DeepSeek-R1
DeepSeek-R1 is an open-weight reasoning model from DeepSeek, released on January 20, 2025 under the MIT licence, with 671B total and 37B activated parameters.
Retired. DeepSeek’s API no longer serves DeepSeek-R1: its change log records deepseek-reasoner moving to DeepSeek-R1-0528 on May 28, 2025 and to DeepSeek-V3.1 on August 21, 2025, and its pricing page now lists only DeepSeek-V4 models. The weights remain on Hugging Face.
Released
January 20, 2025
Type
reasoning
License
open-weight
MIT
Context
128,000 tokens
Benchmarks
| AIME 2024DeepSeek’s figure, pass@1. | 79.8% |
| GPQA DiamondDeepSeek’s figure, pass@1. | 71.5% |
| SWE-bench VerifiedDeepSeek’s figure, resolved. | 49.2% |
Scores are the developer's own published figures for this model, under the test conditions noted. Benchmarks with no primary source are not listed.
Capabilities
Architecture
Parameters: 671B total, 37B active (mixture of experts)
Sources
Every figure on this page was checked against these pages on . A figure we could not source is left out.
- DeepSeek: DeepSeek-R1 release (opens in a new tab)
- Model card: DeepSeek-R1 (opens in a new tab)
- DeepSeek API pricing (opens in a new tab)
- DeepSeek API docs: Change log (opens in a new tab)
Spotted something out of date? Tell us.
Links
In the news: DeepSeek-R1 and DeepSeek
MarkTechPost
DeepSeek Open-Sources DeepInfer, the Inference Framework Behind DeepSeek-V4.1-Flash
DeepSeek has open-sourced DeepInfer, the inference framework that powers its production services and the DeepSeek-V4.1-Flash model. This release, available under the Apache 2.0 license, provides developers with a rare view into a frontier lab's real serving stack. The company also released deepseek-recipe, an MIT-licensed token encoding and decoding layer. DeepInfer is described as a high-performance inference engine built primarily on vLLM, incorporating ideas from SGLang and TensorRT-LLM. DeepInfer supports one model, DeepSeek-V4.1-Flash, with a 1 million token context window. It runs on NVIDIA SM100 (Blackwell) GPUs and Huawei Ascend 950 NPUs. Benchmarks utilized 4 GPUs for prefill and 64 for decode. DeepSeek claims the framework achieves production-grade throughput, with a single cluster serving 155 trillion tokens per day. The system offers over 25,000 output tokens/s/GPU performance at 256 requests per rank with a 131,072 context. The engine incorporates DeepSeek's kernels, including FlashMLA, DeepGEMM, DeepEP, TileKernels, DeepSelect, and custom TileLang kernels. System-level features include Attention DP plus Wide-EP Attention for parallelism, and expert parallelism with an expert-parallel load balancer for MoE layers.
MarkTechPost
DeepSeek Just Released V4.1-Flash-1017, Scores 77.0 on DeepSWE, Opens Search API
DeepSeek has released DeepSeek-V4.1-Flash-1017, a new checkpoint of its V4.1-Flash model, available on its website, app, and API. The company states this version offers stronger scores, faster task completion, and shorter outputs compared to the September release, describing it as a Pareto improvement in performance and cost. Alongside the model update, DeepSeek has opened its Search and Fetch APIs to all developers. This provides agent builders with access to a more cost-effective coding model and a first-party web layer accessible through a single API key. DeepSeek-V4.1-Flash-1017 features a 552 billion backbone parameter count and 196 billion Engram parameters, activating 8 billion parameters per token during prefill and 16 billion during decode. It supports a context window of 1 million tokens with up to 384,000 output tokens. The model, served as deepseek-flash on the DeepSeek API, reportedly beats the previous V4.1-Flash on all 11 reported benchmarks, with improvements ranging from 2.0 to 11.1 points. Its best score is 77.0 on DeepSWE v1.1, surpassing Claude Opus 5's score of 74.0. DeepSeek-V4.1-Flash-1017 is an updated post-training checkpoint of DeepSeek-V4.1-Flash, utilizing reinforcement learning with self-feedback for end-to-end project work, including tasks like game creation and 3D modeling. It is tuned for DeepSeek Harness, Tencent CodeBuddy, and WorkBuddy.
MarkTechPost
openKylin Outlines Its Vision for Reimagining Linux in the AI Era at Open Source Summit Europe 2026
At Open Source Summit Europe 2026, openKylin introduced its vision for reimagining Linux in the AI era. Min Liu, Executive Director of openKylin, presented KART, a unified AI runtime, and a system-level agent execution environment. Jianfeng Li discussed building an AI-ready Linux desktop on RISC-V. openKylin, an open-source operating system community originating in China, aims to develop an AI-integrated OS for global users. The project showcased its latest work on AI and operating system integration through presentations and developer exchanges. The community believes the key challenge for AI innovation is no longer just model capability, but the system infrastructure connecting AI with everyday computing. Current AI stacks are fragmented, and hardware-software integration remains complex. To address this, openKylin proposes a shift in operating system management from "resource management" to "intent management." Traditional operating systems manage CPUs and memory, while AI-era operating systems need to manage AI functionalities more directly.
