cd /news/large-language-models/llm-pricing-this-week-deepseek-quiet… · home topics large-language-models article
[ARTICLE · art-107697] src=dev.to ↗ pub= topic=large-language-models verified=true sentiment=· neutral

LLM Pricing This Week: DeepSeek Quietly Drops a Vision Model

LLM Price Watch reports a quiet week with no pricing changes across Claude, GPT, Gemini, DeepSeek, and Grok. The only notable development is the appearance of an experimental vision model, deepseek/deepseek-v4-flash-vision-exp, on OpenRouter, which could signal aggressive pricing in DeepSeek's flash tier.

read2 min views1 publishedAug 23, 2026

Weekly digest from LLM Price Watch — we track live pricing for Claude, GPT, Gemini, DeepSeek, and Grok so you don't have to.

It was a quiet week on the pricing front. No cuts, no hikes, nothing moved across the five providers we track. If you're mid-project and budgeting around current rates, your spreadsheet is still valid.

The only thing worth noting this week is a new model showing up in the wild.

On August 22nd, our tracker spotted deepseek/deepseek-v4-flash-vision-exp appearing on OpenRouter for the first time. It's sitting under DeepSeek's lineup, and the name tells you a few things:

The exp

tag is worth paying attention to. Experimental models from any provider can change behavior, pricing, or disappear entirely without much notice. If you're evaluating this for something that needs to stay stable, treat it as a preview rather than a dependency.

That said, DeepSeek's "flash" tier has generally meant aggressive pricing in the past, and a vision-capable model in that tier is interesting. A lot of use cases — document parsing, receipt extraction, basic image captioning — don't need the heaviest multimodal model available. If DeepSeek prices this the way they've priced their other flash variants, it could be worth benchmarking against GPT-4o mini or Gemini Flash for vision tasks.

We don't have confirmed pricing locked in for this model yet since it just appeared. We'll update the tracker as that settles.

Honestly, a week with no price changes and one experimental model drop is pretty normal. The pace of model releases has been fast enough over the past year that it's easy to assume something significant happens every week, but sometimes the answer is just: nothing moved, here's what's new.

A few things I'd keep in mind heading into next week:

DeepSeek continues to ship fast. Whether or not deepseek-v4-flash-vision-exp

turns into a production model worth using, the cadence from DeepSeek has been consistent. They tend to iterate in public, so experimental models often do graduate to stable relatively quickly.

Vision is getting commoditized. A year ago, image input was a premium feature. Now it's showing up in flash/lite tiers across providers. If you're paying for a heavier model primarily because you need vision and assumed cheaper options didn't have it, it's worth re-checking the landscape.

No price changes this week doesn't mean no price pressure. Several providers have cut prices in 2025, and competition hasn't slowed. Stable prices one week doesn't signal a plateau — it just means nothing happened this particular week.

That's the week. One new experimental model from DeepSeek, everything else held steady. If you want to keep an eye on pricing as it shifts — especially for DeepSeek's new addition once pricing confirms — the tracker is running daily checks.

── more in #large-language-models 4 stories · sorted by recency
── more on @deepseek 3 stories trending now
sponsored brought to you by zahid.host 4,200+ EU-deployed projects
reading about agents? ship yours in a single git push.

Run your AI side-project on zahid.host

EU-based hosting, git-push deploys, automatic HTTPS, no cold starts. Free tier with a custom domain — perfect for shipping the agent you just read about.

$git push zahid main
Live at https://your-agent.zahid.host
Get free account → Pricing
from €0/mo · no card required
LIVE [news/llm-pricing-this-wee…] indexed:0 read:2min 2026-08-23 ·