DeepSeek-V4-Flash-Vision-Exp vs DeepSeek-V4-Pro-0813: Specs & Benchmark Comparison

DeepSeek-V4-Flash-Vision-Exp and DeepSeek-V4-Pro-0813 both come from DeepSeek, and both were released in August 2026. DeepSeek-V4-Pro-0813 has a published size of about 1.6 trillion parameters; DeepSeek has not disclosed the parameter count of DeepSeek-V4-Flash-Vision-Exp.

The two models share 6 published benchmarks. DeepSeek-V4-Pro-0813 leads on 5 of them, DeepSeek-V4-Flash-Vision-Exp on 1. The widest gaps are on AutomationBench, where DeepSeek-V4-Pro-0813 scores 31.8% against 25.7%; Terminal-Bench 2.1, where DeepSeek-V4-Pro-0813 scores 88.0% against 83.9%. Averaged across everything we track, DeepSeek-V4-Flash-Vision-Exp sits at 50.4% and DeepSeek-V4-Pro-0813 at 60.6%.

DeepSeek-V4-Flash-Vision-Exp is the cheaper API at $0.22 per million input tokens and $0.66 per million output tokens, roughly 6 times cheaper than DeepSeek-V4-Pro-0813 at $1.32 and $3.96. Both accept a context window of about 1M tokens. DeepSeek-V4-Flash-Vision-Exp accepts text and images as input, while DeepSeek-V4-Pro-0813 accepts text. On tooling, only DeepSeek-V4-Pro-0813 supports function calling and only DeepSeek-V4-Pro-0813 offers structured output.

CharacteristicDeepSeek-V4-Flash-Vision-ExpDeepSeek-V4-Pro-0813
CompanyDeepSeekDeepSeek
Release DateAugust 21, 2026August 13, 2026
Parameters1.6T
MultimodalYesNo
Context (input)1.0M1.0M
Context (output)393K393K
Input Price / 1M$0.22$1.32
Output Price / 1M$0.66$3.96
Average Score50.4%60.6%
Benchmarks
AutomationBench25.7%31.8%
Terminal-Bench 2.183.9%88.0%
NL2Repo57.7%61.5%
DSBench-Hard63.6%67.2%
DeepSWE59.3%62.7%
Agents' Last Exam27.3%25.7%

Visual Benchmark Comparison

DeepSeek-V4-Flash-Vision-Exp
DeepSeek-V4-Pro-0813
AutomationBench0.3 vs 0.3
0.3
0.3
Terminal-Bench 2.10.8 vs 0.9
0.8
0.9
NL2Repo0.6 vs 0.6
0.6
0.6
DSBench-Hard0.6 vs 0.7
0.6
0.7
DeepSWE0.6 vs 0.6
0.6
0.6
Agents' Last Exam0.3 vs 0.3
0.3
0.3

Verdict

DeepSeek-V4-Flash-Vision-Exp leads in 2 out of 4 comparison categories.

Overall Performance

Both models show comparable average scores: DeepSeek-V4-Flash-Vision-Exp — 0.5, DeepSeek-V4-Pro-0813 — 0.6.

API Cost

DeepSeek-V4-Flash-Vision-Exp is 6.0x cheaper: input $0.22/1M vs $1.32/1M tokens.

Context Window

DeepSeek-V4-Flash-Vision-Exp supports a larger context: 1M vs 1M tokens.

Recency

Both models were released around the same time: 8/21/2026 and 8/13/2026.

More About These Models

Related Comparisons

Frequently Asked Questions

Which is better for coding — DeepSeek-V4-Flash-Vision-Exp or DeepSeek-V4-Pro-0813?
Direct comparison on the SWE-Bench benchmark is not available. We recommend reviewing other metrics on the comparison page.
Which model is cheaper — DeepSeek-V4-Flash-Vision-Exp or DeepSeek-V4-Pro-0813?
DeepSeek-V4-Flash-Vision-Exp is cheaper for input: $0.22 per 1M tokens vs $1.32.
Which has a larger context window — DeepSeek-V4-Flash-Vision-Exp or DeepSeek-V4-Pro-0813?
DeepSeek-V4-Flash-Vision-Exp supports a larger context: 1,048,576 tokens vs 1,000,000.

The DeepSeek-V4-Flash-Vision-Exp and DeepSeek-V4-Pro-0813 comparison is updated for 2026. Data includes benchmark results, API pricing, context window size and other specifications. For more detailed information, visit the DeepSeek-V4-Flash-Vision-Exp or DeepSeek-V4-Pro-0813 page. See also the complete list of AI model comparisons.