DeepSeek V4 Flash 0731

DeepSeekOpen weightsReasoningReleased Jul 31, 2026Input: text

Compare with Claude Opus 5.5 Source scores ↗OpenRouter ↗

Overall score

—

Not enough evidence to rank yet

Context window

1,000,000 tokens

Price per 1M tokens

$0.14 in · $0.28 out

Source: BenchLM

Blended price

$0.17

3:1 input:output tokens

Output speed

—

Not measured yet

Category scores

Scores per capability area with the model's rank among ranked models in that category.

Higher is better. #N is the model's rank in that category, where it is eligible to be ranked.

Benchmark results (35)

As reported in system cards, papers and public evaluations. Position compares this model with every other model reporting the same benchmark.

Agentic & tool use

Terminal-Bench 2.1	82.7	#24 of 68
BrowseComp	73.2	#30 of 50
MCP Atlas	69.0	#26 of 42
MazeBench	0	#20 of 42
Terminal-Bench 2.0	56.9	#23 of 39
HLE w/ tools	45.1	#24 of 27
CyberGym	76.7	#17 of 25
AutomationBench	25.1	#25 of 25
Toolathlon	47.8	#13 of 22
Toolathlon-Verified	70.3	#17 of 22
Agents' Last Exam	25.2	#16 of 17

Coding

SWE-bench Verified	79.0	#21 of 98
SWE-bench Pro	52.6	#55 of 69
DeepSWE	54.4	#42 of 50
NL2Repo	54.2	#9 of 31
VulcanBench v3	88.4	#3 of 14
OpenHarmony Bench	53.8	#8 of 12
Codeforces	3052.0	#4 of 4
LiveCodeBench Pass@1-COT	91.6	#2 of 2
DSBench-FullStack	68.7	#2 of 2
DSBench-Hard	59.6	#2 of 2

Reasoning

ARC-AGI-2	61.4	#23 of 41
ARC-AGI-1	89.00	#23 of 36
MRCR 1M	78.7	#2 of 4
CorpusQA 1M	60.5	#2 of 2

Knowledge

GPQA	88.1	#33 of 117
HLE	34.8	#37 of 69
MMLU-Pro	86.2	#12 of 67
Chinese-SimpleQA	78.9	#2 of 2

Math

HMMT Feb 2026	94.8	#3 of 24
IMOAnswerBench	88.4	#4 of 9
Apex	33.0	#6 of 8
Apex Shortlist	85.7	#4 of 4

Multilingual

SWE Multilingual

73.3

#28 of 46

Multimodal

Design Arena Website

1208

#55 of 91

Pricing by source

USD per 1M tokens.

Source	Input	Output	Cached in
List price (BenchLM)	$0.14	$0.28	$0.003
OpenRouter deepseek/deepseek-v4-flash-0731	$0.13	$0.39	$0.004

Historical. DeepSeek V4 Flash 0731 used the shared `deepseek-v4-flash` API ID at $0.14 cache-miss input / $0.0028 cache-hit input / $0.28 output per million tokens; BenchLM represents the model at its highest published reasoning effort. DeepSeek retired the V4 Flash SKU on September 10, 2026 when it launched DeepSeek-V4.1-Flash as `deepseek-flash`; the legacy `deepseek-v4-flash` ID is still accepted but temporarily routes to V4.1 Flash and bills at the V4.1 Flash rate. The rates stored here are historical: the last standalone V4 Flash figures BenchLM recorded from the pricing page checked August 12, 2026, before DeepSeek's peak/off-peak schedule took effect at 16:00 UTC on August 16, 2026.

Open weights

From the Hugging Face repository.

Parameters: 304B
Licence: mit
Downloads (30 days): 4.2M
Repository: deepseek-ai/DeepSeek-V4-Flash-0731 ↗

Speed

Measured by us through OpenRouter (median of trials).

Not measured yet.