DeepSeek-V4.1-Flash Benchmark Details

DeepSeek-V4.1-Flash currently shows benchmark results led by HLE (3 / 197, score 63.90), Terminal-Bench 2.1 (1 / 53, score 90.60), CodeForces (1 / 21, score 3471). This page also compares it with 3 competitor models and 2 predecessor or same-series models, including performance and pricing views when available.

Benchmark Results

DeepSeek-V4.1-Flash

Benchmark Results

General Knowledge

3 evaluations

Benchmark / mode

Score

Rank/total

HLE

Max

39.10

78 / 197

HLE

Max

36.80

88 / 197

HLE

MaxTools

63.90

3 / 197

Other

1 evaluations

Benchmark / mode

Score

Rank/total

GPQA Diamond

Max

90.90

39 / 274

Coding and Software Engineer

5 evaluations

Benchmark / mode

Score

Rank/total

CodeForces

Max

3471

1 / 21

DeepSWE

MaxTools

74.20

2 / 38

NL2Repo-Bench

MaxTools

65.40

2 / 16

SEC-Bench Pro

MaxTools

62.80

4 / 5

Program Bench

MaxTools

20.30

8 / 11

AI Agent - Tool Usage

4 evaluations

Benchmark / mode

Score

Rank/total

Terminal-Bench 2.1

MaxTools

90.60

1 / 53

CyberGym

MaxTools

88.10

1 / 8

Terminal-Bench 4.0

MaxTools

31.20

7 / 16

Terminal-Bench 3.0

MaxTools

4 / 11

Agent Level Benchmark

1 evaluations

Benchmark / mode

Score

Rank/total

Agents' Last Exam

MaxTools

31.80

6 / 19

Productivity Knowledge

1 evaluations

Benchmark / mode

Score

Rank/total

AutomationBench

MaxTools

54.80

1 / 17

Multimodal Understanding

3 evaluations

Benchmark / mode

Score

Rank/total

BabyVision

MaxTools

89.60

3 / 9

Chartography

MaxTools

78.90

4 / 8

ZeroBench Main

MaxTools

3 / 6

Math and Reasoning

1 evaluations

Benchmark / mode

Score

Rank/total

MathArena Apex

Max

65.60

1 / 3

Other

1 evaluations

Benchmark / mode

Score

Rank/total

ExploitGym (budget unspecified)

MaxTools

15.30

2 / 4

Compare with other models

Competitor Comparison

Benchmark scores for DeepSeek-V4.1-Flash compared against top models in its class

DeepSeek-V4.1-FlashKimi K3 GLM-5.3 Claude Opus 5

Benchmark categories:

The chart shows each model’s highest score per benchmark within the current filter. Out-of-100 benchmarks use raw heights; out-of-range benchmarks are scaled within that benchmark while labels keep the original scores.

12 benchmarks with comparable scores. Each model shows its best score; mode label is displayed below.

Benchmark	DeepSeek-V4.1-FlashCurrent	Kimi K3	GLM-5.3	Claude Opus 5
HLE Accuracy 综合评估	63.90Thinking Level · High ｜ Tools	59.80Thinking Level · High ｜ Tools	62.50Thinking Level · High ｜ Tools	63.60Thinking Level · High ｜ Tools
GPQA Diamond Accuracy 科学与综合推理	90.90Thinking Level · High	92.90Thinking Level · High	88.10Thinking Level · High	93.40Thinking Level · High
DeepSWE Pass@1 (DeepSWE v1.1) 编程与软件工程	74.20Thinking Level · High ｜ Tools	67.50Thinking Level · High ｜ Tools	66.90Thinking Level · High ｜ Tools	74.00Thinking Level · High ｜ Tools
NL2Repo-Bench Average test pass rate 编程与软件工程	65.40Thinking Level · High ｜ Tools	58.00Thinking Level · High ｜ Tools	58.00Thinking Level · High ｜ Tools	75.30Thinking Level · High ｜ Tools
Program Bench Score 编程与软件工程	20.30Thinking Level · High ｜ Tools	17.50Thinking Level · High ｜ Tools	19.00Thinking Level · High ｜ Tools	37.00Thinking Level · High ｜ Tools
CyberGym Score AI Agent - 工具使用	88.10Thinking Level · High ｜ Tools	80.00Thinking Level · High ｜ Tools	84.50Thinking Level · High ｜ Tools	--
Terminal-Bench 2.1 Accuracy AI Agent - 工具使用	90.60Thinking Level · High ｜ Tools	88.30Thinking Level · High ｜ Tools	88.20Thinking Level · High ｜ Tools	89.10Thinking Level · High ｜ Tools
Terminal-Bench 3.0 Accuracy AI Agent - 工具使用	30.00Thinking Level · High ｜ Tools	17.70Thinking Level · High ｜ Tools	28.30Thinking Level · High ｜ Tools	43.30Thinking Level · High ｜ Tools
Terminal-Bench 4.0 Resolution rate (%) AI Agent - 工具使用	31.20Thinking Level · High ｜ Tools	12.60Thinking Level · High ｜ Tools	37.90Thinking Level · High ｜ Tools	51.80Thinking Level · High ｜ Tools
Agents' Last Exam Score Agent能力评测	31.80Thinking Level · High ｜ Tools	27.60Thinking Level · High ｜ Tools	28.50Thinking Level · High ｜ Tools	28.60Thinking Level · High ｜ Tools
AutomationBench Pass Rate 生产力知识	54.80Thinking Level · High ｜ Tools	46.70Thinking Level · High ｜ Tools	48.80Thinking Level · High ｜ Tools	50.30Thinking Level · High ｜ Tools
BabyVision Accuracy 多模态理解	89.60Thinking Level · High ｜ Tools	85.70Thinking Level · High ｜ Tools	--	94.10Thinking Level · High ｜ Tools

4 additional benchmarks remain in the chart above.

