[Integrations](https://spiral.ai/integrations) › [Codex](https://spiral.ai/integrations/codex) › [DeepSeek V4 Flash Vision Exp](https://spiral.ai/integrations/codex/deepseek-v4-flash-vision-exp)

# How to Use DeepSeek V4 Flash Vision Exp in Codex

DeepSeek V4 Flash Vision Exp pairs image understanding with coding and reasoning for screenshot, chart, and document tasks. Compared with DeepSeek V4 Flash 0731, it adds visual understanding while keeping comparable text capabilities.

|  |  |
| --- | --- |
| Model | DeepSeek V4 Flash Vision Exp |
| Available through | OpenRouter |
| Category | LLM |
| Price | Up to $0.00705 per run at the default inputs. $0.572 per 1M input tokens, $1.716 per 1M output tokens. |
| Returns | text, json |
| Runs | In the background |
| Setup | About 2 minutes, no API key |

## Prompts to try with DeepSeek V4 Flash Vision Exp in Codex

- Use DeepSeek V4 Flash Vision Exp to inspect the failing screen and this project's form components. Transcribe the visible error, trace the relevant validation path, and propose the smallest patch with a regression check.
- Use DeepSeek V4 Flash Vision Exp to compare this page screenshot with the acceptance criteria in the project. List visible mismatches, locate the responsible components, and patch only the confirmed issues.
- Use DeepSeek V4 Flash Vision Exp to examine this chart screenshot and the chart-rendering code. Check axis labels, units, and legend mappings, then propose tests for any mismatch.
- Use DeepSeek V4 Flash Vision Exp to draft an extractor for the supplied document images. Return JSON with labeled values and evidence text, use null for unreadable values, and add validation in the project's existing style.
- Use DeepSeek V4 Flash Vision Exp with maximum reasoning effort to investigate this screenshot and the relevant error logs. Separate observations from hypotheses, inspect the implicated code, and verify the proposed fix.

## How to set up DeepSeek V4 Flash Vision Exp in Codex

Pick the way that suits you. Each takes about two minutes.

### Use Spiral's MCP server

1. **Add the hosted server**

```bash
codex mcp add spiral --url https://api.spiral.ai/mcp
```

2. **Sign in**: Codex opens the browser: sign in to Spiral and click Authorize. No API key. `codex mcp list` then shows Spiral.

```bash
codex mcp login spiral
```

### Use the Spiral skill

1. **Send Codex this message**: Your agent reads Spiral's skill and sets itself up.

```
set up https://spiral.ai/SKILL.md
```

2. **Sign in or paste a key**: Approve Spiral in the browser, or give the agent a key from the console.

### Use the Spiral CLI

1. **Install and sign in**: Installs the Spiral CLI, signs you in and adds its skill to your coding agents.

```bash
npx -y spiral-cli@latest init
```

### Call the Spiral API

1. **Create an API key**: In the console, under API keys.

2. **Run DeepSeek V4 Flash Vision Exp**: One request; the result comes back by polling or a webhook.

```bash
curl https://api.spiral.ai/v1/runs \
  -H "Authorization: Bearer $SPIRAL_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: $(uuidgen)" \
  -d '{
    "provider": "openrouter",
    "model": "deepseek/deepseek-v4-flash-vision-exp",
    "input": {
      "messages": [
        {
          "role": "system"
        }
      ]
    }
  }'
```

Then paste one of the prompts above: Codex finds DeepSeek V4 Flash Vision Exp on Spiral, shows the price and runs it.

Setup guide in the docs: https://spiral.ai/docs/quickstarts/codex

## DeepSeek V4 Flash Vision Exp tools and prices

- [DeepSeek: DeepSeek V4 Flash Vision Exp](https://spiral.ai/tools/openrouter/deepseek/deepseek-v4-flash-vision-exp) (OpenRouter): Up to $0.220432 per run at the default inputs. $0.28028 per 1M input tokens, $0.84084 per 1M output tokens.

