What is DeepSeek V4?
DeepSeek V4 is DeepSeek’s open weights for the preview family; api variants available model for low-cost long-context agents, coding, flexible reasoning and self-hosting experiments. It was released April 24, 2026 preview; V4 Pro GA August 13, 2026. This guide separates documented specifications from vendor benchmark claims and gives teams a practical way to decide whether the model belongs in a real evaluation.
QUICK VERDICT: DeepSeek V4 combines unusually low API prices with a million-token context window and long output limits. Governance, deployment and regional requirements deserve as much attention as benchmark quality.
Field | Verified value |
|---|---|
Provider | DeepSeek |
Release date | April 24, 2026 preview; V4 Pro GA August 13, 2026 |
Availability | Available; V4 Pro and V4 Flash |
License | Open weights for the preview family; API variants available |
Context window | 1 million tokens |
Maximum output | 384,000 tokens maximum |
Modalities | Text input and text output; an experimental Flash Vision variant accepts images |
API pricing | V4 Pro peak rates: $0.044 cached input, $1.32 uncached input and $3.96 output per million tokens; off-peak is half |
Access | DeepSeek app, web, OpenAI-compatible API, Anthropic-compatible API and released weights |
Best fit | low-cost long-context agents, coding, flexible reasoning and self-hosting experiments |
Last verified | September 2, 2026 |
What changed with DeepSeek V4
DeepSeek V4 should not be evaluated as a name change. The important differences are the model’s reasoning controls, tool behavior, context policy, access route and economic profile. Those details determine whether a benchmark result transfers to production.
Area | What changed or matters |
|---|---|
Context | One-million-token window across V4 API variants |
Output | Up to 384K output tokens |
Reasoning | Low, high and max effort for Pro and Flash |
Compatibility | OpenAI Responses and Anthropic-compatible API support |
The correct comparison baseline is DeepSeek V3 family. Teams planning a new deployment should also include DeepSeek V4 Pro and V4 Flash where access allows. Testing only the newest model or only the incumbent hides the migration cost, output-style changes and tool-use regressions that often matter more than a small benchmark gap.
DeepSeek V4 capabilities
The clearest fit is low-cost long-context agents, coding, flexible reasoning and self-hosting experiments. That does not mean every task in those categories should use the model. A production system combines the base model with prompts, retrieval, tools, permissions, memory, retry logic and human review. The model card describes only one layer of that system.
Capability | Practical implication |
|---|---|
Long context | 1 million tokens. Test retrieval accuracy at realistic lengths rather than assuming every token receives equal attention. |
Output capacity | 384,000 tokens maximum. Long output is useful only when the verification process can keep up. |
Modalities | Text input and text output; an experimental Flash Vision variant accepts images. Confirm format-specific accuracy with your own files. |
Agent use | Use explicit tool schemas, narrow permissions, approval gates and recoverable operations. |
Reasoning | Record the exact effort level because cost, latency and quality can change materially. |
For coding, evaluate repository navigation, test creation, regression rate, review burden and recovery after a failed tool call. For research, measure citation correctness, source coverage and whether the model distinguishes evidence from inference. For computer or browser use, record every unsafe click, wrong field, lost state and unapproved action.
Benchmarks and evidence
The evidence below comes from DeepSeek or named launch partners. It is useful for identifying intended strengths, but it is not equivalent to an independent head-to-head test. Prompts, tools, reasoning budgets, sampling, infrastructure and scoring rules can differ.
Evidence | Reported result | How to interpret it |
|---|---|---|
Agent upgrades | DeepSeek reports production gains for V4 Pro | Vendor claim without one universal score |
Cost | Off-peak rates are 50% below peak | Current first-party pricing policy |
Long context | 1M context and 384K maximum output are documented | Capacity specification, not accuracy proof |
Treat benchmark results as a shortlist signal. Before purchasing or migrating, reproduce representative work under one frozen protocol. Keep model snapshots, reasoning effort, tool access and token budgets constant. Report repeated-run pass rates and total cost per accepted result, not a single best attempt.
Pricing, access and deployment
DeepSeek app, web, OpenAI-compatible API, Anthropic-compatible API and released weights. The current pricing reference is V4 Pro peak rates: $0.044 cached input, $1.32 uncached input and $3.96 output per million tokens; off-peak is half. Provider pricing changes frequently and can include cache rates, batch discounts, regional premiums, long-context multipliers, priority processing and tool-call fees. Recheck the official pricing page before budgeting.
