Where is the DeepSeek V4 Pro 0813 documentation?

The DeepSeek V4 Pro 0813 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 0813 cost?

DeepSeek V4 Pro 0813 costs $0.0999/M input tokens and $8.00/M output tokens, with separate rates for Cache Read at $0.0999/M tokens.

What is the context length of DeepSeek V4 Pro 0813?

DeepSeek V4 Pro 0813 has a 1,048,576 token context window. It supports up to 393,216 completion tokens.

Does DeepSeek V4 Pro 0813 support tool calling and structured outputs?

Yes. DeepSeek V4 Pro 0813 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 0813?

DeepSeek V4 Pro 0813 is served by 20 providers on OpenRouter: Relace, Wafer, Alibaba Cloud Int., Baidu Qianfan, StreamLake, DeepSeek, NovitaAI, GMICloud and 12 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 0813 released?

DeepSeek V4 Pro 0813 was released on August 12, 2026.

DeepSeek: DeepSeek V4 Pro 0813

deepseek/deepseek-v4-pro-0813

Model weights

Compare

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

Modalities

In / out price

$0.0999 / $8per 1M

Context

1.0M

Released

Aug 12, 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


	$1.32	$3.96	$0.044	1.24s	112 tok/s	100.00%
	$1.32	$3.96	$0.13	0.43s	115 tok/s	99.57%
	$0.111	$4.44	$0.097	0.25s	93 tok/s	99.65%
	$1.31	$3.96	$0.044	0.55s	92 tok/s	99.42%
	$0.285	$4.30	$0.284	0.26s	90 tok/s	99.57%
	$1.35	$2.40	$0.195	0.45s	55 tok/s	99.96%
50% off	$1.32$0.6587	$3.96$1.976	$0.042$0.02096	1.00s	35 tok/s	99.76%
50% off	$1.32$0.66	$3.96$1.98	$0.044$0.022	1.43s	37 tok/s	100.00%
20% off	$1.32$1.056	$3.96$3.168	$0.044$0.0352	2.07s	46 tok/s	99.92%
	$1.35	$2.40	$0.195	0.42s	47 tok/s	98.17%
	$1.32	$3.96	$0.044	0.51s	46 tok/s	99.66%
	$1.32	$3.96	$0.04	0.46s	55 tok/s	99.07%
	$1.32	$3.96	$0.044	0.60s	31 tok/s	88.35%
25% off	$1.32$0.99	$3.96$2.97	$0.044$0.033	1.70s	27 tok/s	99.94%
	$0.66	$1.98	$0.066	1.02s	22 tok/s	99.97%
Fast	$1.98	$5.94	$0.066	0.32s	70 tok/s	100.00%
	$1.32	$3.96	$0.132	1.37s	21 tok/s	99.96%
	$1.32	$3.96	$0.044	1.09s	22 tok/s	99.90%
	$0.0999	$8.00	$0.0999	0.82s	14 tok/s	99.97%
	$1.65	$4.95	$0.165	1.47s	19 tok/s	98.98%
	$1.32	$3.96	$0.132	0.54s	55 tok/s	98.57%
	$1.30	$2.60	$0.10	1.39s	31 tok/s	99.78%
	$0.66	$1.98	$0.022	1.20s	53 tok/s	100.00%
	$0.5808	$1.7424	$0.05808	1.06s	22 tok/s	99.85%

Input $1.32Output $3.96
112 tok/s100.00% uptime
Input $1.32Output $3.96
115 tok/s99.57% uptime
Input $0.111Output $4.44
93 tok/s99.65% uptime
Input $1.31Output $3.96
92 tok/s99.42% uptime
Input $0.285Output $4.30
90 tok/s99.57% uptime
Input $1.35Output $2.40
55 tok/s99.96% uptime
Input $0.6587Output $1.976
50% off
35 tok/s99.76% uptime
Input $0.66Output $1.98
50% off
37 tok/s100.00% uptime
Input $1.056Output $3.168
20% off
46 tok/s99.92% uptime
Input $1.35Output $2.40
47 tok/s98.17% uptime
Input $1.32Output $3.96
46 tok/s99.66% uptime
Input $1.32Output $3.96
55 tok/s99.07% uptime
Input $1.32Output $3.96
31 tok/s88.35% uptime
Input $0.99Output $2.97
25% off
27 tok/s99.94% uptime
Input $0.66Output $1.98
22 tok/s99.97% uptime
Input $1.98Output $5.94
Fast
70 tok/s100.00% uptime
Input $1.32Output $3.96
21 tok/s99.96% uptime
Input $1.32Output $3.96
22 tok/s99.90% uptime
Input $0.0999Output $8.00
14 tok/s99.97% uptime
Input $1.65Output $4.95
19 tok/s98.98% uptime
Input $1.32Output $3.96
55 tok/s98.57% uptime
Input $1.30Output $2.60
31 tok/s99.78% uptime
Input $0.66Output $1.98
53 tok/s100.00% uptime
Input $0.5808Output $1.7424
22 tok/s99.85% 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