MarkTechPost
Reflection AI Introduces Beam: A 501B Open-Weight MoE Model With 23B Active Parameters for Coding and Agentic Workloads
Reflection AI has introduced Beam, its inaugural open-weight model, on March 18, 2024. Beam is a sparse Mixture-of-Experts (MoE) architecture boasting a total of 501 billion parameters, with 23 billion active parameters per token. This model is specifically engineered for coding, reasoning, and agentic workloads. According to the Reflection AI team, Beam is positioned to directly compete with larger open models such as GLM 5.2, while demonstrating a significant advantage in inference compute efficiency, using 3 to 4 times less compute on reasoning benchmarks. Currently, Beam is not available for self-hosting but is undergoing final red-teaming. Early access is being managed through a waitlist on the Reflection platform. Beam functions as a general agent model, developed from the ground up by Reflection AI, with a primary focus on enterprise-level coding and agentic applications. Reflection AI frames Beam as a significant advancement in the Western open-weight model landscape. The research team acknowledges that while Kimi K3 may still lead in raw capability, Beam's key selling proposition is its efficiency during inference. Users can adjust a "reasoning effort" parameter, where lower settings yield concise answers and higher settings enable more extensive reasoning for complex tasks. This allows teams to tailor the model's effort to the specific difficulty of a task and their available compute budget. The pretraining of Beam involved an extensive dataset of 23.8 trillion tokens, sourced from the web, public repositories, and proprietary licensed datasets. The Reflection AI team reported that their curation process removed approximately 95% of raw internet tokens, while retaining about 1.8 trillion high-quality tokens that might have been discarded by conventional filtering methods. The model's architecture integrates local and global attention mechanisms with finely routed experts. Its load balancing strategy builds upon DeepSeek-V3's auxiliary-loss-free balancing, incorporating a cosine decay for expert-bias updates. This approach ensured that the most utilized expert experienced a load of only 1.04 times the average by the conclusion of pretraining. Throughout the 52 layers, residual norms were maintained within bounds through depth-based scaling, SandwichNorm, attention gating, and FP32 residual accumulation. The entire pretraining process was completed in under four weeks, utilizing 6,144 NVIDIA GB300 NVL72 GPUs. The reported goodput reached 92.3% near the end of the training period, with nine semi-automatic rewinds implemented. During mid-training, the effective context window was extended to 1 million tokens.
MarkTechPost
DeepSeek Harness v0.2 Brings Official Desktop Apps to Its Open-Source Agent Harness
DeepSeek has released an official desktop application for its open-source agent harness, DeepSeek Harness (dsh), with the v0.2 preview. Installers are available for macOS, specifically for Apple silicon, and for Windows 64-bit operating systems. This preview version is deployable immediately and can be downloaded from deepseek.com/harness or accessed via the command line using 'npx @deepseek-ai/dsh web'. DeepSeek has cautioned that compatibility-breaking changes are expected in future releases. The v0.2 preview enhances the harness's capabilities to target both everyday work and coding tasks. Key built-in features include preloaded, commonly used tools for office productivity and software development. A new plugin management interface allows users to install, configure, enable, and disable plugins through a dedicated page. For code and file review, a right-hand sidebar provides previews of generated files and code changes. The harness now supports document work, enabling users to input documents, spreadsheets, or PDFs and request the generation of charts or slides. An automation task plugin has been introduced, which schedules recurring prompts and includes a run history with editable frequency settings. The desktop build also integrates the dsh command-line interface, eliminating the need for separate Node.js or pnpm installations. Users can authenticate using a DeepSeek account or by providing an API key. According to the release candidate 1 (rc.1) notes, account-based models gain access to web search functionality without requiring an additional API key. The architecture of DeepSeek Harness is designed around a plugin-centric approach, where the model adapter, tool registry, and agent loop are all implemented as plugins, allowing for modularity and replacement of any layer. DeepSeek first open-sourced dsh in August 2026 under the MIT license. The harness runs on Cordis, as detailed in the research paper ‘A Programming Paradigm for Spatiotemporal Composability’. The harness is not restricted to DeepSeek's own models, with a provider guide available for integrating third-party providers and custom OpenAI-compatible endpoints. The project's GitHub repository has garnered significant attention, exceeding 240,000 stars and 24,000 forks. DeepSeek Harness also serves as DeepSeek's internal evaluation framework. The API changelog indicates that Code Agent benchmarks, such as the Terminal Bench 2.1, have been run using DeepSeek Harness in a minimal configuration. For instance, the DeepSeek-V4-Flash-0731 model achieved a score of 82.7 on Terminal Bench 2.1 within this setup. The harness supports four distinct session modes.