Where is the DeepSeek V4 Pro 0423 documentation?

The DeepSeek V4 Pro 0423 documentation on OpenRouter includes the Does DeepSeek V4 Have Vision? blog post, the Why Use OpenRouter for DeepSeek blog post, and the DeepSeek V4 Is Earning Agentic Token Share blog post. The quickstart covers API keys and your first request, and the API reference lists every parameter.

How much does DeepSeek V4 Pro 0423 cost?

DeepSeek V4 Pro 0423 costs $0.2079/M input tokens and $0.4157/M output tokens, with separate rates for Cache Read at $0.01722/M tokens.

What is the context length of DeepSeek V4 Pro 0423?

DeepSeek V4 Pro 0423 has a 1,048,576 token context window. It supports up to 393,216 completion tokens.

Does DeepSeek V4 Pro 0423 support tool calling and structured outputs?

Yes. DeepSeek V4 Pro 0423 accepts tools and tool_choice for function calling. It also supports structured outputs via a JSON schema in response_format.

Which providers serve DeepSeek V4 Pro 0423?

DeepSeek V4 Pro 0423 is served by 12 providers on OpenRouter: Baidu Qianfan, StreamLake, Relace, GMICloud, SiliconFlow, DigitalOcean, DeepInfra, Alibaba Cloud Int. and 4 more. Requests are routed to the best available provider, with automatic failover to the others, and you can pin or exclude providers with provider routing.

When was DeepSeek V4 Pro 0423 released?

DeepSeek V4 Pro 0423 was released on April 24, 2026.

DeepSeek: DeepSeek V4 Pro 0423

deepseek/deepseek-v4-pro

Model weights

Compare

DeepSeek V4 Pro is a large-scale Mixture-of-Experts model from DeepSeek with 1.6T total parameters and 49B activated parameters, supporting a 1M-token context window. It is designed for advanced reasoning, coding, and long-horizon agent workflows, with strong performance across knowledge, math, and software engineering benchmarks.

Built on the same architecture as DeepSeek V4 Flash, it introduces a hybrid attention system for efficient long-context processing. Reasoning efforts high and xhigh are supported; xhigh maps to max reasoning. It is well suited for complex workloads such as full-codebase analysis, multi-step automation, and large-scale information synthesis, where both capability and efficiency are critical

Modalities

In / out price

88% off

$0.2079 / $0.4157per 1M

Context

1.0M

Released

Apr 24, 2026

Providers

Different companies host the same model. OpenRouter routes your request to one of them based on the routing mode you pick: Balanced (price + speed), Nitro (fastest), or Floor (cheapest).

Latency / throughput


	$0.90	$1.50	$0.25	0.77s	98 tok/s	100.00%
88% off	$1.69$0.2079	$3.38$0.4157	$0.14$0.01722	1.24s	54 tok/s	99.76%
88% off	$1.74$0.2088	$3.48$0.4176	$0.145$0.0174	1.45s	40 tok/s	100.00%
	$1.60	$3.20	$0.135	2.28s	53 tok/s	100.00%
	$1.044	$2.088	$0.2088	0.92s	49 tok/s	100.00%
	$1.416	$2.832	$0.118	2.13s	39 tok/s	98.12%
	$1.00	$1.50	$0.135	1.41s	23 tok/s	99.89%
45% off	$1.74$0.957	$3.48$1.914	$0.145$0.07975	2.92s	26 tok/s	99.89%
	$1.56	$3.12	$0.13	1.89s	40 tok/s	99.03%
	$1.30	$2.60	$0.10	1.16s	63 tok/s	99.92%
	$1.91	$3.83	$0.16	1.66s	99 tok/s	99.94%
	$1.65	$3.301	$0.33	1.63s	51 tok/s	99.81%
	$1.68	$3.38	$0.13	1.91s	37 tok/s	99.25%

Input $0.90Output $1.50
98 tok/s100.00% uptime
Input $0.2079Output $0.4157
88% off
54 tok/s99.76% uptime
Input $0.2088Output $0.4176
88% off
40 tok/s100.00% uptime
Input $1.60Output $3.20
53 tok/s100.00% uptime
Input $1.044Output $2.088
49 tok/s100.00% uptime
Input $1.416Output $2.832
39 tok/s98.12% uptime
Input $1.00Output $1.50
23 tok/s99.89% uptime
Input $0.957Output $1.914
45% off
26 tok/s99.89% uptime
Input $1.56Output $3.12
40 tok/s99.03% uptime
Input $1.30Output $2.60
63 tok/s99.92% uptime
Input $1.91Output $3.83
99 tok/s99.94% uptime
Input $1.65Output $3.301
51 tok/s99.81% uptime
Input $1.68Output $3.38
37 tok/s99.25% uptime

Pricing

The average price customers actually pay for this model, next to the prices providers post. Caching and discounts mean the price actually paid is often well below the listed one.

