DeepSeek in any tab: V4 Pro in a browser sidebar

Choose DeepSeek V4 Pro and the sidebar answers with it while it can read the page you are on.

Reasoning questions beside the page

DeepSeek's docs list thinking and non-thinking modes for V4 Pro, with thinking as the default. That suits maths, code explanations and step-by-step questions on the page you are reading, such as a tutorial or a problem set.

One limit to know: V4 Pro does not take images

DeepSeek's docs say image input is supported on DeepSeek Flash and not on DeepSeek V4 Pro, and Copilotly's catalogue marks V4 Pro as text-only. For a chart, screenshot or photo, switch to DeepSeek Flash or to Gemini or Claude in the same menu.

Everything else works the same

Rewrite text in place, web search with cited sources, deep research and chat with files are part of Copilotly, not of any one model. Whether a given file turn runs well on a text-only model is something to test on your own documents.

We have not found one. as of November 2026, DeepSeek's developer documentation covers an API with OpenAI-compatible and Anthropic-compatible endpoints, and DeepSeek runs its own web chat. If DeepSeek publishes an extension later, its own site is where it would be announced.

Beware look-alike extensions

Because there is no official extension, store search results fill up with third-party ones using DeepSeek's name. Check who publishes any extension, what permissions it asks for, and whether the listing says it is unofficial, before you install it and before you sign in anywhere through it.

Where Copilotly fits

• A sidebar on the page. DeepSeek in a panel beside what you are reading, rather than a separate site.
• Every model in one menu. GPT, Claude, Gemini, Grok and Mistral too.
• Any browser that runs the extension, including Firefox.
• Bring your own key. Use your DeepSeek API key and pay DeepSeek directly.

Copilotly is independent of DeepSeek. If you want DeepSeek's own app and nothing else, DeepSeek's site is the direct route.

DeepSeek's pricing page lists two current models (as of November 2026).

Model	Tier	Free plan	Pro plan
DeepSeek Flash	fast	Included, within your token allowance	Included, within your token allowance
DeepSeek V4 Pro	premium	10 messages a day, then DeepSeek Flash	Unlimited

There is no frontier-tier DeepSeek model in the list. Pro is $6.99/week, $19.99/month or $149.99/year, with a 7-day free trial for first-time subscribers.

DeepSeek's own specifications

	deepseek-flash	deepseek-v4-pro
Context length	1M	1M
Thinking mode	Thinking and non-thinking	Thinking and non-thinking
Image input	Supported	Not supported
API output, per 1M tokens (off-peak / peak)	$0.6 / $1.2	$1.98 / $3.96

DeepSeek publishes off-peak and peak rates, with off-peak at half the peak price. Those are API prices you would pay on your own key; Copilotly's plans are priced separately. See pricing.

On the free plan

Yes. The free plan includes DeepSeek Flash and 10 premium-model messages a day for DeepSeek V4 Pro. A free account is required; a card is not. After 10 messages the chat continues on DeepSeek Flash.

On your own DeepSeek API key

Every model on this page can also run on your own API key. Requests made on your own key are never metered by Copilotly, whatever the model's tier, and that includes frontier models on the free plan, which are otherwise Pro-only. You pay the provider directly at their API rates. The steps are on the bring your own key page.

If you already have a DeepSeek API key, connect it and DeepSeek V4 Pro runs unmetered in the sidebar, billed by DeepSeek at the rates above. Create keys from your DeepSeek platform account.

A note on privacy

Your messages go to the model provider to be answered. Avoid pasting confidential material into any assistant unless your own policy allows it, and read the provider's data terms for the model you choose.

Step-by-step reasoning on a budget

DeepSeek V4 Pro is the one to try when you want strong step-by-step reasoning and care about cost, which is how Copilotly's picker describes it. DeepSeek also lists a 1M-token context, so it can take long pages and files as text.

Quick, cheap answers and images

DeepSeek Flash is for fast answers, and it is the DeepSeek model that accepts images, so use it for a screenshot or chart.

Check before you rely on it

As with any model, verify facts and code before relying on them. For claims about recent events, turn on web search and open the sources. When an answer matters, ask Claude or GPT the same question and compare.

Two facts set DeepSeek V4 Pro apart in this list: it has no image input, and its API price is low. DeepSeek's own pricing page lists a 1M-token context for it.

Model	Context window	Image input	Source (Nov 2026)
GPT-6.1 Terra	1.05M	Yes	OpenAI docs
Claude Sonnet 5.5	1M	Yes	Copilotly catalogue
Gemini 4 Argon	1M	Yes, plus audio, video and PDF	Google AI Studio docs
DeepSeek V4 Pro	1M	No (Flash: yes)	DeepSeek docs
Grok 4.7	500K	Yes	xAI docs
Mistral Large 4	1M	Multimodal	Mistral docs

Image input is the practical difference

Every other flagship in the table accepts images. If your question is about a chart, a screenshot or a scanned page, V4 Pro cannot see it. Switch to DeepSeek Flash, which DeepSeek lists as supporting images, or to another provider.

Price, if you use your own key

DeepSeek lists V4 Pro output at $1.98 per million tokens off-peak and $3.96 at peak, with Flash at $0.6 and $1.2. Compare that with the vendors' own price pages before you decide on a key. On Copilotly's plans the tier sets your limit instead.

Ask for the working, then check it

For maths and code, ask DeepSeek V4 Pro to show each step. Recompute one number yourself or run the code. A reasoning model can still reach a confident wrong answer.

Turn images into text if you must use V4 Pro

If you want V4 Pro to work on a table that is an image, copy or retype the table as text, or use Flash for that step.

Use web search for current facts

Training data goes stale. For libraries, prices or news, use web search and open the sources it cites.

Keep sensitive material out

Your messages go to the provider that answers. Leave out passwords, customer data and anything your employer forbids sharing with outside services.

Safety depends on what you send and on your organisation's rules, not only on the model. Some employers and governments restrict particular AI services; check your own policy before using DeepSeek, or any model, for work material.

What goes where

In Copilotly, your messages are sent to the provider that answers, here DeepSeek, to generate the reply. If you use your own DeepSeek key, the request is made on that key and DeepSeek's data terms for the API apply. Read those terms on DeepSeek's site, because they are the provider's to set and to change.

Lower-risk habits

• Leave out passwords, access keys, customer records and anything under a confidentiality agreement.
• Use incognito chats in Copilotly when you do not want a conversation kept in your history.
• Pick a different provider's model from the same menu for material your policy keeps away from DeepSeek.

This is general guidance, not legal or compliance advice. If your company has an AI policy, follow it.

DeepSeek in Firefox works through the Firefox Add-ons listing, and DeepSeek in Opera uses the Chrome Web Store build, as do Brave, Arc and Vivaldi.

Where	Status	How
Chrome	Supported	Chrome Web Store
Edge	Supported	Its own Edge Add-ons listing
Firefox	Supported	Its own Firefox Add-ons listing
Brave	Supported	Installs the Chrome Web Store build
Opera	Supported	Installs the Chrome Web Store build
Arc, Vivaldi	Supported	Install the Chrome Web Store build
Safari	Not yet	There is no Safari build in the App Store
iPhone, Android	Supported	The Copilotly iOS and Android apps, which work on the free plan