Snaxbyte-sized news, on the go
Mistral previewed a trillion-parameter model and promised open weights within the month. It has not said yet what licence those weights will carry. OpenAI put GPT-6 in front of every ChatGPT tier at once. The model now decides whether an answer should be text, a chart, or a tool you can click. Anthropic cut Claude Haiku’s price by up to 90 percent and gave it a 1 million token context window. The cheap-model tier of the price war just got a lot more capable. Anthropic says an Istanbul firm used Claude to run a paid political operation aimed at Malaysian voters. Its report found no sign the campaign reached real communities. The White House ordered federal agencies to stop saying “AI.” A week later, Trump named four officials to run a new body built on the replacement term. Claude for Government left public beta on 30 September 2026. Anthropic is now selling Washington a spending cap, not just a chatbot. OpenAI says it will not release a flagship model it had been testing. The company that races itself on capability now says it will also lose to its own safety bar. DeepSeek shipped a new small model on a new architecture. It undercuts its own flagship on price, not just rivals. Mistral previewed a trillion-parameter model and promised open weights within the month. It has not said yet what licence those weights will carry. OpenAI put GPT-6 in front of every ChatGPT tier at once. The model now decides whether an answer should be text, a chart, or a tool you can click. Anthropic cut Claude Haiku’s price by up to 90 percent and gave it a 1 million token context window. The cheap-model tier of the price war just got a lot more capable. Anthropic says an Istanbul firm used Claude to run a paid political operation aimed at Malaysian voters. Its report found no sign the campaign reached real communities. The White House ordered federal agencies to stop saying “AI.” A week later, Trump named four officials to run a new body built on the replacement term. Claude for Government left public beta on 30 September 2026. Anthropic is now selling Washington a spending cap, not just a chatbot. OpenAI says it will not release a flagship model it had been testing. The company that races itself on capability now says it will also lose to its own safety bar. DeepSeek shipped a new small model on a new architecture. It undercuts its own flagship on price, not just rivals.
Model prices /1M out
GPT-5.6 Luna $1.20 —
Qwen3.8-Max $6.00 —
Mistral Medium 3.5 $7.50 —
Claude Sonnet 5 $10.00 —
GPT-5.6 Terra $12.00 —
GPT-5.6 Sol $20.00 —
Claude Opus 5 $25.00 —
Claude Fable 5 $50.00 —
AI APP PROFILE

DeepSeek

Remarkable capability for nothing, on terms that make it unsuitable for anything you would not publish.

Source review Last reviewed October 7, 2026 By DeepSeek

The facts at a glance

ProviderDeepSeek
TypeReasoning assistant
Free optionYes. The web and mobile apps are free to use
Paid priceNo consumer subscription. Paid access is API only, from USD 0.15 per million input tokens (cache miss, off-peak, under DeepSeek's peak/off-peak schedule)
AvailabilityWeb, iOS, Android, API

Pricing last checked October 7, 2026. Prices change and vary by region. Check the official pricing page ↗

Assessment

What it does well

  • Strong reasoning performance at no cost in the consumer app
  • API pricing is roughly a thirtieth of the Western flagships
  • The underlying V4-Flash weights are MIT licensed, so you can self host and avoid the data question entirely

Where it falls short

  • There is no consumer paid tier, so there is no plan with better data terms to upgrade to
  • DeepSeek states it stores personal data in the People's Republic of China
  • The training opt out is by email to privacy@deepseek.com rather than a setting, which is a meaningful practical barrier
  • DeepSeek warns a significant API price increase is expected

Privacy and data control

DeepSeek states it collects text and voice input, prompts, uploaded files, and photos and uses them to train and improve its machine learning models, stores personal data in the People's Republic of China, and provides a training opt out only by emailing privacy@deepseek.com.

Read the vendor privacy page ↗

The longer read

The honest recommendation

The capability is real and it costs nothing. The data terms are the strictest trade off on this site: content used for training by default, stored in China, with an opt out that requires sending an email. For homework, hobby code, and anything you would happily publish, that is a fair exchange. For client work, personal information, or anything under an NDA, use the MIT licensed weights on your own hardware instead, or use something else.

What to watch

Known limitations and defaults worth knowing before you commit. Each one says where it came from and when we last checked it.

  1. The training opt out is an email, not a setting

    Documented warning
    What can happen
    DeepSeek provides a training opt out only by emailing privacy@deepseek.com. There is no toggle in the product.
    Why it matters
    An opt out that requires composing an email and waiting is one that most users will never complete, which makes the training default effectively permanent in practice.
    Who should be careful
    Anyone who would otherwise flip a training switch and consider the matter closed.
    What to do about it
    If you need the opt out, send the email before you start using it in earnest, and keep the reply.

  2. Personal data is stored in the People's Republic of China

    Documented warning
    What can happen
    DeepSeek states it stores personal data in the People's Republic of China, and collects text and voice input, prompts, uploaded files, and photos to train and improve its models.
    Why it matters
    For many organisations the storage location is a compliance question that is settled before any capability question, and it is not something a setting can change.
    Who should be careful
    Anyone under data residency obligations, and anyone handling customer or employee records.
    What to do about it
    Check the residency requirement before evaluating the product. Where the requirement is binding, use the open weights route and host it yourself rather than the hosted app.

  3. No consumer paid tier, so no better data terms to buy

    Documented warning
    What can happen
    There is no consumer paid plan, which means there is no upgrade that comes with different data handling.
    Why it matters
    On most products, paying is the route to stronger terms. Here that route does not exist, so the free terms are the only terms.
    Who should be careful
    Teams who expected to solve the privacy question by buying the business plan.
    What to do about it
    Decide on the published terms as they stand, or use the API or self-hosted weights where you control the boundary.

How it performs by outcome

We assess products against a job, not a leaderboard. These are the outcomes this entry is compared under.

Worth comparing against

Grok 4.5 Terminal and agentic coding work where token efficiency matters DeepSeek-V4-Flash Teams that need to run a capable model on their own infrastructure Qwen3.8-Max Multi day autonomous coding projects Kimi K3 Research teams that need frontier scale weights they can inspect

How this profile was produced

Source review. Every fact here is drawn from the provider documentation linked below and checked on the date shown. We have not yet run this entry through a controlled test, and no scoring is implied.

Read the full review methodology →

Primary sources