deepseek
DeepSeek: DeepSeek V4 Flash Vision Exp (batch)
The DeepSeek V4 Flash Vision Exp is an AI model that can process both text and image inputs, with a maximum context length of 1 megabyte and completion tokens limited to approximately 943 thousand per batch. It supports tools and reasoning capabilities. This model may be worth considering for those on a budget who require a balance between input and output costs, which total $0.44 per million tokens. Its blended benchmark score of 80.1 across three independent benchmarks is notable, but more context would be helpful to fully evaluate its performance in specific tasks.
Quality Score
100/100
price + capability + benchmarks
Input Price
$0.11
per 1M tokens · checked 2026-09-09
Output Price
$0.33
per 1M tokens · checked 2026-09-09
Context Window
1,048,576
tokens
Benchmark results
Independent, published benchmarks. Blended score 80.1 across 3 benchmarks, last refreshed 2026-09-09. How scoring works →
| Benchmark | Measures | Score |
|---|---|---|
| AI Index | broad capability composite | 57.8 |
| AI Index Coding | software engineering tasks | 100.0 |
| AI Index Agentic | multi-step tool-using tasks | 79.4 |
- Model ID
- deepseek/deepseek-v4-flash-vision-exp:batch
- Vendor
- deepseek
- Released
- August 2026
- Tokenizer
- DeepSeek
- Input Modalities
- text, image
- Output Modalities
- text
- Max Output
- 943,718 tokens
- Tool Calling
- ✓ supported
- Structured Output
- ✓ supported
- Reasoning Mode
- ✓ supported
- Vision
- ✓ accepts images
- Audio
- no
- Moderated
- no
What it costs in practice
Computed from the current $0.11/M input and $0.33/M output rates. Run your own numbers →
| Job | Tokens | Cost |
|---|---|---|
| Summarize a 50-page report | 30k in / 1.5k out | under $0.01 |
| Classify 1,000 customer emails | 500k in / 50k out | $0.07 |
| A month of a busy support chatbot | 5M in / 2M out | $1.21 |
Strong choice for
Data
JSON Extraction
Pulling structured fields out of unstructured text.
Data
Bulk Data Labeling
Cheaply tagging thousands of items with consistent labels.
Writing
Short-Form Summarization
TL;DRs of articles and emails at scale.
Writing
Email Drafting
Cold emails, replies, and outreach at the right tone.
Education
Language Learning
Conversational practice, grammar drills, vocabulary.
Business
Customer Support
Replying to tickets and chats accurately.
Category rankings
Where DeepSeek: DeepSeek V4 Flash Vision Exp (batch) places across the 34 categories it ranks in. How we rank →
| # | Category | Score |
|---|---|---|
| #3 | JSON ExtractionData · of 25 ranked | 149 |
| #3 | Bulk Data LabelingData · of 25 ranked | 135 |
| #3 | Short-Form SummarizationWriting · of 25 ranked | 131 |
| #3 | Email DraftingWriting · of 25 ranked | 127 |
| #3 | Language LearningEducation · of 25 ranked | 127 |
| #3 | Customer SupportBusiness · of 25 ranked | 135 |
| #3 | Chat CompanionPersonal · of 25 ranked | 131 |
| #3 | Dataset AnnotationResearch · of 25 ranked | 145 |
| #5 | Trivia & General KnowledgePersonal · of 25 ranked | 121 |
| #6 | Sales / Cold EmailBusiness · of 25 ranked | 123 |
| #6 | Job Application DraftingBusiness · of 25 ranked | 123 |
| #6 | Journaling HelperPersonal · of 25 ranked | 123 |
| #6 | Recipe GenerationPersonal · of 25 ranked | 123 |
| #6 | Real-Time ChatLatency · of 25 ranked | 118 |
| #10 | Image CaptioningVision · of 25 ranked | 120 |
| #11 | Self-Hosted / LocalCost · of 25 ranked | 117 |
| #12 | Code CompletionCode · of 25 ranked | 133 |
| #13 | Resume WritingBusiness · of 25 ranked | 117 |
| #13 | SOP / Process DocsBusiness · of 25 ranked | 136 |
| #15 | Regex WritingCode · of 25 ranked | 134 |
| #15 | Language TranslationWriting · of 25 ranked | 134 |
| #15 | Tutoring for KidsPersonal · of 25 ranked | 134 |
| #15 | Cheap Bulk InferenceCost · of 25 ranked | 137 |
| #16 | Social Media PostsWriting · of 25 ranked | 119 |
| #16 | Voice Assistant BackendVoice · of 25 ranked | 123 |
| #17 | Function / Tool CallingAgents · of 25 ranked | 154 |
| #20 | Code DocumentationCode · of 25 ranked | 141 |
| #20 | CSV / Spreadsheet CleanupData · of 25 ranked | 153 |
| #20 | Meeting NotesBusiness · of 25 ranked | 151 |
| #20 | Browser AutomationAgents · of 25 ranked | 169 |
| #22 | Marketing CopyWriting · of 25 ranked | 128 |
| #22 | Standardized Test PrepEducation · of 25 ranked | 128 |
| #23 | SQL GenerationCode · of 25 ranked | 169 |
| #25 | Diagram ExtractionVision · of 25 ranked | 147 |
Similar models
deepseek
DeepSeek: DeepSeek V4 Flash Vision Exp
$0.22 in / $0.66 out
1,048,576 ctx
100
deepseek
DeepSeek: DeepSeek V4 Flash 0731
$0.07 in / $0.18 out
1,310,720 ctx
99
deepseek
DeepSeek: DeepSeek V4 Flash 0423
$0.09 in / $0.18 out
1,048,576 ctx
99
deepseek
DeepSeek: DeepSeek V4 Flash 0731 (batch)
$0.11 in / $0.33 out
1,048,576 ctx
99
deepseek
DeepSeek: DeepSeek V4 Pro 0813
$0.58 in / $1.74 out
1,048,576 ctx
98
deepseek
DeepSeek: DeepSeek V4 Pro 0813 (batch)
$0.66 in / $1.98 out
1,048,576 ctx
98
Quick answers
- How much does DeepSeek: DeepSeek V4 Flash Vision Exp (batch) cost?
- $0.11 per million input tokens and $0.33 per million output tokens.
- What is DeepSeek: DeepSeek V4 Flash Vision Exp (batch)'s context window?
- 1,048,576 tokens, roughly 1,572 pages of text in a single request.
- Does DeepSeek: DeepSeek V4 Flash Vision Exp (batch) support tool calling?
- It supports tool calling, structured output, a reasoning mode.
- Can DeepSeek: DeepSeek V4 Flash Vision Exp (batch) process images?
- Yes, it accepts image input.