Weighted average

Input

$0.2046/M tokens

Output

$1.245/M tokens

DeepSeek V4 Pro 0423 — Price history

Price history

No effective price data for this selection

$0.0692	$0.666	86.5%	27.4%
$0.0905	$0.672	79.5%	20.4%
$0.2340	$3.200	93.2%	16.5%
$0.2754	$2.831	87.9%	1.7%
$0.3086	$1.500	79.9%	1.1%
$0.3427	$1.500	85.7%	25.4%
$0.3774	$3.380	84.0%	0.1%
$0.3853	$3.829	87.1%	2.6%
$0.4489	$2.599	70.9%	2.2%
$0.4537	$1.912	57.4%	1.5%
$0.4798	$2.087	67.6%	1.0%
$0.5407	$3.119	71.3%	0.1%
$1.0579	$3.300	44.8%	0.1%

Effective in $0.0692Effective out $0.666
Token share
27.4%
Effective in $0.0905Effective out $0.672
Token share
20.4%
Effective in $0.2340Effective out $3.200
Token share
16.5%
Effective in $0.2754Effective out $2.831
Token share
1.7%
Effective in $0.3086Effective out $1.500
Token share
1.1%
Effective in $0.3427Effective out $1.500
Token share
25.4%
Effective in $0.3774Effective out $3.380
Token share
0.1%
Effective in $0.3853Effective out $3.829
Token share
2.6%
Effective in $0.4489Effective out $2.599
Token share
2.2%
Effective in $0.4537Effective out $1.912
Token share
1.5%
Effective in $0.4798Effective out $2.087
Token share
1.0%
Effective in $0.5407Effective out $3.119
Token share
0.1%
Effective in $1.0579Effective out $3.300
Token share
0.1%

Performance

Throughput is how fast the model writes (tokens per second — higher is better). Latency is total round-trip time (lower is better). TTFT is time-to-first-token — how long before you see anything appear (lower is better).

Throughput

99tok/s

P50, best across providers

Latency

0.77s

P50, best provider

Reasoning effortLatency / throughput

AutoExacto benchmarks

GPQA DiamondTAU-BenchParasail88.0%--Cloudflare91.8%76.4%GMICloud91.3%75.9%NovitaAI91.3%75.5%auto-routing88.4%76.9%

+13 more providers

Uptime

Uptime is the percentage of the past 3 days that at least one provider was responding to requests. Availability is the percentage of time that inference was successfully served. OpenRouter continuously monitors and uses the next-best provider when one returns an error.

Uptime (3d)

100.00%

Availability (3d)

99.46%

Availability over the last 3 days

Last 72 hours

Availability 99.46%

3 Days Ago2 Days AgoYesterdayNow

Availability over the last 24 hours

OpenRouter Availability

99.58%

Without Routing

99.04%

When an error occurs in an upstream provider, we can recover by routing to another healthy provider, if your request filters allow it. You can access per-provider uptime data programmatically through the Endpoints API. Learn more about our load balancing and provider routing options.

Benchmarks

Scores on standardized evaluations and this model’s own outputs on shared benchmark prompts. Higher percentages are better — and each rank shows where this model lands among the models evaluated on that source.

