#DeepSeek
318 related articles
Qwen 4 First Test Leaks: Performance Rivals Opus 5.5, DeepSeek V4 Series Surfaces
Early Qwen 4 test outputs leak, with front-end generation said to rival Opus 5.5; DeepSeek V4 Mini, Flash, and Pro names appear in a code repo; Ant Group's Ling-3.1-flash surfaces too. Here's the latest on model tiering and on-device lightweight models.
DeepSeek V4.1 Flash 1017 Released: Reasoning and Coding Upgrades, Search and Fetch API Now Available
DeepSeek releases V4.1 Flash 1017 with major gains in reasoning, coding, and agentic tasks, plus web search and fetch APIs at 5 yuan per 1,000 requests. Tencent Hunyuan open-sources High Memory 2.0, Perplexity open-sources multimodal decision models, and MOFCOM seeks comments on AI export catalog.
DeepSeek 1017 Hands-On: All 11 Benchmarks Up, Costs Down 91%, Ready to Ship Projects
DeepSeek 1017 hands-on: all 11 official benchmarks up, averaging 5.2 points; terminal +11.1, coding +8.8, four leads over Opus 5; costs down 76-91%, 460M tokens ≈ 24 yuan; builds full projects like a web racing game from few prompts; retains 552B params, 1M context, MIT open source.
AI Model Selection Guide: DeepSeek vs. Qwen3 in Practice
How should AI developers choose models? Based on a Bilibili tutorial, this article breaks down the Function Calling differences between DeepSeek V3 and R1, Qwen3's programmatic thinking mode switching and agent capabilities, and outlines a multi-model learning path and practical selection logic.
DeepSeek V4.1 Flash 1017 Breakdown: It Won the Benchmarks, But at What Cost?
DeepSeek V4.1 Flash 1017 update 37 days later: official benchmarks improve but third-party tests show the main win is shorter output. Breakdown of benchmarks, cost accounting, and hidden costs of a 5-hour outage, with upgrade advice for three user types.
Behind the US Ban on DeepSeek: 20% API Share and China's Hardware Breakthrough
The US DHS emergency directive orders federal agencies to remove DeepSeek, amid DeepSeek's global API call share exceeding 20% and surpassing Google and OpenAI, plus Changxin Storage's 4F2 architecture raising the capacity ceiling of domestic HBM. This article unpacks why open-source low-cost models pose a threat and why the focus of large-model competition is shifting to unit inference cost.
DeepSeek Open-Sources DeepInfer Inference Framework; Overseas Users Call It 'the Real OpenAI'
Three days after V4.1 Flash, DeepSeek open-sourced the DeepInfer inference framework to cut inference compute costs and energy use. Overseas developers call it 'the real OpenAI.'
DeepSeek's Open-Source Infra Map: What Ordinary Programmers Should Understand
A visual guide to DeepSeek's open-source infrastructure: Flash MLA, DeepGEMM operators, DeepEP communication, 3FS storage, and DeepInfer inference—what each library solves, how they fit together, and what ordinary programmers can take away.
DeepSeek Turned into a Yandere Blue Fish? A Bilibili Fan-Creation Recap
A Bilibili uploader turns DeepSeek into a "big blue fat fish" with a yandere persona through onomatopoeia collage and role-play. This article breaks down the video's narrative techniques and analyzes the communication logic and brand impact of community repackaging of large models.
DeepSeek Open-Sources DeepInfer: How Inference Frameworks Make AI Serving Cheaper
DeepSeek open-sources the DeepInfer inference framework, revealing a cluster handling 155 trillion tokens per day. This article uses a restaurant analogy to explain prefill/decode separation, 96% KV cache hit rate, DeepSpark speculative decoding, and expert load balancing, analyzes how inference frameworks lower AI serving costs, and whether ordinary developers can use it directly.
Behind the U.S. Government Procurement Ban: Why Does DeepSeek Affect All Chinese Model Companies?
U.S. government procurement rules restrict DeepSeek and extend to third-party integrations, revealing supply chain review logic. This article analyzes the cost drivers of U.S. companies adopting Chinese models and the replicable impact on Chinese model companies going overseas.
DeepSeek Responds to the $38 Windfall Claim: The Compute Ledger and the Open-Source Gambit
DeepSeek responds to the $38 per GPU-hour windfall claim: $38 is theoretical full-load revenue; real costs include H100 purchases, data center electricity, and depreciation. The open-source framework is hard for OpenAI to migrate due to CUDA ecosystem lock-in, truly benefiting domestic compute and smaller players.
DeepSeek Open-Sources DeepInfer: How a 72-Hour Context Cache Rewrites AI Inference Costs
DeepSeek releases open-source inference framework DeepInfer, using 72-hour context caching to slash AI inference costs—up to $38 revenue per GPU per hour. We break down the tech, hardware strategy, and the open vs. closed source battle.
Model Fingerprint Fiasco: The DeepSeek Distills GPT Rumor and the Open-Source Counterattack
DeepSeek accused of distilling GPT and Claude? Model fingerprint tests produce absurd time-travel results, and the tool's built-in fingerprint library contains only GPT and Claude. Meanwhile, DeepSeek open-sources DeepInfer, countering doubts with transparent engineering.
DeepSeek Open-Sources DeepInfer Inference Engine — Why Did It Trigger a US Ban?
DeepSeek open-sources its inference engine DeepInfer for commercial use with impressive performance, yet it was banned by US federal civilian agencies. This article analyzes DeepInfer's technical features, the significance of open sourcing, and the game behind the ban.
DeepSeek Open-Sources Inference Framework DeepInfer: The Secret to Halving AI Inference Costs
DeepSeek open-sources DeepInfer inference framework, revealing online inference engineering secrets: PD separation, KV Cache, MoE load balancing, and speculative decoding—how it earns $38/GPU/hour, avoids recomputing 96% of input, plus open-source boundaries and industry impact.
Nexus-CLI: A Pure Go DeepSeek Bridge Tool
Nexus-CLI is a pure Go bridge tool integrating DeepSeek R1/V3 with Claude Code and Ollama. This article analyzes its architecture, core capabilities, and early community feedback.
DeepSeek Open-Sources DeepInfer Inference Engine: What It Can Run, What It Lacks
DeepSeek open-sourced DeepInfer, the inference engine behind its API, but it's offline-only, supports only V4.1-Flash, and runs only on Blackwell or Ascend. We break down the three constraints.
DeepSeek Open-Sources Inference Engine DeepInfer: $22M Daily Revenue per Cluster, 81-Day Payback
DeepSeek open-sources its inference engine DeepInfer, claiming 155T tokens/day on a single cluster, $22M daily revenue, ~80% gross margin, and 81-day payback. An analysis of its performance data, technical logic, and potential impact on the AI inference industry.
DeepSeek V4.1 Flash Hits 77% Coding Pass Rate, Search API Opens at Half Price
Two days after an outage, DeepSeek releases V4.1 Flash 1017 with 77% coding pass rate, beating GPT-5.6 and Claude Opus 5, and opens Search and Fetch APIs at 5 yuan per thousand calls—half the price of competitors.