## Settings you can ask for

| Setting | Allowed values | Default | Description |
| --- | --- | --- | --- |
| seed |  | — |  |
| stop |  | — |  |
| min_p | 0 to 1 | — |  |
| tools |  | — |  |
| top_k | 0 to … | — |  |
| top_p | 0 to 1 | — |  |

## What can you do with DeepSeek V4 Flash Vision Exp in Codex?

Turn screenshots into code changes by pairing visual evidence with the relevant project files. Use DeepSeek V4 Flash Vision Exp to draft a patch and define checks for the reported problem.

### Turn a screenshot into a bug report

Ask DeepSeek V4 Flash Vision Exp to transcribe the visible error, describe the affected screen, and separate observed behavior from possible causes.

### Review an interface against its requirements

Give DeepSeek V4 Flash Vision Exp a screenshot and written acceptance criteria. Request a checklist of visible matches, mismatches, and requirements the screenshot cannot establish.

### Explain a chart

Have DeepSeek V4 Flash Vision Exp identify the axes, units, trends, and outliers in a chart image. Require it to distinguish readable values from estimates.

### Extract information from a document image

Ask DeepSeek V4 Flash Vision Exp to turn a document image into a labeled record, leaving unreadable or missing values empty for review.

### Draft image descriptions

Use DeepSeek V4 Flash Vision Exp to draft concise descriptions of product photos or instructional screenshots. Specify the audience and ask it to exclude unsupported assumptions.

### Connect visual bugs to source code

Let DeepSeek V4 Flash Vision Exp examine a failing screen alongside relevant project files, propose a small patch, and identify how to verify it.

### Prepare recurring dashboard briefs

Give DeepSeek V4 Flash Vision Exp a dashboard screenshot and the reporting criteria. Request a short brief with visible metrics, notable patterns, and questions that need more data.

## Overview

- [OpenRouter](https://spiral.ai/providers/openrouter)
- [All Codex guides](https://spiral.ai/integrations/codex)

## DeepSeek V4 Flash Vision Exp in other agents

- [Amp](https://spiral.ai/integrations/amp/deepseek-v4-flash-vision-exp)
- [ChatGPT](https://spiral.ai/integrations/chatgpt/deepseek-v4-flash-vision-exp)
- [Claude](https://spiral.ai/integrations/claude-desktop/deepseek-v4-flash-vision-exp)
- [Claude Code](https://spiral.ai/integrations/claude-code/deepseek-v4-flash-vision-exp)
- [Cline](https://spiral.ai/integrations/cline/deepseek-v4-flash-vision-exp)
- [Cursor](https://spiral.ai/integrations/cursor/deepseek-v4-flash-vision-exp)
- [Devin](https://spiral.ai/integrations/devin/deepseek-v4-flash-vision-exp)
- [dots](https://spiral.ai/integrations/dots/deepseek-v4-flash-vision-exp)
- [Factory Droid](https://spiral.ai/integrations/factory-droid/deepseek-v4-flash-vision-exp)
- [Gemini](https://spiral.ai/integrations/gemini/deepseek-v4-flash-vision-exp)
- [Gemini CLI](https://spiral.ai/integrations/gemini-cli/deepseek-v4-flash-vision-exp)
- [Google Antigravity](https://spiral.ai/integrations/antigravity/deepseek-v4-flash-vision-exp)
- [goose](https://spiral.ai/integrations/goose/deepseek-v4-flash-vision-exp)
- [Grok](https://spiral.ai/integrations/grok/deepseek-v4-flash-vision-exp)
- [Grok Build](https://spiral.ai/integrations/grok-build/deepseek-v4-flash-vision-exp)
- [Hermes](https://spiral.ai/integrations/hermes/deepseek-v4-flash-vision-exp)
- [Instinct](https://spiral.ai/integrations/instinct/deepseek-v4-flash-vision-exp)
- [Manus](https://spiral.ai/integrations/manus/deepseek-v4-flash-vision-exp)
- [Muse](https://spiral.ai/integrations/muse/deepseek-v4-flash-vision-exp)
- [Muse Code](https://spiral.ai/integrations/muse-code/deepseek-v4-flash-vision-exp)
- [n8n](https://spiral.ai/integrations/n8n/deepseek-v4-flash-vision-exp)
- [Notion](https://spiral.ai/integrations/notion/deepseek-v4-flash-vision-exp)
- [OpenClaw](https://spiral.ai/integrations/openclaw/deepseek-v4-flash-vision-exp)
- [opencode](https://spiral.ai/integrations/opencode/deepseek-v4-flash-vision-exp)
- [Perplexity](https://spiral.ai/integrations/perplexity/deepseek-v4-flash-vision-exp)
- [Raycast](https://spiral.ai/integrations/raycast/deepseek-v4-flash-vision-exp)
- [VS Code](https://spiral.ai/integrations/vscode/deepseek-v4-flash-vision-exp)
- [Windsurf](https://spiral.ai/integrations/windsurf/deepseek-v4-flash-vision-exp)
- [Zapier](https://spiral.ai/integrations/zapier/deepseek-v4-flash-vision-exp)