Benchmark score summary for DeepSeek: DeepSeek V4 Pro 0423 (Artificial Analysis, Design Arena and Vals AI)
Source	Benchmark	Score
Artificial Analysis	DeepSeek V4 Pro 0813 (High) Intelligence Index	20.8
Artificial Analysis	DeepSeek V4 Pro 0813 (High) HLE	41.5%
Artificial Analysis	DeepSeek V4 Pro 0813 (High) AA-LCR	79.3%
Artificial Analysis	DeepSeek V4 Pro 0813 (High) GDPval-AA	41.8%
Artificial Analysis	DeepSeek V4 Pro 0813 (High) CritPt	18.6%
Artificial Analysis	DeepSeek V4 Pro 0813 (High) SciCode	51.7%
Artificial Analysis	DeepSeek V4 Pro 0813 (High) AA-Omniscience Accuracy	45.1%
Artificial Analysis	DeepSeek V4 Pro 0813 (High) AA-Omniscience Non-Hallucination Rate	5.9%
Artificial Analysis	DeepSeek V4 Pro 0813 (Non-reasoning) Intelligence Index	12.6
Artificial Analysis	DeepSeek V4 Pro 0813 (Non-reasoning) HLE	10.6%
Artificial Analysis	DeepSeek V4 Pro 0813 (Non-reasoning) AA-LCR	50.7%
Artificial Analysis	DeepSeek V4 Pro 0813 (Non-reasoning) GDPval-AA	28.3%
Artificial Analysis	DeepSeek V4 Pro 0813 (Non-reasoning) CritPt	0.0%
Artificial Analysis	DeepSeek V4 Pro 0813 (Non-reasoning) SciCode	39.9%
Artificial Analysis	DeepSeek V4 Pro 0813 (Non-reasoning) Terminal-Bench 4.0	1.5%
Artificial Analysis	DeepSeek V4 Pro 0813 (Non-reasoning) AA-Omniscience Accuracy	38.6%
Artificial Analysis	DeepSeek V4 Pro 0813 (Non-reasoning) AA-Omniscience Non-Hallucination Rate	6.1%
Artificial Analysis	DeepSeek V4 Pro 0424 (High) Coding Index	58.7
Artificial Analysis	DeepSeek V4 Pro 0424 (High) GPQA Diamond	90.5%
Artificial Analysis	DeepSeek V4 Pro 0424 (High) HLE	35.2%
Artificial Analysis	DeepSeek V4 Pro 0424 (High) IFBench	71.3%
Artificial Analysis	DeepSeek V4 Pro 0424 (High) τ²-Bench Telecom	94.2%
Artificial Analysis	DeepSeek V4 Pro 0424 (High) AA-LCR	70.3%
Artificial Analysis	DeepSeek V4 Pro 0424 (High) τ-Bench Banking	26.2%
Artificial Analysis	DeepSeek V4 Pro 0424 (High) GDPval-AA	33.0%
Artificial Analysis	DeepSeek V4 Pro 0424 (High) CritPt	10.0%
Artificial Analysis	DeepSeek V4 Pro 0424 (High) Terminal-Bench Hard	41.7%
Artificial Analysis	DeepSeek V4 Pro 0424 (High) Terminal-Bench 2.1	64.8%
Artificial Analysis	DeepSeek V4 Pro 0424 (Max) Coding Index	59.4
Artificial Analysis	DeepSeek V4 Pro 0424 (Max) Agentic Index	26.3
Artificial Analysis	DeepSeek V4 Pro 0424 (Max) GPQA Diamond	88.8%
Artificial Analysis	DeepSeek V4 Pro 0424 (Max) HLE	37.5%
Artificial Analysis	DeepSeek V4 Pro 0424 (Max) IFBench	76.5%
Artificial Analysis	DeepSeek V4 Pro 0424 (Max) τ²-Bench Telecom	96.2%
Artificial Analysis	DeepSeek V4 Pro 0424 (Max) AA-LCR	74.7%
Artificial Analysis	DeepSeek V4 Pro 0424 (Max) τ-Bench Banking	30.1%
Artificial Analysis	DeepSeek V4 Pro 0424 (Max) GDPval-AA	33.3%
Artificial Analysis	DeepSeek V4 Pro 0424 (Max) CritPt	12.9%
Artificial Analysis	DeepSeek V4 Pro 0424 (Max) SciCode	50.8%
Artificial Analysis	DeepSeek V4 Pro 0424 (Max) Terminal-Bench Hard	46.2%
Artificial Analysis	DeepSeek V4 Pro 0424 (Max) Terminal-Bench 4.0	14.6%
Artificial Analysis	DeepSeek V4 Pro 0424 (Max) Terminal-Bench 2.1	64.0%
Artificial Analysis	DeepSeek V4 Pro 0424 (Non-reasoning) GPQA Diamond	71.7%
Artificial Analysis	DeepSeek V4 Pro 0424 (Non-reasoning) HLE	8.2%
Artificial Analysis	DeepSeek V4 Pro 0424 (Non-reasoning) IFBench	45.8%
Artificial Analysis	DeepSeek V4 Pro 0424 (Non-reasoning) τ²-Bench Telecom	91.2%
Artificial Analysis	DeepSeek V4 Pro 0424 (Non-reasoning) AA-LCR	53.0%
Artificial Analysis	DeepSeek V4 Pro 0424 (Non-reasoning) CritPt	0.9%
Artificial Analysis	DeepSeek V4 Pro 0424 (Non-reasoning) Terminal-Bench Hard	36.4%
Design Arena	DeepSeek-V4-Pro Agents Arena Full Stack Elo	948
Design Arena	DeepSeek-V4-Pro Agents Arena Godotgamedev Elo	1059
Design Arena	DeepSeek-V4-Pro Agents Arena Webapps Elo	1000
Design Arena	DeepSeek-V4-Pro Models Arena 3D Elo	1233
Design Arena	DeepSeek-V4-Pro Models Arena Asciiart Elo	1143
Design Arena	DeepSeek-V4-Pro Models Arena Code Categories Elo	1231
Design Arena	DeepSeek-V4-Pro Models Arena Data Visualization Elo	1200
Design Arena	DeepSeek-V4-Pro Models Arena Game Development Elo	1221
Design Arena	DeepSeek-V4-Pro Models Arena SVG Elo	1128
Design Arena	DeepSeek-V4-Pro Models Arena UI Component Elo	1211
Design Arena	DeepSeek-V4-Pro Models Arena Website Elo	1227
Vals AI	DeepSeek V4 CaseLaw v2	59.4%
Vals AI	DeepSeek V4 Code Migration	26.2%
Vals AI	DeepSeek V4 CorpFin v2	61.4%
Vals AI	DeepSeek V4 EMB	51.6%
Vals AI	DeepSeek V4 Finance Agent (v2)	44.1%
Vals AI	DeepSeek V4 GPQA Diamond	89.4%
Vals AI	DeepSeek V4 Harvey's Legal Agent Benchmark v1	3.8%
Vals AI	DeepSeek V4 IOI v1	35.8%
Vals AI	DeepSeek V4 LiveCodeBench	87.5%
Vals AI	DeepSeek V4 LegalBench	80.3%
Vals AI	DeepSeek V4 Legal Research Bench	23.1%
Vals AI	DeepSeek V4 MedCode	40.5%
Vals AI	DeepSeek V4 MedScribe	75.1%
Vals AI	DeepSeek V4 MMLU Pro	87.2%
Vals AI	DeepSeek V4 ProgramBench	0.0%
Vals AI	DeepSeek V4 ProofBench v1.1	16.0%
Vals AI	DeepSeek V4 Public Benefits Bench v1.1	62.9%
Vals AI	DeepSeek V4 SkillsBench	51.3%
Vals AI	DeepSeek V4 SWE-bench	77.4%
Vals AI	DeepSeek V4 Tax Agent Bench	25.2%
Vals AI	DeepSeek V4 TaxEval v2	72.1%
Vals AI	DeepSeek V4 Terminal-Bench 4.0	11.1%
Vals AI	DeepSeek V4 Vals Index	38.6%
Vals AI	DeepSeek V4 Vibe Code Bench v1.1	49.9%

