Compare models

Context, limits, capabilities, price and use cases side by side — up to 4 models at a time. The URL holds your selection, so you can share it.

DeepSeek V4 Flash Vision ExpDeepSeek V4 FlashDeepSeek V4 Pro Add a modelChat with these 3

AttributeHide identical rows	DeepSeek V4 Flash Vision Expdeepseek/deepseek-v4-flash-vision-exp10% off	DeepSeek V4 Flashdeepseek/deepseek-v4-flash10% off	DeepSeek V4 Prodeepseek/deepseek-v4-pro10% off
Specs
Context window	1M	1M	1M
Max output	384K	384K	384K
Input modalities	TextImage	Text	Text
Output modalities	Text	Text	Text
Capabilities
Streaming	Supported	Supported	Supported
Tool calling	Supported	Supported	Supported
Structured output (JSON)	Supported	Supported	Supported
Image input	Supported	Not supported	Not supported
Thinking mode	Supported	Supported	Supported
Price / 1M tokens
Input	$ 0.45$ 0.41	$ 0.45$ 0.41	$ 1.35$ 1.22
Output	$ 1.35$ 1.22	$ 1.35$ 1.22	$ 4.05$ 3.65
Cache read	$ 0.015$ 0.013	$ 0.015$ 0.013	$ 0.045$ 0.041
Cache write	$ 0.45$ 0.41	$ 0.45$ 0.41	$ 1.35$ 1.22
What it is good at
Best for	Vision-dependent agents Charts and document images Coding agents that read screenshots	Cost-sensitive coding agents High-volume tool workflows	Complex long-horizon agents Research and hard reasoning
Published benchmarks
Terminal-Bench 2.1pass_ratesetups differ by: harness, temperature, top_p	83.9%effort max · DeepSeek Harness minimal(setup not fully disclosed)	78.7%effort max(setup not fully disclosed)	78.7%effort max(setup not fully disclosed)
Agents' Last Examscore	27.3%(setup not fully disclosed)	—	—
ApexBenchpass_at_1	36.5%(setup not fully disclosed)	—	—
AutomationBench (Public)score	25.7%(setup not fully disclosed)	—	—
DSBench-Hardscore	63.6%(setup not fully disclosed)	—	—
DeepSWEresolved_rate	59.3%effort max · DeepSeek Harness minimal(setup not fully disclosed)	—	—
NL2Reposcore	57.7%effort max · DeepSeek Harness minimal(setup not fully disclosed)	—	—
SciCodescore	—	49.9%effort max(setup not fully disclosed)	49.2%effort max(setup not fully disclosed)
GDPval-AA v2elo	—	1558.4effort max(setup not fully disclosed)	1590.3effort max(setup not fully disclosed)
AA-LCRscore	—	74.3%effort max(setup not fully disclosed)	75.3%effort max(setup not fully disclosed)
Chartographyscore	64.3%(setup not fully disclosed)	—	—
ZeroBenchpass_at_5	35%(setup not fully disclosed)	—	—
AA Intelligence Index v4.1.1score	—	51.8effort max(setup not fully disclosed)	53effort max(setup not fully disclosed)
Humanity's Last Examaccuracy	—	38.6%effort max(setup not fully disclosed)	39.3%effort max(setup not fully disclosed)
CritPtscore	—	16.6%effort max(setup not fully disclosed)	18%effort max(setup not fully disclosed)
GPQA Diamondaccuracy	—	90.8%effort max(setup not fully disclosed)	92.8%effort max(setup not fully disclosed)
tau3-Bankingscore	—	39.4%effort max(setup not fully disclosed)	39.6%effort max(setup not fully disclosed)