Page last updated on
DeepSeek-V4-Pro
By DeepSeek
Flagship MoE model with 1.6T total parameters and 1M context. Built for advanced reasoning, coding, and long-horizon agent workflows.
Specifications
- Total parameters
- 1.6T (49B active)
- Active parameters
- 49B
- Architecture
- MoE
- Architecture class
- DeepseekV4ForCausalLM
- Attention
- Compressed Sparse Attention (CSA) + Heavily Compressed Attention (HCA)
- Context window
- 1M
- Vocab size
- 129,280
- Precision
- FP4 + FP8 Mixed
- Modality
- Text
- License
- MIT
- Released
- April 2026
- Recommended hardware
- 8× H2008× B200
- Best for
- Advanced reasoning, coding, and long-horizon agent workflows
Good to know
-
This page covers the DeepSeek-V4-Pro preview, released on April 24, 2026. It has been superseded. The official build, DeepSeek-V4-Pro-0813, went live on August 12, 2026 and now serves the deepseek-v4-pro API name.
-
DeepSeek shipped the official version of the smaller sibling model first. On July 31, 2026, DeepSeek-V4-Flash-0731 entered public beta on the API. The V4-Pro API and the models behind the DeepSeek app and website were not changed by that update, and DeepSeek said the official V4-Pro version will follow as soon as possible.
-
V4-Flash-0731 keeps the same architecture and parameter count as V4-Flash-Preview, at 284B total and 13B active. Only post-training was redone. DeepSeek reports that the result beats DeepSeek-V4-Pro-Preview by a wide margin on several agent benchmarks, including 82.7 on Terminal Bench 2.1, 54.4 on DeepSWE, 76.7 on Cybergym, and 70.3 on Toolathlon verified. Because the official V4-Pro build is not out yet, that comparison is against the preview, not the finished flagship.
-
V4-Flash-0731 also adds native support for the Responses API format, with a specific adaptation for Codex, the OpenAI coding assistant that ships as a CLI, a ChatGPT desktop app, and a VS Code extension. V4-Pro still runs through the Chat Completions and Anthropic-compatible endpoints.
Architecture
The defining choice of DeepSeek-V4-Pro is the attention stack. It interleaves Compressed Sparse Attention (CSA) with Heavily Compressed Attention (HCA).
- CSA compresses blocks of the KV cache and attends to only a few relevant entries.
- HCA squeezes much larger spans into single entries. That way the model keeps a full 1M-token history within reach, without paying dense-attention memory for it.
The payoff is efficiency at long context. At 1M tokens, DeepSeek reports that V4-Pro uses 27% of the per-token inference FLOPs and 10% of the KV-cache memory of DeepSeek-V3.2. The model also adds manifold-constrained hyper-connections (mHC) for cleaner residual flow. It trains with the Muon optimizer on more than 32T tokens. The released checkpoint uses mixed-precision quantization: MoE expert weights in FP4, and most of the rest in FP8.
One thing to watch while V4-Pro sits in preview: the July 31, 2026 official release of DeepSeek-V4-Flash suggests DeepSeek is treating post-training as the main lever for agent work. V4-Flash-0731 shares the architecture and the 284B/13B parameter split of the preview build, and the only change was a fresh post-training run. That alone was enough to push the cheap model past the preview of the expensive one on Terminal Bench 2.1, DeepSWE, Cybergym, and Toolathlon. The same treatment applied to a 1.6T backbone is the interesting part of the pending V4-Pro release.
For more models like this one, browse the full open source LLM directory or read the open source LLM ecosystem statistics.
Benchmarks
DeepSeek published a full benchmark set for V4-Pro. It is strongest at coding and tool use. It leads on LiveCodeBench (93.5%) and Codeforces (3206). It sits within a point of the best score on SWE-bench Verified (80.6%). On broad knowledge tests like GPQA Diamond and HLE, it gives a little ground to the closed frontier models.
