 Flash 0731 vs Qwen 3.8 Max 18 shared benchmarks DeepSeek V4 Flash 0731 vs Mistral Large 4 9 shared benchmarks DeepSeek V4 Flash 0731 vs Muse Spark Open 4 shared benchmarks DeepSeek V4 Flash 0731 vs Muse Spark 1.2 16 shared benchmarks DeepSeek V4 Flash 0731 vs Step 5 Preview 8 shared benchmarks DeepSeek V4 Flash 0731 vs Gemini 3.5 Flash 16 shared benchmarks DeepSeek V4 Flash 0731 vs Grok 4.5 17 shared benchmarks DeepSeek V4 Flash 0731 vs Hy4 Preview 13 shared benchmarks DeepSeek V4 Flash 0731 vs GLM 5.3 Flash 17 shared benchmarks