New Sign up free, 10 calls on us. Up to $1, no card needed.

DeepSeek V4 Pro

Released 2026-04-24

chatCodeReasoningTool callingPrompt caching

DeepSeek V4 Pro is the flagship of the open-source DeepSeek-V4 series: a Mixture-of-Experts model with 1.6T total parameters and 49B activated, pre-trained on more than 32T tokens.

Input
text $1.32/M
Output
text $3.96/M
Cache read
$0.132/M
Context
1M
vs GPT-4o
~74% cheaper

Benchmarks

Above average4 / 10
DeepSeek V4 Pro other models measured peer average no peer scored higher
SWE-Bench Pro
59%
CoWorkBench max
66.3%
Humanity's Last Exam no tools
37.7%
MCP-Mark
57.1%

Vendor-published: Alibaba (Qwen) Anthropic DeepSeek Google MiniMax Moonshot OpenAI Tencent Z.ai

Price in context

Where the price sits among 60 comparable models

Input$1.32/M
$0.05 · Qwen3 VL Flash GPT-5.4 Pro · $30
Output$3.96/M
$0.275 · DeepSeek V4 Flash GPT-5.4 Pro · $180
Cached read$0.132/M
$0.0028 · DeepSeek V4 Flash GPT-5.4 Pro · $15

The bar shows how this model’s price compares with every other model of the same kind on Synthorai. The cheapest and the most expensive are named at each end. These are base rates; batch, region and cache-write discounts are on the pricing page.

Specs & limits

Tokens

Context window (vendor spec) 1,000,000
Max output (vendor spec) 393,216

Prompt caching

How it caches automatic
Lifetime no fixed TTL (evicted when unused)

Thinking

Vendor control thinking.type + reasoning_effort
Accepted values thinking.type enabled · disabled; reasoning_effort high · max (low and medium map to high, xhigh maps to max)
Default enabled, with reasoning_effort high; some complex agent requests are automatically set to max applied when the request sets nothing
Can be turned off Yes
Thinking behaviour Trace returns in reasoning_content, and on a turn where the model called a tool it must be passed back in every later turn; otherwise it is ignored. Thinking mode also silently ignores temperature, top_p, presence_penalty and frequency_penalty. DeepSeek publishes this default for its own API; other platforms that host the model document different defaults, so confirm against the platform you are routed to.
Parameter reasoning_effort
Values minimal · low · medium · high the gateway's parameter surface - the vendor mapping above applies

Model

Modalities text → text
Parameters 1.6T total · 49B active MoE
License MIT
  • 1.6T/49B active MoE flagship, 1M context, Thinking + Non-Thinking modes
  • hybrid sparse attention cuts single-token FLOPs to 27% and KV cache to 10% vs V3.2 at 1M context

per DeepSeek official docs ↗

One prompt, measured through the gateway

PROMPT Explain to a CFO, in exactly three sentences, why our LLM bill doubled after we enabled reasoning. CHECK exactly 3 sentences

DeepSeek V4 Pro passed · 3 sentences

Enabling reasoning causes the model to generate a lengthy internal chain‑of‑thought before it delivers a visible answer, and we are billed for every one of those hidden reasoning tokens just as we are for the final output. In most interactions, the volume of these unseen tokens at least equals the visible response, effectively doubling the total tokens consumed per query. That doubled token consumption translates directly into a doubled per‑query cost, which is why our overall LLM bill doubled.

out 641 tok (+545 thinking) latency 11.5 s

Instruction following (exactly three sentences - countable), audience shaping (CFO register), and the hidden-thinking billing gap the token meters below expose.

PROMPT This function is supposed to return the longest run of consecutive equal items, but callers report it is off by one on single-element inputs and misses runs that end at the last index. Fix it and explain the bug in one paragraph. CHECK fix passes tests