DS-V4-Pro Max vs frontier models
Scores published by DeepSeek in the V4-Pro model card. DS-V4-Pro Max runs in Think Max mode. Higher is better on every benchmark shown. Scores marked n/a were not reported.
Each model keeps the same color across every benchmark, stacked top to bottom in this order. Hover or focus a row to read all six scores.
Separate axis. Rating is not a percentage, so this row is plotted as points on a fitted scale starting at 2,900.
Separate axis. Elo is not a percentage, so this row is plotted as points on a fitted scale starting at 1,000.
- Opus-4.6 Max
- GPT-5.4 xHigh
- Gemini-3.1-Pro High
- K2.6 Thinking
- GLM-5.1 Thinking
- DS-V4-Pro Max
View all scores as a table
| Benchmark | Opus-4.6 Max | GPT-5.4 xHigh | Gemini-3.1-Pro High | K2.6 Thinking | GLM-5.1 Thinking | DS-V4-Pro Max |
|---|---|---|---|---|---|---|
| Knowledge & Reasoning | ||||||
| MMLU-Pro (EM) | 89.1 | 87.5 | 91.0 | 87.1 | 86.0 | 87.5 |
| SimpleQA-Verified (Pass@1) | 46.2 | 45.3 | 75.6 | 36.9 | 38.1 | 57.9 |
| Chinese-SimpleQA (Pass@1) | 76.4 | 76.8 | 85.9 | 75.9 | 75.0 | 84.4 |
| GPQA Diamond (Pass@1) | 91.3 | 93.0 | 94.3 | 90.5 | 86.2 | 90.1 |
| HLE (Pass@1) | 40.0 | 39.8 | 44.4 | 36.4 | 34.7 | 37.7 |
| LiveCodeBench (Pass@1) | 88.8 | n/a | 91.7 | 89.6 | n/a | 93.5 |
| HMMT 2026 Feb (Pass@1) | 96.2 | 97.7 | 94.7 | 92.7 | 89.4 | 95.2 |
| IMOAnswerBench (Pass@1) | 75.3 | 91.4 | 81.0 | 86.0 | 83.8 | 89.8 |
| Apex (Pass@1) | 34.5 | 54.1 | 60.9 | 24.0 | 11.5 | 38.3 |
| Apex Shortlist (Pass@1) | 85.9 | 78.1 | 89.1 | 75.5 | 72.4 | 90.2 |
| Codeforces (Rating) | n/a | 3,168 | 3,052 | n/a | n/a | 3,206 |
| Long Context | ||||||
| MRCR 1M (MMR) | 92.9 | n/a | 76.3 | n/a | n/a | 83.5 |
| CorpusQA 1M (ACC) | 71.7 | n/a | 53.8 | n/a | n/a | 62.0 |
| Agentic | ||||||
| Terminal Bench 2.0 (Acc) | 65.4 | 75.1 | 68.5 | 66.7 | 63.5 | 67.9 |
| SWE Verified (Resolved) | 80.8 | n/a | 80.6 | 80.2 | n/a | 80.6 |
| SWE Pro (Resolved) | 57.3 | 57.7 | 54.2 | 58.6 | 58.4 | 55.4 |
| SWE Multilingual (Resolved) | 77.5 | n/a | n/a | 76.7 | 73.3 | 76.2 |
| BrowseComp (Pass@1) | 83.7 | 82.7 | 85.9 | 83.2 | 79.3 | 83.4 |
| HLE w/ tools (Pass@1) | 53.1 | 52.0 | 51.6 | 54.0 | 50.4 | 48.2 |
| MCPAtlas Public (Pass@1) | 73.8 | 67.2 | 69.2 | 66.6 | 71.8 | 73.6 |
| Toolathlon (Pass@1) | 47.2 | 54.6 | 48.8 | 50.0 | 40.7 | 51.8 |
| GDPval-AA (Elo) | 1,619 | 1,674 | 1,314 | 1,482 | 1,535 | 1,554 |
Source: the DS-V4-Pro model card. Best score in each row is marked.