Apps

Public apps that send the most traffic to this model. Good signal for what real production workloads look like — and a hint at which use cases this model is best suited for.

Hermes Agent

Hermes Agent is an open-source, self-improving AI agent by Nous Research that runs persistently with memory across sessions, and builds reusable skills from experience. It comes with 40+ built-in tools, including web search, browser automation, and vision, plus scheduled automations and subagents.

652Btokens

OpenClaw

OpenClaw is an open-source AI agent that connects to your messaging apps and takes real actions on your behalf, from running commands and browsing the web to managing files and sending emails.

129Btokens

DeepSeek Harness

new

56.1Btokens

There are many coding agents, but this one is yours.

45.4Btokens

Janitor AI

Janitor AI is a chatbot platform where users create and chat with custom AI characters for interactive roleplay, storytelling, and immersive fiction.

44.1Btokens

Rest of OpenRouter

27.4B

Top 5 Apps

14.2B

Not currently available through EU in-region routing. In-region routing docs

More models from DeepSeek

DeepSeek V4.1 Flash 1017

DeepSeek V4.1 Flash 1017 is the October 17, 2026 snapshot of DeepSeek V4.1 Flash, served by DeepSeek under the deepseek-flash API model name. V4.1 Flash is a sparse mixture-of-experts model from DeepSeek, and the first built on the company's Causal Encoder-Decoder (CED) architecture. It activates 8B parameters on input and 16B on output from a 552B-parameter backbone, an asymmetric split that keeps per-token compute low relative to the model's total size. Image understanding is native to the architecture, with visual and text embeddings trained jointly from the start of pre-training rather than added afterward as in the earlier experimental V4 Flash Vision Exp.