DeepSeek V4 Pro passed · 8/8 cases

```python def longest_run(xs): if not xs: return 0 best = 1 cur = 1 for i in range(1, len(xs)): if xs[i] == xs[i - 1]: cur += 1 best = max(best, cur) else: cur = 1 return best ``` **Explanation:** The bug stems from initializing `cur = 0` and only incrementing it when two adjacent elements match, which counts the number of *additional* equal elements rather than the actual run length (which is `cur + 1`). As a result, the returned `best` is always one less than the true longest run—most obviously returning 0 for a single-element list instead of 1. […]

out 2418 tok (+2131 thinking) latency 35.6 s

Whether the fix is actually correct (runnable), explanation density, and token efficiency on a bounded task.

PROMPT Extract a JSON object with fields {company, quarter, revenue_usd_m, growth_pct, guidance} from this text. Use null for anything not stated; add no other fields. CHECK valid JSON, schema exact

DeepSeek V4 Pro passed · 5/5 fields, guidance null

{ "company": "Meridian Robotics", "quarter": "September quarter", "revenue_usd_m": 412, "growth_pct": "eleven and change", "guidance": null }

out 1204 tok (+1153 thinking) latency 19.9 s

Schema obedience (no invented fields), hallucination pressure (guidance is explicitly withheld), and structured-output path differences.

PROMPT Write a 120-word product announcement for an API gateway feature that caches prompts across providers. Forbidden words: "seamless", "unlock", "game-changer", "revolutionize", "empower". Exactly one sentence must be a question. CHECK 120 words, 0 banned words

DeepSeek V4 Pro no answer to grade · no answer text within 16,384 tokens (all of it went to thinking)

The model returned no answer text - the whole token budget went to hidden thinking.

out 8193 tok (+8192 thinking) latency 106.2 s

Constraint obedience (word budget, banned-word list, the single question), style fingerprint, and length control.

Use DeepSeek V4 Pro in 30 seconds

OpenAI-compatible: swap the base_url, keep your SDK. POST /v1/chat/completions

from openai import OpenAI

client = OpenAI(
    base_url="https://synthorai.io/v1",
    api_key="sk-syn-...",
)

resp = client.chat.completions.create(
    model="deepseek-v4-pro",
    messages=[{"role": "user", "content": "Summarize this diff"}],
    reasoning_effort="medium",
)
print(resp.choices[0].message.content)

About DeepSeek V4 Pro

  • It supports a 1M-token context window and dual thinking/non-thinking modes, and DeepSeek highlights strong agentic coding, broad world knowledge, and reasoning across mathematics, STEM, and programming.
  • The model card names the pieces behind that: a hybrid attention stack pairing Compressed Sparse Attention with Heavily Compressed Attention, Manifold-Constrained Hyper-Connections, and the Muon optimizer, with MoE expert weights held in FP4 and most other parameters in FP8.
  • New sparse-attention designs reduce long-context inference FLOPs and KV cache to a fraction of DeepSeek-V3.2's.
  • Thinking is a request parameter rather than a separate model name: a thinking object switches it on or off, a reasoning_effort setting selects the documented Non-think, Think High and Think Max behaviours, and the chain of thought comes back in reasoning_content beside the answer.
  • DeepSeek's guide adds a rule worth coding for: on any turn where the model called a tool, that turn's reasoning_content must be passed back in later turns.
  • Function calling and JSON output both work with thinking on, output runs to 384K tokens, and prefix caching is automatic.
  • Weights are openly released under the MIT License.
  • On Synthorai, DeepSeek V4 Pro is available through the OpenAI-compatible chat completions endpoint.

FAQ

Is the DeepSeek V4 Pro API free to try?

Yes: new accounts get 10 trial calls and up to $1 in free credit, no card required. At $1.32/M input tokens, that credit alone covers roughly 94 requests of ~8K tokens against DeepSeek V4 Pro.

What is DeepSeek V4 Pro best at?

1.6T-parameter MoE with 49B activated; pre-trained on more than 32T tokens; 1M context with dual thinking modes. See the About section for the full picture from the vendor's own release notes.

How much does DeepSeek V4 Pro cost?

DeepSeek V4 Pro costs $1.32 per million input tokens and $3.96 per million output tokens on Synthorai. That is the provider's list price, with no platform markup. Cached input tokens bill at $0.132/M.

Does DeepSeek V4 Pro support prompt caching?

Yes, automatically: DeepSeek-served prompts cache with no code changes. Cached input tokens bill at $0.132/M vs $1.32/M uncached (TTL no fixed TTL (evicted when unused)). Prompt caching guide →

How do I get access to DeepSeek V4 Pro?

Point your existing OpenAI SDK at base_url="https://synthorai.io/v1", set model="deepseek-v4-pro", and you're done. One API key covers every model on the gateway.

Is DeepSeek V4 Pro open source?

Yes: the weights are published under the MIT license (official repository linked in the About section). Or skip the GPUs: the hosted version here is pay-as-you-go with no infrastructure to run. Running open-weight models →

Related models

Compare

Every value on this page is transcribed from the vendor's own documentation, linked above, and carries the date it was checked. Prices are compared across the catalogue; specification values that vendors define differently are shown with the difference stated rather than charted. Nothing here is measured by us, and nothing is scored.

Get your API key Compare your cost →