## More Codex guides

- [DeepSeek V4 Flash Latest](https://spiral.ai/integrations/codex/deepseek-v4-flash-latest)
- [DeepSeek V4 Pro 0423](https://spiral.ai/integrations/codex/deepseek-v4-pro-0423)
- [DeepSeek V4 Flash 0731 (batch)](https://spiral.ai/integrations/codex/deepseek-v4-flash-0731-batch)
- [DeepSeek V4 Pro 0813](https://spiral.ai/integrations/codex/deepseek-v4-pro-0813)
- [DeepSeek V4 Flash 0731](https://spiral.ai/integrations/codex/deepseek-v4-flash-0731)
- [DeepSeek V4.1 Flash](https://spiral.ai/integrations/codex/deepseek-v4-1-flash)

## Questions about DeepSeek V4 Flash Vision Exp in Codex

**Can Codex use DeepSeek V4 Flash Vision Exp?**

Yes. Add Spiral to Codex as an MCP server, and Codex can find and run DeepSeek V4 Flash Vision Exp like any built-in tool.

**How much does DeepSeek V4 Flash Vision Exp cost in Codex?**

Up to $0.00705 per run at the default inputs. $0.572 per 1M input tokens, $1.716 per 1M output tokens. You pay from your Spiral balance as you go, with no subscription, and a failed run is not charged. Your Codex plan is separate.

**Do I need an OpenRouter account to use DeepSeek V4 Flash Vision Exp in Codex?**

No. Spiral runs DeepSeek V4 Flash Vision Exp on its own OpenRouter account. You sign in to Spiral once and never paste a key.

**How long does it take to set up DeepSeek V4 Flash Vision Exp in Codex?**

About two minutes: add the hosted server, then sign in.

**Does Codex wait while DeepSeek V4 Flash Vision Exp runs?**

DeepSeek V4 Flash Vision Exp runs in the background. Codex starts the run, checks on it and gives you the text, json when it's ready.

**What is DeepSeek V4 Flash Vision Exp useful for?**

Use it to explain screenshots, interpret charts, and answer questions about document images. Ask for visible evidence alongside each conclusion so you can check the result.

**How does it differ from DeepSeek V4 Flash 0731?**

DeepSeek V4 Flash Vision Exp adds visual understanding while retaining comparable text capabilities. Choose it when the task depends on what appears in an image.

**How much text can it handle?**

OpenRouter documents a 1,048,576-token context window and up to 262,144 completion tokens. Keep the requested answer focused instead of treating that ceiling as a target length.

**Can it generate images or analyze video?**

It returns text and does not accept video input. Use it for written analysis of still images.

**Which reasoning and sampling settings should I test?**

DeepSeek's text agent evaluations used maximum reasoning effort, temperature 1.0, and top_p 0.95. Use those as a reference configuration, then compare results on representative tasks.