$0.1983/M tokens

Output

$3.326/M tokens

DeepSeek V4 Pro 0813 — Price history

Price history

No effective price data for this selection

	$0.0998	$8.00	69.0%	3.2%
	$0.1012	$4.44	69.5%	22.9%
	$0.1051	$3.17	93.2%	0.6%
	$0.1721	$3.96	90.0%	20.4%
	$0.1763	$1.98	75.8%	6.2%
	$0.1781	$1.98	75.5%	15.6%
	$0.1800	$3.96	89.3%	0.8%
	$0.1863	$2.97	84.0%	3.6%
	$0.2247	$1.98	68.0%	4.4%
	$0.2842	$4.30	71.3%	4.3%
	$0.2865	$3.14	80.0%	1.3%
	$0.2879	$3.96	80.7%	8.7%
	$0.3966	$3.96	77.6%	3.9%
	$0.4584	$3.96	67.3%	0.4%
	$0.4587	$2.60	70.1%	2.2%
	$0.5755	$3.65	56.6%	0.2%
	$0.6880	$3.96	49.5%	0.3%
	$0.8245	$2.40	45.5%	0.2%
Fast	$0.8697	$5.94	58.0%	0.2%
	$0.9113	$2.40	38.0%	0.3%
	$0.9991	$3.96	25.1%	0.2%
	$1.1121	$3.96	17.5%	0.1%
	$1.1473	$3.96	14.5%	0.1%
	$1.2101	$4.95	29.6%	0.1%

Effective in $0.0998Effective out $8.00
Token share
3.2%
Effective in $0.1012Effective out $4.44
Token share
22.9%
Effective in $0.1051Effective out $3.17
Token share
0.6%
Effective in $0.1721Effective out $3.96
Token share
20.4%
Effective in $0.1763Effective out $1.98
Token share
6.2%
Effective in $0.1781Effective out $1.98
Token share
15.6%
Effective in $0.1800Effective out $3.96
Token share
0.8%
Effective in $0.1863Effective out $2.97
Token share
3.6%
Effective in $0.2247Effective out $1.98
Token share
4.4%
Effective in $0.2842Effective out $4.30
Token share
4.3%
Effective in $0.2865Effective out $3.14
Token share
1.3%
Effective in $0.2879Effective out $3.96
Token share
8.7%
Effective in $0.3966Effective out $3.96
Token share
3.9%
Effective in $0.4584Effective out $3.96
Token share
0.4%
Effective in $0.4587Effective out $2.60
Token share
2.2%
Effective in $0.5755Effective out $3.65
Token share
0.2%
Effective in $0.6880Effective out $3.96
Token share
0.3%
Effective in $0.8245Effective out $2.40
Token share
0.2%
Effective in $0.8697Effective out $5.94
FastToken share
0.2%
Effective in $0.9113Effective out $2.40
Token share
0.3%
Effective in $0.9991Effective out $3.96
Token share
0.2%
Effective in $1.1121Effective out $3.96
Token share
0.1%
Effective in $1.1473Effective out $3.96
Token share
0.1%
Effective in $1.2101Effective out $4.95
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

115tok/s

P50, best across providers

Latency

0.25s

P50, best provider

Reasoning effortLatency / throughput

AutoExacto benchmarks

GPQA DiamondTAU-BenchBaseten90.8%--Parasail89.4%82.7%Phala92.3%78.7%StreamLake92.2%78.7%NovitaAI91.4%78.7%Parasail (Fast)90.9%74.7%

+19 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)

