DeepSeek V4 Pro 0813 on Run BiOS

DeepSeek-V4-Pro is the flagship model of the DeepSeek-V4 series, a family of Mixture-of-Experts (MoE) language models built by DeepSeek (DeepSeek-AI) and described in the technical report "DeepSeek-V4: Towards Highly Efficient Million-Token Context Intelligence" (2026). DeepSeek-V4-Pro has 1.6T total parameters with 49B activated per token, and supports a context length of one million tokens. DeepSeek released the V4 series as a preview on April 24, 2026, describing DeepSeek-V4-Pro as delivering "performance rivaling the world's top closed-source models."

DeepSeek reports that DeepSeek-V4-Pro-Max — the maximum reasoning effort mode of DeepSeek-V4-Pro — "significantly advances the knowledge capabilities of open-source models, firmly establishing itself as the best open-source model available today," achieving top-tier performance in coding benchmarks and significantly bridging the gap with leading closed-source models on reasoning and agentic tasks. DeepSeek's announcement further characterizes DeepSeek-V4-Pro as offering:

• Enhanced agentic capabilities — open-source state of the art in agentic coding benchmarks.
• Rich world knowledge — leads all current open models, trailing only Gemini-3.1-Pro.
• World-class reasoning — beats all current open models in Math/STEM/Coding, rivaling top closed-source models.

Architecture

DeepSeek-V4-Pro is a Mixture-of-Experts Transformer. Per the technical report, the DeepSeek-V4 series incorporates several key upgrades in architecture and optimization:

1. Hybrid attention architecture — a hybrid attention mechanism combining Compressed Sparse Attention (CSA) and Heavily Compressed Attention (HCA) to dramatically improve long-context efficiency. In the 1M-token context setting, DeepSeek-V4-Pro requires only 27% of single-token inference FLOPs and 10% of KV cache compared with DeepSeek-V3.2.
2. Manifold-Constrained Hyper-Connections (mHC) — incorporated to strengthen conventional residual connections, enhancing stability of signal propagation across layers while preserving model expressivity.
3. Muon optimizer — employed for faster convergence and greater training stability.

Published configuration details for DeepSeek-V4-Pro from the technical report:

• 61 Transformer layers; hidden dimension 7168.
• The first two layers use HCA; subsequent layers use CSA and HCA in an interleaved manner.
• MoE layers are used in all Transformer blocks, with the Hash routing strategy for the first 3 MoE layers. Each MoE layer consists of 1 shared expert and 384 routed experts with an intermediate hidden dimension of 3072 per expert; 6 routed experts are activated for each token.
• Multi-token prediction (MTP) depth is set to 1.
• mHC expansion factor is set to 4.
• The FFNs follow the DeepSeekMoE paradigm with fine-grained routed experts and shared experts, an auxiliary-loss-free load-balancing strategy augmented by a slight sequence-wise balance loss, and affinity scores computed with a Sqrt(Softplus(.)) activation.

The released DeepSeek-V4-Pro checkpoint uses FP4 + FP8 mixed precision: MoE expert parameters use FP4 precision, while most other parameters use FP8. (The companion base checkpoint, DeepSeek-V4-Pro-Base, is published in FP8 mixed precision.)

Training

Per the technical report, both DeepSeek-V4 models were pre-trained on more than 32T diverse and high-quality tokens, followed by a comprehensive post-training pipeline. Post-training features a two-stage paradigm: independent cultivation of domain-specific experts (through supervised fine-tuning and reinforcement learning with GRPO), followed by unified model consolidation via on-policy distillation, integrating distinct proficiencies across diverse domains into a single model. The Muon optimizer is used for the majority of parameters, with AdamW for the embedding module, the prediction head module, and the weights of all RMSNorm modules.