Third-party evaluations
The first third-party evaluations landed right after the open-source preview. They line up with what DeepSeek reported. V4 sits in the top tier of open models, and it is strongest at coding.
API pricing
Price is where V4 pulls ahead. V4-Pro serves the same 1M-token context and up to 384K output tokens as the frontier models. Output costs $0.87 per million tokens. That is roughly 94% below GPT-5.4 at $15, and 97% below Claude Opus 4.8 at $25.
| Model | Input · cache hit | Input · cache miss | Output |
|---|---|---|---|
| DeepSeek-V4-Pro | $0.003625 | $0.435 | $0.87 |
| DeepSeek-V4-Flash | $0.0028 | $0.14 | $0.28 |
Per 1M tokens, as of August 6, 2026. Official pricing.
V4-Flash goes lower still, at $0.28 per million output tokens. That is about 99% cheaper than Opus 4.8, and among the least expensive models at any scale. Prompt caching drops V4-Pro input to $0.003625 per million tokens on a cache hit. Workloads that reuse a system prompt, tool schema, or repository context pay a small fraction of the cache-miss price.
The table above shows the rates that applied through August 15, 2026. They are no longer current. At 16:00 UTC on August 16, 2026, alongside the official DeepSeek-V4-Pro-0813 release, DeepSeek replaced the flat rate with a peak and off-peak structure. V4-Pro output went to $3.96 per million tokens at peak and $1.98 off-peak, against $0.87 before. Peak hours are 01:00 to 04:00 and 06:00 to 10:00 UTC. See the DeepSeek-V4-Pro-0813 page for the full current tables.
Sources: DeepSeek pricing · Claude pricing · OpenAI pricing
What people are saying
“🚀 DeepSeek-V4 Preview is officially live & open-sourced! Welcome to the era of cost-effective 1M context length. DeepSeek-V4-Pro: 1.6T total / 49B active params. Performance rivaling the world's top closed-source models.”
DeepSeek
@deepseek_ai
X Apr 24, 2026
“IMO DeepSeek v4 demonstrated utter confidence and competence by not benchmaxxing... just showed up, demonstrated SOTA long context efficiency techniques (CSA, HCA, mHC, flash at 8% cost of pro...), dropped the best open base models in the world, peaced out. BYO posttraining... bravo.”
swyx
@swyx · Latent Space
X Apr 2026
“A few more notes on DeepSeek-V4: it seems to be a ~GPT-5.2/Opus 4.5+ tier model... at 1.6T params they now have a model that's in the same weight class as GPT-5.4... the technical paper is a big deal... the arch is very efficient for long context.”
Lisan al Gaib
@scaling01
X Apr 24, 2026
“DeepSeek v4 is out, and it's groundbreaking... it's the first open model to have solved long context... beats Opus 4.6 on LiveCodeBench... tops GPT-5.4 on Codeforces... the open landscape is competitive across the board, exciting times!”
Merve Noyan
@mervenoyann · Hugging Face
X Apr 2026
“🚀 DeepSeek-V4 Preview is officially live & open-sourced! Welcome to the era of cost-effective 1M context length. DeepSeek-V4-Pro: 1.6T total / 49B active params. Performance rivaling the world's top closed-source models.”
DeepSeek
@deepseek_ai
X Apr 24, 2026
“IMO DeepSeek v4 demonstrated utter confidence and competence by not benchmaxxing... just showed up, demonstrated SOTA long context efficiency techniques (CSA, HCA, mHC, flash at 8% cost of pro...), dropped the best open base models in the world, peaced out. BYO posttraining... bravo.”
swyx
@swyx · Latent Space
X Apr 2026
“A few more notes on DeepSeek-V4: it seems to be a ~GPT-5.2/Opus 4.5+ tier model... at 1.6T params they now have a model that's in the same weight class as GPT-5.4... the technical paper is a big deal... the arch is very efficient for long context.”
