DeepSeek's V4-Flash-Vision rivals Opus 4.8 on agent tests

Share
DeepSeek's V4-Flash-Vision rivals Opus 4.8 on agent tests

DeepSeek keeps moving fast on its low-cost track, and its latest move pairs image understanding with the company's signature price. Today it unveiled V4-Flash-Vision Exp, an experimental multimodal model that adds vision to its V4 Flash series, and DeepSeek says it stacks up against Anthropic's flagship on visual-agent benchmarks.

DeepSeek's V4-Flash-Vision Exp nearly matches Opus 4.8 on multimodal agent tests.

The new model, live on DeepSeek's API platform as deepseek-v4-flash-vision-exp, layers vision on top of the text reasoning that V4 Flash already shipped. On DeepSeek's own internal multimodal-agent benchmarks, the company says the vision variant scores close to Claude Opus 4.8 — and in at least one widely shared framing it edged ahead. The Decoder notes the model is built for agent workflows: it can describe images, pull text out of screenshots, and analyze diagrams, handling JPEG, PNG, GIF, and WebP formats. It's designed to drop into existing agent frameworks and combine what it sees with tool use, the use case DeepSeek is clearly betting on.

On the numbers, V4-Flash-Vision Exp beats its predecessor across six of seven text benchmarks, and the biggest jump is in image analysis. DeepSeek measured more than 10% improvement on two of four visual tests, and bested Opus 4.8 on two specific evals: ALE, a suite of more than 1,000 multi-step tasks where models interact with apps, write code, and interpret files, and ZeroBench, a set of 100 intentionally hard image-analysis tasks. For now the bargain framing is the whole pitch — images are tokenized at up to 384 tokens each and billed at V4-Flash pricing, and DeepSeek says a fuller "Vision-Full" release is coming down the line, with nothing announced yet on open weights.

DeepSeek is reportedly preparing a listing that CNBC-tracked reports call a potential record IPO.

The model drop lands against a steady drumbeat that DeepSeek is readying the biggest AI IPO in history, reportedly in the tens of billions of dollars in valuation. A 21财经 report framed it as an AI giant about to file its listing application, with coverage pointing to a number that would make it the largest-ever AI company offering — several outlets put the buzz around an $86 billion valuation. DeepSeek hasn't confirmed specifics, so treat the number as reported rather than final, but the direction is clear: the outfit best known for cutting the cost of frontier AI is now looking to raise serious public capital. That pairs oddly but powerfully with today's release — ruthless pricing as a pre-IPO growth story.

What to watch: whether the open-weights version arrives, since that's the move that would really put pressure on the paid tier's moat.

Open weights on a model this cheap would reset the multimodal race again — do you think DeepSeek releases them, or keeps vision closed for the IPO? Write to us via our contact page.

Get SiliconNoon in your inbox

One email a day with the AI stories that matter — what happened, the numbers, and why it matters. No spam, unsubscribe in one click.

Read more

Shanda founder's 27B local model takes the agent fight to DeepSeek

Shanda founder's 27B local model takes the agent fight to DeepSeek

StartLux's first preview beats much bigger models on agent benchmarks, VAST ships production-ready 3D scene generation from a single image, and Hugging Face gives coding agents a memory the user controls. Shanda co-founder Chen Danian's StartLux unveiled a 27-billion-parameter local agent model on Wednesday that, by its own numbers, runs within striking distance of DeepSeek's trillion-parameter flagship on tool-using tasks.

Independent run confirms DeepSeek V4 Flash's 82.7% score

Independent run confirms DeepSeek V4 Flash's 82.7% score

DeepSeek's flagship agentic benchmark claim just survived its first outside audit — and the numbers held up to the decimal point. An independent, fully public evaluation run reproduced DeepSeek V4 Flash 0731's 82.7% score on Terminal-Bench 2.1, matching the lab's claim exactly. The run, published by the author of the Ante evaluation harness (disclosure noted in the post), used a public, downloadable harness rather than DeepSeek's own tooling: 368 successful trials out of 445, 82.7% accuracy…

SpaceXAI's Grok Voice 3.0 tops the speech index at $0.08 a minute

SpaceXAI's Grok Voice 3.0 tops the speech index at $0.08 a minute

Voice agents are having their price-war moment: a new frontier speech model landed today at the same per-minute rate as the one it replaces, and it is already answering the phones at Starlink. Meanwhile the AI buildout keeps pulling power plants out of the ground, and a license-plate camera lawsuit tests whether surveillance guardrails mean anything. SpaceXAI released Grok Voice 3.0, a speech-to-speech model it says ranks first on Artificial Analysis' independent Speech-to-Speech Index with a…

Best model scores 12% on SemiAnalysis's new frontier benchmark

Best model scores 12% on SemiAnalysis's new frontier benchmark

Every few months a new leaderboard arrives claiming the models have finally been measured on real work. This one arrives with a number nobody can spin: the best model in the world scores 12%. SemiAnalysis released GDPx on Sunday, a 206-task benchmark split across 11 domains it considers the places frontier intelligence will actually earn its trillion dollars — chip design, hardware design, cybersecurity, biology, finance, legal, quant trading, recursive self-improvement, materials science…

vivo's 3B on-device model tops phone LLM chart, nears cloud scores

vivo's 3B on-device model tops phone LLM chart, nears cloud scores

A SuperCLUE benchmark released Friday found that a small model living entirely on a phone is competitive with the cloud. vivo's BlueLM 3.5 Nano 3B took the top spot on the group's on-device leaderboard with 89.86 points, edging past the field while scoring within a small margin of flagship cloud systems it wasn't even competing against. vivo's on-device BlueLM 3B topped SuperCLUE's phone-LLM chart — and nearly matched the big cloud models.

DeepMind alumni's Inherent says its Faraday agent beats GPT-5.5 at reproducing research

DeepMind alumni's Inherent says its Faraday agent beats GPT-5.5 at reproducing research

A London lab founded by Google DeepMind alumni says its new agent, Faraday, outperformed Claude Opus 4.8 and GPT-5.5 at independently reproducing the findings of published scientific papers — and it did it on a 27B-parameter Qwen 3.6 model, a fraction of the frontier systems' size. Inherent emerged from stealth just weeks ago with a $50 million seed round, and this benchmark claim is its first real proof point. The company is upfront that the task was narrow: replication, not discovery.