See how ChatGPT, Perplexity and Google AI Overviews describe you today
Free AI Visibility AuditDeepSeek V3.1 is a text-only hybrid reasoning model from DeepSeek that can run in either a thinking mode or a direct non-thinking mode, selected through the prompt template rather than by switching to a different model.
DeepSeekVerified 24/09/2026Released 21/08/202501 / decision
Each figure sits next to the middle of the field, so it can be read as dear or cheap, wide or narrow, rather than floating on its own. Compared against the 126 models we track, not against an absolute standard.
Capability
Not measured
No published benchmark scores yet
Input / 1M tokens
$0.25
field median $0.30
Output / 1M tokens
$0.95
field median $1.25
Context
164K
field median 524K
02 / overview
Four questions, answered separately, because somebody arrives at one of them rather than at the top of the page.
The job it was built for, and the job it is not for.
DeepSeek V3.1 is a large open-weight language model built to handle both quick answers and step-by-step reasoning in one place, with the mode chosen at prompt time. It extends the DeepSeek V3 base with a two-phase long-context training stage, so it is aimed at work that spans long documents and multi-step tasks. It handles text only, so there is no image, audio or video input to build on.
Where it sits in its line, and when a sibling is the better pick.
DeepSeek V3.1 first appeared in August 2025 as the successor to DeepSeek V3, and the headline change over its predecessor is the hybrid thinking and non-thinking behaviour. It is the right pick when you want reasoning available on demand without paying frontier prices for every call, and the wrong pick when your workload involves images, audio or anything beyond text.
The API, the apps it powers, and what its limits let you do.
You reach DeepSeek V3.1 through the DeepSeek API, and because the weights are open it also turns up on third-party inference providers and in self-hosted stacks. The two modes are triggered by prompt templates, so a single integration can switch between fast responses and longer reasoning without re-routing to another model. The context window is large enough to hold long source documents, extended chat history or a sizeable codebase slice in a single request, and the output ceiling is generous enough for full drafts and long structured returns.
What changes because this exists, or why it does not.
The practical change is that reasoning becomes a per-request decision rather than an architectural one. Teams that previously ran a cheap model for routine calls and escalated to a separate reasoning model for hard ones can collapse that into one endpoint, which removes a routing layer and the prompt drift that comes with maintaining two sets of instructions.
Follows DeepSeek V3 0324. Superseded by DeepSeek V3.1 Terminus. See the whole line.
03 / evidence
Split by category, so a strong number never hides a thin evidence base. Verified means we read it on the benchmark's own published results; a provider's claim about its own model is shown and labelled rather than dropped.
No published benchmark scores for this model yet.
We publish a score only where we can link the result to where it was published. Until a benchmark result for this model exists in a source we read, this section stays empty rather than being filled with a provider's marketing figure.
How we decide what counts as evidence04 / ledger
Every published row, grouped by category, each compared with the best published score on the same benchmark. 'Is 64% good' is a question nobody can answer; '26 points behind the leader' is one anybody can.
Nothing in the ledger yet.
Each row here carries a score, the benchmark it came from, what the leading model scored on the same test, and a link to the published result. Rows appear as results are published and read.
How we decide what counts as evidence05 / capability
Where this model is strong, and against how many peers. Ranks are against models with evidence in that category, not against everything we track: ranking against models nobody tested would rank who published, not who is better.
No category scores to shape yet.
A category score is the weighted mean of the benchmarks published for it. With no published rows there is nothing to average, and an empty chart drawn at zero would say something false.
How we decide what counts as evidence06 / cost
List API rates as last read from the provider, with the source on every row, plus every change we have recorded since we started tracking it.
| Charge | Price | Unit | Read on | Source |
|---|---|---|---|---|
| Input | $0.25 | per 1M tokens | 2026-09-24 | Check |
| Output | $0.95 | per 1M tokens | 2026-09-24 | Check |
DeepSeek V3.1 sits at the low end of the market on both input and output, which is the main reason teams pick it over Western reasoning models for high-volume work. Because the weights are open, the API price is effectively a ceiling, self-hosting or a cheaper inference host can bring the running cost down further.
| Workload | Input / month | Output / month | Cost |
|---|---|---|---|
| A small product team | 20M tokens | 5M tokens | $9.75 |
| A busy support assistant | 200M tokens | 40M tokens | $88.00 |
| A document pipeline | 1000M tokens | 100M tokens | $345.00 |
List API rates, no caching and no batch discount, which both providers offer and which change the answer a great deal. Treat these as the ceiling, not the bill.
07 / specs
As listed by the provider's own catalogue and re-read every few hours. Anything absent is absent there too.
| Context window | 163,840 tokens |
|---|---|
| Maximum output | 32,768 tokens |
| Modalities | text |
| Released | 21/08/2025 |
| Status | Current |
| Catalogue identifier | deepseek/deepseek-chat-v3.1 |
08 / lineage
What this model replaced, what replaced it, and what else its provider has in the field.
Also from DeepSeek
09 / line
Every model in this family in release order, so a page from eight months ago says in one glance that two newer ones exist.
Came before
← DeepSeek V3 0324Came after
DeepSeek V3.1 Terminus →Ordered by release date and worked out from the naming, so a new member slots in as soon as its page exists. A retired model keeps its page and its place in the line.
10 / notes
What this model changes for a brand trying to be cited in AI answers, and every change we have logged since it launched.
What it changes for you
DeepSeek V3.1 is cheap enough to sit behind high-volume summarisation, research and answer-generation features, which means more of the tools that describe your category to a buyer will be running on it. Because it is open-weight and widely re-hosted, you cannot assume there is one canonical surface to optimise for, what matters is that your product pages carry clean, extractable claims that any model can lift. Its long context also rewards depth, a model reading your full documentation in one pass will represent you better than one working from a snippet.
Buyers meet DeepSeek V3.1 mostly through the API and through products built on it rather than through a branded consumer assistant, and its open weights mean it sits quietly inside tools whose users never see the model name. It also shows up in self-hosted and on-premise deployments where teams want reasoning without sending data to a US provider.
Nothing published here yet. Changes appear within hours of a provider announcing them.
11 / questions
The things people ask about this model, answered from what is on this page rather than from anywhere else.
Models change what gets cited. We measure whether AI answers name your brand or your competitors across every assistant, and show you what to change.