It is suited for coding, terminal, and computer-use agents, along with long-horizon tasks that must run to completion across many steps and long-context analysis. Compressed KV caching cuts cache memory to roughly a quarter of the previous Flash generation, significantly reducing costs on agentic workloads. DeepSeek positions it as the cost-efficient tier of the V4.1 family and reports that it exceeds V4 Pro on performance, speed, and task completion time.

Text1.0M context$0.038 / $0.50

DeepSeek V4.1 Flash 1017 (batch)

Text1.0M context$0.05 / $0.15

DeepSeek Pro Latest

This model always redirects to the latest model in the DeepSeek Pro family.

Text1.0M context

DeepSeek Flash Latest

This model always redirects to the latest model in the DeepSeek Flash family.

Text1.0M context

DeepSeek V4.1 Flash

DeepSeek V4.1 Flash is a sparse mixture-of-experts model from DeepSeek, and the first built on the company's Causal Encoder-Decoder (CED) architecture. It activates 8B parameters on input and 16B on output from a 552B-parameter backbone, an asymmetric split that keeps per-token compute low relative to the model's total size. Image understanding is native to the architecture, with visual and text embeddings trained jointly from the start of pre-training rather than added afterward as in the earlier experimental V4 Flash Vision Exp.

Text1.0M context$0.0507 / $0.2028

DeepSeek V4.1 Flash (batch)

Text1.0M context$0.112 / $0.336

DeepSeek V4 Flash Vision Exp

DeepSeek V4 Flash Vision Exp is an experimental vision-enabled version of DeepSeek V4 Flash 0731 from DeepSeek, adding image understanding while matching the base model on text capabilities including agents, reasoning, and world knowledge. It is a sparse mixture-of-experts model with 13B active parameters out of 284B total.

It is suited for document and chart understanding, visual question answering, and multimodal agent workflows that interleave text and images.

Text1.0M context$0.2156 / $0.6468

DeepSeek V4 Pro 0813

DeepSeek V4 Pro 0813 is a large-scale mixture-of-experts model from DeepSeek. This is the GA release of DeepSeek V4 Pro.

Text1.0M context$0.0999 / $8

DeepSeek V4 Flash Latest

This model always redirects to the latest model in the DeepSeek V4 Flash family.

Text1.0M context

DeepSeek V4 Flash 0731

DeepSeek V4 Flash 0731 is a sparse mixture-of-experts model from DeepSeek, with 13B active parameters out of 284B total. This re-post-trained revision is suited for coding, reasoning, and agent workflows. This is the GA release of DeepSeek V4 Flash.

Text1.0M context$0.0434 / $0.30

DeepSeek V4 Flash 0731 (batch)

Text1.0M context$0.048 / $0.144

DeepSeek V4 Flash 0423

DeepSeek V4 Flash is an efficiency-optimized Mixture-of-Experts model from DeepSeek with 284B total parameters and 13B activated parameters, supporting a 1M-token context window. It is designed for fast inference and high-throughput workloads, while maintaining strong reasoning and coding performance.

The model includes hybrid attention for efficient long-context processing. Reasoning efforts high and xhigh are supported; xhigh maps to max reasoning. It is well suited for applications such as coding assistants, chat systems, and agent workflows where responsiveness and cost efficiency are important.

Text1.0M context$0.014 / $0.45

DeepSeek V3.2 Speciale

DeepSeek-V3.2-Speciale is a high-compute variant of DeepSeek-V3.2 optimized for maximum reasoning and agentic performance. It builds on DeepSeek Sparse Attention (DSA) for efficient long-context processing, then scales post-training reinforcement learning to push capability beyond the base model. Reported evaluations place Speciale ahead of GPT-5 on difficult reasoning workloads, with proficiency comparable to Gemini-3.0-Pro, while retaining strong coding and tool-use reliability. Like V3.2, it benefits from a large-scale agentic task synthesis pipeline that improves compliance and generalization in interactive environments.

Text131K context

DeepSeek V3.2

DeepSeek-V3.2 is a large language model designed to harmonize high computational efficiency with strong reasoning and agentic tool-use performance. It introduces DeepSeek Sparse Attention (DSA), a fine-grained sparse attention mechanism that reduces training and inference cost while preserving quality in long-context scenarios. A scalable reinforcement learning post-training framework further improves reasoning, with reported performance in the GPT-5 class, and the model has demonstrated gold-medal results on the 2025 IMO and IOI. V3.2 also uses a large-scale agentic task synthesis pipeline to better integrate reasoning into tool-use settings, boosting compliance and generalization in interactive environments.