Cost driver | What to measure |
|---|---|
Input tokens | Prompt, retrieved context, tool results and repeated history |
Output tokens | Visible answer plus any billable reasoning or generated artifacts |
Caching | Eligible repeated prefixes, cache-read price and expiration policy |
Tools | Search, computer use, code execution and third-party API fees |
Retries | Failed runs, verifier loops and human rework |
Success-adjusted cost | Total spend divided by deliverables that pass review |
A cheaper token price can lose to a more expensive model if it takes more steps, retries more often or produces work that needs heavy correction. Conversely, a frontier model can be wasteful when a smaller model already passes the task rubric. Route by measured task difficulty rather than brand prestige.
DeepSeek V4 limitations
- Peak and off-peak pricing complicates budgeting.
- Open-weight and hosted variants may not behave identically.
- The vision variant is experimental.
- Regional compliance and data handling require separate review.
- Very long outputs can magnify error and verification costs.
High-stakes medical, legal, financial, security and scientific work requires qualified review. Store the exact model identifier, prompt version, tools, source documents and approvals for each consequential run. Build a rollback path before granting write access to repositories, browsers, databases or cloud infrastructure.
Risk | Minimum control |
|---|---|
Hallucination | Require source checks or executable tests |
Prompt injection | Separate untrusted content from instructions and restrict tools |
Over-permission | Use least privilege and approval gates |
Silent model change | Pin snapshots where possible and run regression tests |
Data exposure | Review retention, regional processing and provider terms |
Runaway cost | Set token, time, tool-call and retry budgets |
How to evaluate DeepSeek V4
Create 20 to 50 tasks from real work. Freeze the tasks and rubric before testing. Include easy tasks, normal tasks, edge cases and adversarial inputs. Give each model equivalent tools and enough budget to finish, but cap time and retries. Repeat non-deterministic runs so one lucky result does not decide the winner.
Test area | Record |
|---|---|
Task completion | Pass or fail plus rubric score |
Reliability | Repeated-run success and variance |
Quality | Factuality, instruction adherence and usefulness |
Tool use | Wrong calls, retries, recovery and permission errors |
Efficiency | Wall time, tokens, cache use, tool fees and human review |
Safety | Unsafe actions, injection response and sensitive-data handling |
Migration | Prompt changes, integration work and output-style regressions |
Compare DeepSeek V4 with DeepSeek V3 family and DeepSeek V4 Pro and V4 Flash. Choose the least expensive configuration that meets the acceptance threshold with an adequate safety margin. Re-run the suite after any model snapshot, system prompt, retrieval or tool change.
Frequently asked questions
Is DeepSeek V4 available now?
Available; V4 Pro and V4 Flash. The documented access routes are DeepSeek app, web, OpenAI-compatible API, Anthropic-compatible API and released weights. Availability can differ by region, plan and partner platform.
Is DeepSeek V4 open source?
DeepSeek V4 has an open-weight route. Open weights do not necessarily mean the training data, training code and full software stack are open source. Review the exact license before commercial deployment.
How much does DeepSeek V4 cost?
V4 Pro peak rates: $0.044 cached input, $1.32 uncached input and $3.96 output per million tokens; off-peak is half. Budget with measured end-to-end workloads because token rates alone omit retries, tools, caching, long-context multipliers and review time.
What is DeepSeek V4 best used for?
Its strongest documented fit is low-cost long-context agents, coding, flexible reasoning and self-hosting experiments. Start with a supervised pilot and keep human sign-off for consequential work.
Should I migrate from DeepSeek V3 family?
Only after a side-by-side evaluation. Measure pass rate, total cost, latency, output style, tool reliability and the engineering work required to migrate. A newer model is not automatically the better operational choice.
Related model guides
Start with the broader Best AI Models directory.
Compare with the Kimi K3 guide.
Compare with the Qwen 3.8 guide.
Official sources and update policy
DeepSeek V4 Preview announcement.
DeepSeek V4 Pro GA announcement.
Checked September 2, 2026. We update this guide when the provider changes the model specification, pricing, access, licensing or safety documentation. Vendor benchmarks are attributed and are not presented as independent testing. No vendor payment or affiliate relationship determined inclusion.