Model Variants

The DeepSeek-V4 series as published by DeepSeek:

Model	#Total Params	#Activated Params	Context Length	Precision
DeepSeek-V4-Flash-Base	284B	13B	1M	FP8 Mixed
DeepSeek-V4-Flash	284B	13B	1M	FP4 + FP8 Mixed*
DeepSeek-V4-Pro-Base	1.6T	49B	1M	FP8 Mixed
DeepSeek-V4-Pro	1.6T	49B	1M	FP4 + FP8 Mixed*

*FP4 + FP8 Mixed: MoE expert parameters use FP4 precision; most other parameters use FP8.

Reasoning Modes

DeepSeek reports that DeepSeek-V4-Pro and DeepSeek-V4-Flash both support three reasoning effort modes:

Reasoning Mode	Characteristics	Typical Use Cases	Response Format
Non-think	Fast, intuitive responses	Routine daily tasks, low-risk decisions	</think> summary
Think High	Conscious logical analysis, slower but more accurate	Complex problem-solving, planning	<think> thinking </think> summary
Think Max	Push reasoning to its fullest extent	Exploring the boundary of model reasoning capability	Special system prompt + <think> thinking </think> summary

For the Think Max reasoning mode, DeepSeek recommends a context window of at least 384K tokens, and recommends sampling parameters of temperature = 1.0 and top_p = 1.0.

Benchmarks

All tables below are reproduced in full from DeepSeek's official model card for the DeepSeek-V4 series. Bold marks the best score in a row, as published.

Base Model

Benchmark (Metric)	# Shots	DeepSeek-V3.2-Base	DeepSeek-V4-Flash-Base	DeepSeek-V4-Pro-Base
Architecture	-	MoE	MoE	MoE
# Activated Params	-	37B	13B	49B
# Total Params	-	671B	284B	1.6T
World Knowledge
AGIEval (EM)	0-shot	80.1	82.6	83.1
MMLU (EM)	5-shot	87.8	88.7	90.1
MMLU-Redux (EM)	5-shot	87.5	89.4	90.8
MMLU-Pro (EM)	5-shot	65.5	68.3	73.5
MMMLU (EM)	5-shot	87.9	88.8	90.3
C-Eval (EM)	5-shot	90.4	92.1	93.1
CMMLU (EM)	5-shot	88.9	90.4	90.8
MultiLoKo (EM)	5-shot	38.7	42.2	51.1
Simple-QA verified (EM)	25-shot	28.3	30.1	55.2
SuperGPQA (EM)	5-shot	45.0	46.5	53.9
FACTS Parametric (EM)	25-shot	27.1	33.9	62.6
TriviaQA (EM)	5-shot	83.3	82.8	85.6
Language & Reasoning
BBH (EM)	3-shot	87.6	86.9	87.5
DROP (F1)	1-shot	88.2	88.6	88.7
HellaSwag (EM)	0-shot	86.4	85.7	88.0
WinoGrande (EM)	0-shot	78.9	79.5	81.5
CLUEWSC (EM)	5-shot	83.5	82.2	85.2
Code & Math
BigCodeBench (Pass@1)	3-shot	63.9	56.8	59.2
HumanEval (Pass@1)	0-shot	62.8	69.5	76.8
GSM8K (EM)	8-shot	91.1	90.8	92.6
MATH (EM)	4-shot	60.5	57.4	64.5
MGSM (EM)	8-shot	81.3	85.7	84.4
CMath (EM)	3-shot	92.6	93.6	90.9
Long Context
LongBench-V2 (EM)	1-shot	40.2	44.7	51.5

DeepSeek-V4-Pro-Max vs Frontier Models

Benchmark (Metric)	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	-	91.7	89.6	-	93.5
Codeforces (Rating)	-	3168	3052	-	-	3206
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
Long Context
MRCR 1M (MMR)	92.9	-	76.3	-	-	83.5
CorpusQA 1M (ACC)	71.7	-	53.8	-	-	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	-	80.6	80.2	-	80.6
SWE Pro (Resolved)	57.3	57.7	54.2	58.6	58.4	55.4
SWE Multilingual (Resolved)	77.5	-	-	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
GDPval-AA (Elo)	1619	1674	1314	1482	1535	1554
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