Lisan al Gaib
@scaling01
X Apr 24, 2026
“DeepSeek v4 is out, and it's groundbreaking... it's the first open model to have solved long context... beats Opus 4.6 on LiveCodeBench... tops GPT-5.4 on Codeforces... the open landscape is competitive across the board, exciting times!”
Merve Noyan
@mervenoyann · Hugging Face
X Apr 2026
Run it yourself
vllm serve deepseek-ai/DeepSeek-V4-Pro
sglang serve --model-path deepseek-ai/DeepSeek-V4-ProFrequently asked questions
Is DeepSeek-V4-Pro open source?
Yes. DeepSeek-V4-Pro ships under the MIT license, with open weights available on Hugging Face. You can use it commercially, fine-tune it, and redistribute it without restrictions.
What hardware do I need to run DeepSeek-V4-Pro?
The recommended setups are 8× H200 or 8× B200 GPU nodes. Mixed FP4 + FP8 precision keeps memory requirements manageable despite the 1.6T total parameter count.
What is the difference between DeepSeek-V4-Pro and DeepSeek-V4-Flash?
V4-Pro is the flagship, with 1.6T total parameters and 49B active per token. V4-Flash is a lighter sibling at 284B total parameters and 13B active, aimed at cost-sensitive workloads. Both support a 1M token context window and ship under the MIT license. Since July 31, 2026 the two are also at different release stages: V4-Flash has an official build, DeepSeek-V4-Flash-0731, while V4-Pro is still the April preview.
Is DeepSeek-V4-Pro out of preview yet?
Yes. DeepSeek-V4-Pro-0813 went live on the API on August 12, 2026 and superseded this preview build. The API name did not change, so calls to deepseek-v4-pro now reach the official checkpoint. API pricing moved to a peak and off-peak structure on August 16, 2026.
Which DeepSeek model works with Codex?
DeepSeek-V4-Flash-0731. The official V4-Flash build natively supports the Responses API format and was adapted for Codex, the OpenAI coding assistant available as a CLI, a ChatGPT desktop app, and a VS Code extension. V4-Pro is called through the standard Chat Completions and Anthropic-compatible endpoints.
Is DeepSeek-V4-Flash better than DeepSeek-V4-Pro at agent tasks?
On the numbers DeepSeek published on July 31, 2026, V4-Flash-0731 scores far above V4-Pro-Preview on several agent benchmarks, including Terminal Bench 2.1 at 82.7 and DeepSWE at 54.4. The caveat is timing. Those scores come from a re-post-trained official build measured against a three-month-old preview, so the fair comparison arrives when the official V4-Pro ships.
How does DeepSeek-V4-Pro perform on coding benchmarks?
Launch coverage reports 80.6% on SWE-bench Verified, the strongest open-weights result at release and roughly on par with leading closed models.
Keep exploring
2026-08-13
- Parameters
- 1.6T (49B active)
- Architecture
- MoE
- Context window
- 1M
- Modality
- Text
- License
- MIT
- Recommended Hardware
- 8× H2008× B200
View full details
2026-07
- Parameters
- 284B (13B active)
- Architecture
- MoE
- Context window
- 1M
- Modality
- Text
- License
- MIT
- Recommended Hardware
- 8× H2008× MI325X
View full details
vllm.ai
DeepSeek V4 in vLLM: Efficient Long-context Attention
lmsys.org
DeepSeek-V4 on Day 0: From Fast Inference to Verified RL with SGLang and Miles
newsletter.semianalysis.com
DeepSeekV4 1.6T Day 0 to Day 43 Performance Over Time
From the blog
Is DeepSeek Better Than ChatGPT? 25+ DeepSeek Stats (2026)
From the blog