DeepSeek V4 Flash Vision Exp

Tool callingStructured outputReasoningImage input

Hugging Face Website

Experimental multimodal DeepSeek V4 Flash model for image understanding, coding, and agentic work

Context: 1.0M
Released: 1mo ago
Parameters: —
Input / 1M tokens: $0.22
Output / 1M tokens: $0.65
Cached input / 1M tokens: $0.00686

Sources:models.devOpenRouter

Benchmarks

DeepSWE is the primary coding benchmark where an exact model evaluation exists. Other sourced evaluations appear below; missing scores are not zero.

Not evaluated in DeepSWE 1.1 as of 2026-09-22. Results are only shown for exact model versions in the official dataset. Official leaderboard.

No benchmark results have been matched to this model yet.

Specifications

Developer: DeepSeek
Model family: DeepSeek
Type: Language model
Predecessor: —
Successor: —
Released: August 21, 2026
Model ID: deepseek/deepseek-v4-flash-vision-exp
Openness: Open weights
License: MIT
Parameters: —
Input modalities: text, image
Output modalities: text
Modality: text+image->text
Tokenizer: DeepSeek
Max output tokens: 262K
Hugging Face: deepseek-ai/DeepSeek-V4-Flash-Vision-Exp
Source freshness: Metadata synced Oct 8, 2026, 10:17 PM UTC

Additional source evidence

Source model: deepseek/deepseek-v4-flash-vision-exp
Source family: deepseek-flash
Last updated: 2026-09-01
Model context: 1M
Model output limit: 384K
Open weights: Yes
File attachments: Yes
Reasoning: Yes
Tool calling: Yes
Structured output: Yes
Temperature control: Yes

Downloadable weights

Hugging Face

Published benchmarks

Benchmark	Score	Evaluation details	Evidence
Terminal-Bench	83.9 · score	DeepSeek Harness minimal mode · max effort · 2.1	Source
NL2Repo	57.7 · score	DeepSeek Harness minimal mode · max effort	Source
CyberGym	75.3 · score	DeepSeek Harness minimal mode · max effort	Source
DeepSWE	59.3 · score	DeepSeek Harness minimal mode · max effort	Source
Toolathlon-Verified	75.9 · score	DeepSeek Harness minimal mode · max effort	Source
DSBench-Hard	63.6 · score	DeepSeek Harness minimal mode · internal · max effort	Source
AutomationBench	25.7 · score	DeepSeek Harness minimal mode · public · max effort	Source
ApexBench	36.5 · pass@1	—	Source
Agents' Last Exam	27.3 · score	—	Source
Chartography	64.3 · score	—	Source
ZeroBench	35 · pass@5	—	Source

Complete source record (JSON)

{
  "id": "deepseek/deepseek-v4-flash-vision-exp",
  "name": "DeepSeek V4 Flash Vision Exp",
  "description": "Experimental multimodal DeepSeek V4 Flash model for image understanding, coding, and agentic work",
  "family": "deepseek-flash",
  "last_updated": "2026-09-01",
  "attachment": true,
  "reasoning": true,
  "tool_call": true,
  "structured_output": true,
  "temperature": true,
  "modalities": {
    "input": [
      "text",
      "image"
    ],
    "output": [
      "text"
    ]
  },
  "limit": {
    "context": 1000000,
    "output": 384000
  },
  "benchmarks": [
    {
      "name": "Terminal-Bench",
      "score": 83.9,
      "source": "https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash-Vision-Exp",
      "metric": "score",
      "harness": "DeepSeek Harness minimal mode",
      "variant": "max effort",
      "version": "2.1"
    },
    {
      "name": "NL2Repo",
      "score": 57.7,
      "source": "https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash-Vision-Exp",
      "metric": "score",
      "harness": "DeepSeek Harness minimal mode",
      "variant": "max effort"
    },
    {
      "name": "CyberGym",
      "score": 75.3,
      "source": "https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash-Vision-Exp",
      "metric": "score",
      "harness": "DeepSeek Harness minimal mode",
      "variant": "max effort"
    },
    {
      "name": "DeepSWE",
      "score": 59.3,
      "source": "https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash-Vision-Exp",
      "metric": "score",
      "harness": "DeepSeek Harness minimal mode",
      "variant": "max effort"
    },
    {
      "name": "Toolathlon-Verified",
      "score": 75.9,
      "source": "https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash-Vision-Exp",
      "metric": "score",
      "harness": "DeepSeek Harness minimal mode",
      "variant": "max effort"
    },
    {
      "name": "DSBench-Hard",
      "score": 63.6,
      "source": "https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash-Vision-Exp",
      "metric": "score",
      "harness": "DeepSeek Harness minimal mode",
      "variant": "max effort",
      "dataset": "internal"
    },
    {
      "name": "AutomationBench",
      "score": 25.7,
      "source": "https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash-Vision-Exp",
      "metric": "score",
      "harness": "DeepSeek Harness minimal mode",
      "variant": "max effort",
      "dataset": "public"
    },
    {
      "name": "ApexBench",
      "score": 36.5,
      "source": "https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash-Vision-Exp",
      "metric": "pass@1"
    },
    {
      "name": "Agents' Last Exam",
      "score": 27.3,
      "source": "https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash-Vision-Exp",
      "metric": "score"
    },
    {
      "name": "Chartography",
      "score": 64.3,
      "source": "https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash-Vision-Exp",
      "metric": "score"
    },
    {
      "name": "ZeroBench",
      "score": 35,
      "source": "https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash-Vision-Exp",
      "metric": "pass@5"
    }
  ],
  "release_date": "2026-08-21",
  "open_weights": true,
  "license": "MIT",
  "weights": [
    {
      "label": "Hugging Face",
      "url": "https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash-Vision-Exp"
    }
  ]
}

Infrastructure

Provider availability

All providers

Provider	Endpoint	Context	Input /M tokens	Output /M tokens	Cached input /M tokens	Cache write /M tokens	Uptime 1d	Data handling
DeepInfra	deepinfra/fp8	1.0M	$0.22	$0.65	$0.00686	—	100.00%	ZDR: Yes Training: Unknown Prompt logging: Unknown Retention: Unknown ZDR source Privacy policy Checked 2026-10-15
GMICloud	gmicloud/fp8	1.0M	$0.44	$1.32	$0.01	—	100.00%	ZDR: No Training: Unknown Prompt logging: Unknown Retention: Unknown ZDR source Privacy policy Checked 2026-10-15
SiliconFlow	siliconflow/fp8	1.0M	$0.44	$1.32	$0.03	—	98.73%	ZDR: Yes Training: Unknown Prompt logging: Unknown Retention: Unknown ZDR source Privacy policy Checked 2026-10-15

Provider data:OpenRouter

Deployment graph

Can I run this?

Vertical Tabs only shows compatibility when it has structured evidence tying this exact model to hardware, runtime, quantization, memory use, and observed performance.

Browse hardware