# Text-agent results use DeepSeek Harness minimal mode, max effort, temperature=1.0, top_p=0.95.
# Open weights (MIT): https://huggingface.co/deepseek-ai/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"
release_date = "2026-08-21"
last_updated = "2026-09-01"
attachment = true
reasoning = true
temperature = true
tool_call = true
structured_output = true
open_weights = true
license = "MIT"

[limit]
context = 1_000_000
output = 384_000

[modalities]
input = ["text", "image"]
output = ["text"]

[[weights]]
label = "Hugging Face"
url = "https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash-Vision-Exp"

[[benchmarks]]
name = "Terminal-Bench"
score = 83.9
metric = "score"
harness = "DeepSeek Harness minimal mode"
variant = "max effort"
version = "2.1"
source = "https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash-Vision-Exp"

[[benchmarks]]
name = "NL2Repo"
score = 57.7
metric = "score"
harness = "DeepSeek Harness minimal mode"
variant = "max effort"
source = "https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash-Vision-Exp"

[[benchmarks]]
name = "CyberGym"
score = 75.3
metric = "score"
harness = "DeepSeek Harness minimal mode"
variant = "max effort"
source = "https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash-Vision-Exp"

[[benchmarks]]
name = "DeepSWE"
score = 59.3
metric = "score"
harness = "DeepSeek Harness minimal mode"
variant = "max effort"
source = "https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash-Vision-Exp"

[[benchmarks]]
name = "Toolathlon-Verified"
score = 75.9
metric = "score"
harness = "DeepSeek Harness minimal mode"
variant = "max effort"
source = "https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash-Vision-Exp"

[[benchmarks]]
name = "DSBench-Hard"
score = 63.6
metric = "score"
harness = "DeepSeek Harness minimal mode"
variant = "max effort"
dataset = "internal"
source = "https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash-Vision-Exp"

[[benchmarks]]
name = "AutomationBench"
score = 25.7
metric = "score"
harness = "DeepSeek Harness minimal mode"
variant = "max effort"
dataset = "public"
source = "https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash-Vision-Exp"

[[benchmarks]]
name = "ApexBench"
score = 36.5
metric = "pass@1"
source = "https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash-Vision-Exp"

[[benchmarks]]
name = "Agents' Last Exam"
score = 27.3
metric = "score"
source = "https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash-Vision-Exp"

[[benchmarks]]
name = "Chartography"
score = 64.3
metric = "score"
source = "https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash-Vision-Exp"

[[benchmarks]]
name = "ZeroBench"
score = 35
metric = "pass@5"
source = "https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash-Vision-Exp"