99.72%

Availability over the last 3 days

Last 72 hours

Availability 99.72%

3 Days Ago2 Days AgoYesterdayNow

Availability over the last 24 hours

OpenRouter Availability

99.75%

Without Routing

57.89%

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 0813 (Artificial Analysis and Vals AI)
Source	Benchmark	Score
Artificial Analysis	DeepSeek V4 Pro 0813 (Max) Intelligence Index	23.1
Artificial Analysis	DeepSeek V4 Pro 0813 (Max) Coding Index	68.8
Artificial Analysis	DeepSeek V4 Pro 0813 (Max) Agentic Index	41.3
Artificial Analysis	DeepSeek V4 Pro 0813 (Max) GPQA Diamond	92.8%
Artificial Analysis	DeepSeek V4 Pro 0813 (Max) HLE	41.0%
Artificial Analysis	DeepSeek V4 Pro 0813 (Max) AA-LCR	80.3%
Artificial Analysis	DeepSeek V4 Pro 0813 (Max) τ-Bench Banking	39.6%
Artificial Analysis	DeepSeek V4 Pro 0813 (Max) GDPval-AA	47.8%
Artificial Analysis	DeepSeek V4 Pro 0813 (Max) CritPt	18.0%
Artificial Analysis	DeepSeek V4 Pro 0813 (Max) SciCode	51.0%
Artificial Analysis	DeepSeek V4 Pro 0813 (Max) Terminal-Bench 4.0	14.1%
Artificial Analysis	DeepSeek V4 Pro 0813 (Max) Terminal-Bench 2.1	78.7%
Artificial Analysis	DeepSeek V4 Pro 0813 (Max) AA-Omniscience Accuracy	45.2%
Artificial Analysis	DeepSeek V4 Pro 0813 (Max) AA-Omniscience Non-Hallucination Rate	5.9%
Vals AI	DeepSeek V4 Pro 0813 Code Migration v1.1	52.0%
Vals AI	DeepSeek V4 Pro 0813 CorpFin v2	65.4%
Vals AI	DeepSeek V4 Pro 0813 CyberBench v1.1	72.9%
Vals AI	DeepSeek V4 Pro 0813 EMB v1.1	65.1%
Vals AI	DeepSeek V4 Pro 0813 Finance Agent v2.1	57.1%
Vals AI	DeepSeek V4 Pro 0813 Vals Finance Index	52.9%
Vals AI	DeepSeek V4 Pro 0813 GPQA Diamond	92.4%
Vals AI	DeepSeek V4 Pro 0813 HLAB-Vals-1.1	10.8%
Vals AI	DeepSeek V4 Pro 0813 Immigration	61.7%
Vals AI	DeepSeek V4 Pro 0813 IOI	51.6%
Vals AI	DeepSeek V4 Pro 0813 LiveCodeBench	87.5%
Vals AI	DeepSeek V4 Pro 0813 LegalBench	82.4%
Vals AI	DeepSeek V4 Pro 0813 Vals Legal Index	69.0%
Vals AI	DeepSeek V4 Pro 0813 Legal Research Bench	40.9%
Vals AI	DeepSeek V4 Pro 0813 MedCode	42.5%
Vals AI	DeepSeek V4 Pro 0813 MedScribe	80.2%
Vals AI	DeepSeek V4 Pro 0813 MMLU Pro	87.0%
Vals AI	DeepSeek V4 Pro 0813 ProgramBench	0.0%
Vals AI	DeepSeek V4 Pro 0813 ProofBench v1.1	50.0%
Vals AI	DeepSeek V4 Pro 0813 SkillsBench	53.8%
Vals AI	DeepSeek V4 Pro 0813 SRE Bench	1.9%
Vals AI	DeepSeek V4 Pro 0813 SWE-bench	96.4%
Vals AI	DeepSeek V4 Pro 0813 Tax Agent Bench	26.8%
Vals AI	DeepSeek V4 Pro 0813 TaxEval v2	73.1%
Vals AI	DeepSeek V4 Pro 0813 Terminal-Bench 4.0	14.1%
Vals AI	DeepSeek V4 Pro 0813 Terminal-Bench-Science	4.3%
Vals AI	DeepSeek V4 Pro 0813 Vals Index	53.0%
Vals AI	DeepSeek V4 Pro 0813 Vibe Code Bench 1-100	17.5%
Vals AI	DeepSeek V4 Pro 0813 Vibe Code Bench v1.1	82.3%

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.

355Btokens

marshal-harness

new

129Btokens

Claude Code

Claude Code is Anthropic's agentic coding tool that reads your entire codebase, plans and executes changes across files, runs tests, and iterates on failures, all from natural language prompts.

126Btokens

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

119Btokens

ISEKAI ZERO

AI adventures. Travel with your favorite characters

95.7Btokens

Rest of OpenRouter

35.6B

Top 5 Apps

9.95B

DeepSeek: DeepSeek Pro versions

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 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 Pro 0423

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

Text1.0M context$0.2079 / $0.4157

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 0813

Providers