Standard API Pricing: DeepSeek-V4.1-Flash vs. Peer Models

Shows standard text input and output pricing side by side for each model. If extended-context pricing exists, the chart keeps the base rate and explains the threshold below.

Source: DataLearnerAI. Standard text prices shown here use the default supplier.

These models use different currencies or billing units, so the page falls back to raw price values instead of a shared bar chart.

Kimi K3

Supplier: Moonshot AI

Standard input: ¥20 / 1M tokens

Standard output: ¥100 / 1M tokens

GLM-5.3

Supplier: 智谱AI

Standard input: $1.4 / 1M tokens

Standard output: $4.4 / 1M tokens

Claude Opus 5

Supplier: Anthropic

Standard input: $5 / 1M tokens

Standard output: $25 / 1M tokens

Model	Supplier	Standard input	Standard output	Base price applies to
Kimi K3	Moonshot AI	¥20 / 1M tokens	¥100 / 1M tokens	—
GLM-5.3	智谱AI	$1.4 / 1M tokens	$4.4 / 1M tokens	—
Claude Opus 5	Anthropic	$5 / 1M tokens	$25 / 1M tokens	—

Version History

How each version of the DeepSeek-V4.1-Flash series stacks up on benchmark tests

DeepSeek-V4.1-FlashDeepSeek-V4-Flash-Vision-Exp DeepSeek-V4-Flash

Benchmark categories:

12 benchmarks with comparable scores. Each model shows its best score; mode label is displayed below.· Click a row to view its trend chart.

Benchmark	DeepSeek-V4.1-FlashCurrent	DeepSeek-V4-Flash-Vision-Exp	DeepSeek-V4-Flash
HLE Accuracy 综合评估	63.90Thinking Level · High ｜ Tools	--	51.50Thinking Level · High ｜ Tools
GPQA Diamond Accuracy 科学与综合推理	90.90Thinking Level · High	--	89.90Thinking Level · High
CodeForces Accuracy 编程与软件工程	3471.00Thinking Level · High	--	3289.00Thinking Level · High
DeepSWE Pass@1 (DeepSWE v1.1) 编程与软件工程	74.20Thinking Level · High ｜ Tools	59.30Thinking Level · High ｜ Tools	54.40Thinking Level · High ｜ Tools
NL2Repo-Bench Average test pass rate 编程与软件工程	65.40Thinking Level · High ｜ Tools	57.70Thinking Level · High ｜ Tools	54.20Thinking Level · High ｜ Tools
SEC-Bench Pro Score (%) 编程与软件工程	62.80Thinking Level · High ｜ Tools	--	30.90Thinking Level · High ｜ Tools
CyberGym Score AI Agent - 工具使用	88.10Thinking Level · High ｜ Tools	75.30Thinking Level · High ｜ Tools	76.70Thinking Level · High ｜ Tools
Terminal-Bench 2.1 Accuracy AI Agent - 工具使用	90.60Thinking Level · High ｜ Tools	83.90Thinking Level · High ｜ Tools	82.70Thinking Level · High ｜ Tools
Terminal-Bench 3.0 Accuracy AI Agent - 工具使用	30.00Thinking Level · High ｜ Tools	--	7.60Thinking Level · High ｜ Tools
Terminal-Bench 4.0 Resolution rate (%) AI Agent - 工具使用	31.20Thinking Level · High ｜ Tools	--	7.00Thinking Level · High ｜ Tools
Agents' Last Exam Score Agent能力评测	31.80Thinking Level · High ｜ Tools	27.30Thinking Level · High ｜ Tools	25.20Thinking Level · High ｜ Tools
AutomationBench Pass Rate 生产力知识	54.80Thinking Level · High ｜ Tools	25.70Thinking Level · High ｜ Tools	37.70Thinking Level · High ｜ Tools

4 additional benchmarks remain in the chart above.

Single-Benchmark Version Trend

Viewing: HLE · 综合评估

Benchmark

NormalNormal + ToolsThinkingThinking + ToolsDeepDeep + Tools

X-axis shows model and release date, Y-axis shows score; solid lines connect the same mode across versions, while dotted guides align modes within the same generation.

Standard API Pricing Across the DeepSeek-V4.1-Flash Series

Shows standard text input and output pricing side by side for each model. If extended-context pricing exists, the chart keeps the base rate and explains the threshold below.

Source: DataLearnerAI. Standard text prices shown here use the default supplier. · USD / 1M tokens

Model	Supplier	Standard input	Standard output	Base price applies to
DeepSeek-V4-Flash	DeepSeek-AI	$0.14 / 1M tokens	$0.28 / 1M tokens	—