Comparison across Modes

Benchmark (Metric)	V4-Flash Non-Think	V4-Flash High	V4-Flash Max	V4-Pro Non-Think	V4-Pro High	V4-Pro Max
Knowledge & Reasoning
MMLU-Pro (EM)	83.0	86.4	86.2	82.9	87.1	87.5
SimpleQA-Verified (Pass@1)	23.1	28.9	34.1	45.0	46.2	57.9
Chinese-SimpleQA (Pass@1)	71.5	73.2	78.9	75.8	77.7	84.4
GPQA Diamond (Pass@1)	71.2	87.4	88.1	72.9	89.1	90.1
HLE (Pass@1)	8.1	29.4	34.8	7.7	34.5	37.7
LiveCodeBench (Pass@1)	55.2	88.4	91.6	56.8	89.8	93.5
Codeforces (Rating)	-	2816	3052	-	2919	3206
HMMT 2026 Feb (Pass@1)	40.8	91.9	94.8	31.7	94.0	95.2
IMOAnswerBench (Pass@1)	41.9	85.1	88.4	35.3	88.0	89.8
Apex (Pass@1)	1.0	19.1	33.0	0.4	27.4	38.3
Apex Shortlist (Pass@1)	9.3	72.1	85.7	9.2	85.5	90.2
Long Context
MRCR 1M (MMR)	37.5	76.9	78.7	44.7	83.3	83.5
CorpusQA 1M (ACC)	15.5	59.3	60.5	35.6	56.5	62.0
Agentic
Terminal Bench 2.0 (Acc)	49.1	56.6	56.9	59.1	63.3	67.9
SWE Verified (Resolved)	73.7	78.6	79.0	73.6	79.4	80.6
SWE Pro (Resolved)	49.1	52.3	52.6	52.1	54.4	55.4
SWE Multilingual (Resolved)	69.7	70.2	73.3	69.8	74.1	76.2
BrowseComp (Pass@1)	-	53.5	73.2	-	80.4	83.4
HLE w/ tools (Pass@1)	-	40.3	45.1	-	44.7	48.2
MCPAtlas (Pass@1)	64.0	67.4	69.0	69.4	74.2	73.6
GDPval-AA (Elo)	-	-	1395	-	-	1554
Toolathlon (Pass@1)	40.7	43.5	47.8	46.3	49.0	51.8

Agentic and Coding Capabilities

DeepSeek positions DeepSeek-V4-Pro as its strongest agentic and coding model: the announcement describes open-source state-of-the-art results in agentic coding benchmarks, and the benchmark tables above include Terminal Bench 2.0, SWE Verified, SWE Pro, SWE Multilingual, BrowseComp, HLE with tools, MCPAtlas, Toolathlon, LiveCodeBench, and Codeforces results. DeepSeek's announcement states that DeepSeek-V4 is integrated with leading AI coding agents and that it already drives in-house agentic coding at DeepSeek.

Safety Notes

DeepSeek's official model card and technical report for the DeepSeek-V4 series do not publish a dedicated safety evaluation, red-teaming, or responsible-use section. The model is released under the MIT License. As with any general-purpose model, outputs should be reviewed before use in high-stakes applications.

On Run BiOS

On the Run BiOS platform, deepseek-v4-pro-0813 is the frozen pin of the DeepSeek-V4-Pro-0813 checkpoint (the re-post-trained release published by DeepSeek on August 13, 2026): it always serves that checkpoint, while the stable deepseek-v4-pro id follows new checkpoints as DeepSeek publishes them (today the two ids serve the same 0813 weights and are billed identically). It is available with a 1,048,576-token context window, function calling, reasoning with effort levels high and max, and prompt caching for repeated prefixes.