Users can control the reasoning behaviour with the reasoning enabled boolean. Learn more in our docs

Text164K context$0.2088 / $0.3096

DeepSeek V3.2 Exp

DeepSeek-V3.2-Exp is an experimental large language model released by DeepSeek as an intermediate step between V3.1 and future architectures. It introduces DeepSeek Sparse Attention (DSA), a fine-grained sparse attention mechanism designed to improve training and inference efficiency in long-context scenarios while maintaining output quality. Users can control the reasoning behaviour with the reasoning enabled boolean. Learn more in our docs

The model was trained under conditions aligned with V3.1-Terminus to enable direct comparison. Benchmarking shows performance roughly on par with V3.1 across reasoning, coding, and agentic tool-use tasks, with minor tradeoffs and gains depending on the domain. This release focuses on validating architectural optimizations for extended context lengths rather than advancing raw task accuracy, making it primarily a research-oriented model for exploring efficient transformer designs.

Text164K context$0.27 / $0.41

DeepSeek V3.1 Terminus

DeepSeek-V3.1 Terminus is an update to DeepSeek V3.1 that maintains the model's original capabilities while addressing issues reported by users, including language consistency and agent capabilities, further optimizing the model's performance in coding and search agents. It is a large hybrid reasoning model (671B parameters, 37B active) that supports both thinking and non-thinking modes. It extends the DeepSeek-V3 base with a two-phase long-context training process, reaching up to 128K tokens, and uses FP8 microscaling for efficient inference. Users can control the reasoning behaviour with the reasoning enabled boolean. Learn more in our docs

The model improves tool use, code generation, and reasoning efficiency, achieving performance comparable to DeepSeek-R1 on difficult benchmarks while responding more quickly. It supports structured tool calling, code agents, and search agents, making it suitable for research, coding, and agentic workflows.

Text164K context$0.27 / $1

DeepSeek V3.1

DeepSeek-V3.1 is a large hybrid reasoning model (671B parameters, 37B active) that supports both thinking and non-thinking modes via prompt templates. It extends the DeepSeek-V3 base with a two-phase long-context training process, reaching up to 128K tokens, and uses FP8 microscaling for efficient inference. Users can control the reasoning behaviour with the reasoning enabled boolean. Learn more in our docs

It succeeds the DeepSeek V3-0324 model and performs well on a variety of tasks.

Text164K context$0.25 / $0.95

DeepSeek V3.1 Base

This is a base model, trained only for raw next-token prediction. Unlike instruct/chat models, it has not been fine-tuned to follow user instructions. Prompts need to be written more like training text or examples rather than simple requests (e.g., “Translate the following sentence…” instead of just “Translate this”).

DeepSeek-V3.1 Base is a 671B parameter open Mixture-of-Experts (MoE) language model with 37B active parameters per forward pass and a context length of 128K tokens. Trained on 14.8T tokens using FP8 mixed precision, it achieves high training efficiency and stability, with strong performance across language, reasoning, math, and coding tasks.

Text164K context

R1 Distill Qwen 7B

DeepSeek-R1-Distill-Qwen-7B is a 7 billion parameter dense language model distilled from DeepSeek-R1, leveraging reinforcement learning-enhanced reasoning data generated by DeepSeek's larger models. The distillation process transfers advanced reasoning, math, and code capabilities into a smaller, more efficient model architecture based on Qwen2.5-Math-7B. This model demonstrates strong performance across mathematical benchmarks (92.8% pass@1 on MATH-500), coding tasks (Codeforces rating 1189), and general reasoning (49.1% pass@1 on GPQA Diamond), achieving competitive accuracy relative to larger models while maintaining smaller inference costs.

Text131K context

DeepSeek R1 0528 Qwen3 8B

DeepSeek-R1-0528 is a lightly upgraded release of DeepSeek R1 that taps more compute and smarter post-training tricks, pushing its reasoning and inference to the brink of flagship models like O3 and Gemini 2.5 Pro. It now tops math, programming, and logic leaderboards, showcasing a step-change in depth-of-thought. The distilled variant, DeepSeek-R1-0528-Qwen3-8B, transfers this chain-of-thought into an 8 B-parameter form, beating standard Qwen3 8B by +10 pp and tying the 235 B “thinking” giant on AIME 2024.

Text131K context

DeepSeek: DeepSeek V4 Pro 0423

Providers