DoubleChecks

What was claimed

DeepSeek unveiled new multimodal AI model that approaches Anthropic’s Opus 4.8 in multimodal agent performance while retaining top text/reasoning capabilities

Our verdict

Accurate

DeepSeek announced the experimental DeepSeek-V4-Flash-Vision-Exp multimodal model in its API docs and in coverage from multiple outlets.

1 of 3 AI systems agree20 sources citedChecked Aug 22, 2026

Check your own claim

Paste any statement, headline, or AI answer — 3 independent AIs verify it in seconds, with sources.

Key findings

The model approaches Anthropic's Opus 4.8 in multimodal agent performance

Verified95%
1 AI checked

The new DeepSeek model approaches Anthropic’s Opus 4.8 in multimodal agent performance.

Verified95%
4 of 7 AIs agree·ChatGPT: Misleading, ChatGPT: Misleading, Perplexity: Misleading

The model retains top text/reasoning capabilities

Verified98%
1 AI checked

Verified against 20 sources

scmp.compureai.comapi-docs.deepseek.comsiliconangle.comtechcrunch.comdeepseek.aidocs.clore.aiinvesting.comgithub.comarxiv.orgfinance.yahoo.comdev.tomindstudio.aiblog.roboflow.comarxiv.orgapi-docs.deepseek.comreddit.comhuggingface.cojuliangoldie.comaxios.com

Detailed Analysis

The statement is not supported by the most authoritative sources in the provided results. DeepSeek’s cited release is a multimodal version of V4 Flash Vision, and sources say it comes close to Opus 4.8 on multimodal agent benchmarks while preserving text/reasoning ability, but none support the phrase “new multimodal AI model” as written with the specific wording “approaches Anthropic’s Opus 4.8 in multimodal agent performance while retaining top text/reasoning capabilities” as a verified standalone claim. The wording is also misleading because the model is described as experimental and benchmark-specific, not broadly established as a top overall multimodal model.

Why this verdict

  • The statement is not supported by the most authoritative sources in the provided results.
  • DeepSeek’s cited release is a multimodal version of V4 Flash Vision, and sources say it comes close to Opus 4.8 on multimodal agent benchmarks while preserving text/reasoning ability, but none support the phrase “new multimodal AI model” as written with the specific wording “approaches Anthropic’s Opus 4.8 in multimodal agent performance while retaining top text/reasoning capabilities” as a verified standalone claim.
  • The wording is also misleading because the model is described as experimental and benchmark-specific, not broadly established as a top overall multimodal model.

Claims checked

The model retains top text/reasoning capabilities.

Misleading79%
The sources say it matches DeepSeek-V4-Flash on text capabilities, including reasoning and agent tasks. They do not prove it has 'top' text/reasoning capabilities overall.

This claim may mislead readers without added context.

The model approaches Anthropic’s Opus 4.8 in multimodal agent performance.

Verified92%
DeepSeek says its multimodal agent capabilities are close to Opus-4.8, and reporting repeats that claim.

DeepSeek unveiled a new multimodal AI model.

Verified95%
DeepSeek announced the experimental DeepSeek-V4-Flash-Vision-Exp multimodal model in its API docs and in coverage from multiple outlets.

What other AI think

The claim accurately reflects DeepSeek's official announcement about its new multimodal model. All major components are factually correct: DeepSeek did unveil a multimodal model (DeepSeek-V4-Flash-Vision-Exp), it does approach Anthropic's Opus 4.8 in multimodal agent performance, and it does retain top text/reasoning capabilities. The statement is well-supported by multiple recent sources.

Why this verdict

  • The claim accurately reflects DeepSeek's official announcement about its new multimodal model.
  • All major components are factually correct: DeepSeek did unveil a multimodal model (DeepSeek-V4-Flash-Vision-Exp), it does approach Anthropic's Opus 4.8 in multimodal agent performance, and it does retain top text/reasoning capabilities.
  • The statement is well-supported by multiple recent sources.

Claims checked

The model approaches Anthropic's Opus 4.8 in multimodal agent performance

Verified95%
DeepSeek's official statements and multiple news sources confirm the model brings multimodal agent performance 'close to' Opus 4.8. On DeepSeek's own benchmarks, it wins 3 of 11 multimodal agentic tests against Opus 4.8, supporting the 'approaches' characterization.

The model retains top text/reasoning capabilities

Verified98%
DeepSeek explicitly stated the model 'matches DeepSeek-V4-Flash on text capabilities—including agents, reasoning, and world knowledge.' Multiple sources confirm text and reasoning capabilities are preserved from the original V4-Flash model.

DeepSeek unveiled new multimodal AI model

Verified99%
DeepSeek released DeepSeek-V4-Flash-Vision-Exp, an experimental multimodal model, on August 21, 2026. Multiple sources confirm this announcement.
Parts of the statement are supported by public releases and early comparisons. Performance claims versus Anthropic Opus 4.8 are contested across independent benchmarks and vendor reports, so they are not conclusively established.

Why this verdict

  • Parts of the statement are supported by public releases and early comparisons.
  • Performance claims versus Anthropic Opus 4.8 are contested across independent benchmarks and vendor reports, so they are not conclusively established.

Claims checked

The new DeepSeek model approaches Anthropic’s Opus 4.8 in multimodal agent performance.

Misleading60%
Some independent tests and blog comparisons show DeepSeek close to Opus 4.8 on specific agent benchmarks, but other independent evaluations report DeepSeek underperforming; there is no clear, consistent consensus that it reliably 'approaches' Opus 4.8 across multimodal agent workloads.

This claim may mislead readers without added context.

The new DeepSeek model retains top text/reasoning capabilities.

Misleading60%
Benchmarks show DeepSeek is competitive on some text and reasoning tasks, but independent results are mixed and do not uniformly place it as a top-tier leader compared to models like Opus 4.8, GPT-5.5, or GLM-5.2.

This claim may mislead readers without added context.

DeepSeek unveiled a new multimodal AI model.

Verified95%
DeepSeek has recently published and rolled out V4 variants (e.g., V4 Pro / V4 Flash) with multimodal and agent-focused features according to vendor materials and multiple third-party reports.

Share this result

Try DoubleChecks