Cardboard Iguana Security

Properties6
created2026-08-02
updated2026-10-21
typeentity
confidencemedium
sourceshttps://arxiv.org/pdf/2606.19348, https://api-docs.deepseek.com/updates/, https://www.geopolitechs.org/p/deepseek-founder-liang-wenfeng-in, https://www.federalregister.gov/documents/full_text/text/2026/10/20/2026-21447.txt, https://www.bankinfosecurity.com/dhs-bans-deepseek-from-federal-civilian-agencies-a-33146, https://thehackernews.com/2026/10/us-orders-deepseek-removed-from-federal.html
tagsai, llm, architecture, economics, reference

7 min read

DeepSeek

DeepSeek (深度求索) is a Chinese artificial-intelligence research lab founded in Hangzhou in May 2023 by Liang Wenfeng, funded by his quantitative hedge fund High-Flyer. Small (~150 people), flat, and explicitly organized around an AGI goal rather than a product roadmap, DeepSeek pairs fully open weights (MIT license) with an aggressively cheap per-token API. Its stated strategy is restraint: price to recoup hardware costs in roughly ten months, refuse consumer/enterprise monetization fights, and treat both as “side products on the road to AGI.” 1

Model lineage

ReleaseParams (total/active)Key advances
V2 (2024-05)236B / 21BMulti-head Latent Attention (MLA); DeepSeekMoE fine-grained experts
V3 (2024-12)671B / 37BAuxiliary-loss-free load balancing; multi-token prediction; FP8 training + DualPipe; ~2.788M H800-hours total
R1 (2025-01)671B / 37BRL-first reasoning via GRPO (critic-free group-relative advantages)
V3.1 → V3.2 (2025)671B / 37BHybrid thinking/non-thinking single model; agent-focused post-training
V4 Preview (2026-04-24)1.6T / 49B (Pro); 284B / 13B (Flash)CSA+HCA sparse attention; mHC; Muon optimizer; 1M-token context default; MIT license
V4-Flash GA “0731” (2026-07-31)284B / 13BSame architecture, re-post-trained only; beats V4-Pro-Preview on agentic benchmarks

The legacy deepseek-chat/deepseek-reasoner API aliases were hard-retired 2026-07-24; the V4-Pro GA remains pending (“will follow soon”) as of early August 2026. 2

Technical through-line

DeepSeek’s architecture is a continuous refinement of one bet: extreme MoE sparsity + aggressive attention compression + training efficiency, in service of frontier-adjacent capability at minimal compute. The V4 headline is a tiered sparse-attention stack replacing dense attention: HCA (Heavily Compressed Attention) compresses keys ~128× and runs dense attention on the coarse representation in early layers; CSA (Compressed Sparse Attention) compresses ~4× and applies DeepSeek Sparse Attention (DSA) — a lightweight “lightning indexer” scores compressed blocks and selects top-k tokens (1024 Pro / 512 Flash) for full attention in later layers. At 1M-token context, V4-Pro needs only ~27% of the single-token inference FLOPs of V3.2 (Flash ~10%), and mixed-precision KV storage (RoPE dims BF16, rest FP8) roughly halves the cache again — which is what makes a million-token context the default tier rather than a premium one. Supporting pieces: mHC (Manifold-Constrained Hyper-Connections) — residual streams widened 4× with the mixing matrix constrained to the doubly-stochastic manifold via Sinkhorn iterations — and the Muon optimizer, under which DeepSeek reports zero loss spikes across the ~32T-token pre-training run. 3

Business and compute position

  • Pricing as doctrine. V4-Pro launched 2026-04-24 at $1.74/$3.48 per M tokens, cut ~75% to $0.435/$0.87 within five weeks; Flash sits at $0.14/$0.28 with cache-hit input orders of magnitude lower. Liang’s pricing rule is hardware recoup in ~10 months, explicitly not profit maximization; a peak/off-peak surge model (2× Beijing business hours) is announced but not yet active. 4
  • Funding. June 2026 brought the first-ever external round (~$7.4B from Tencent, CATL, and others at a ~$50B valuation), with a second round and onshore IPO rumored — a research lab becoming a real company on Chinese strategic-capital terms.
  • Ascend pivot. V4 is optimized for Huawei Ascend silicon; Liang frames the China–US gap as “mainly resources… on people, almost no gap,” chasing with roughly 1/20th the compute, and treats domestic-chip adoption (Huawei 950 supernodes) as both an export-control hedge and an ecosystem investment. 5
  • Reality check. V4 is not the absolute capability leader — it trails the top US closed models on composite evals — and inside China it is the price floor, not the quality leader, with Moonshot, Zhipu, Alibaba, MiniMax and others pressing. Liang predicts consolidation: “There are too many base-model companies in China; it will converge.”

US government restrictions (2026)

On July 24, 2026, DHS Secretary Markwayne Mullin signed Exclusion and Removal Order 2026-01 under the Federal Acquisition Supply Chain Security Act of 2018; DHS filed it publicly on October 19 and published it in the Federal Register on October 20 (91 FR 66346, docket DHS-2026-0628). The order excludes IT products and services from Hangzhou DeepSeek Artificial Intelligence Co., Ltd and its affiliates from federal civilian agency procurement, and orders removal from civilian agency IT of all DeepSeek products plus third-party products and services “developed using DeepSeek implementations.” It followed a year of agency-level device bans (by February 2025: Congress, NASA, the Navy, the Pentagon, and Texas; then New York and Virginia) and the House Select Committee’s April 2025 “DeepSeek Unmasked” report; Australia, South Korea, and Taiwan keep DeepSeek off government systems as well. The order binds civilian agencies and their contractors, not the public or the Department of Defense. 67

The derivative clause sits awkwardly with DeepSeek’s MIT-licensed open weights: see adversary-software-supply-chain-bans for the instrument and the open-provenance problem “developed using DeepSeek implementations” poses.

Philosophy (leaked investor call, July 2026)

A leaked four-hour Liang Wenfeng investor call (talk dated 2026-05-20, public 2026-07-23; two independent English translations) is the richest statement of DeepSeek’s self-understanding: vision over organization (“we don’t really have an organization… that vision isn’t even written down”), open source as the point rather than the tactic (“I want [competitors] to be able to deploy it. I only worry they can’t”), restraint as strategy (“People who take more get beaten by people who take less”), and a roadmap of CoT → agents → continual learning → a gradual “singularity” → embodied intelligence — while refusing video, 3D, and world models as off the “main line of intelligence.” The call is internally consistent with every observable decision (price cuts, MIT license, tiny team), though authenticity is corroborated rather than officially confirmed. 8

Connections

  • chinese-innovation-patent-ecosystem: R1’s January 2025 release was the archetypal Chinese “Sputnik moment” in the patent-literature sense; DeepSeek is the firm-level proof point for that page’s measurement debates.
  • gongkai: DeepSeek’s open-weights, ecosystem-first strategy is gongkai’s quid-pro-quo logic in AI form — give away the artifacts, own the ecosystem position. Liang’s “open source is a cultural behavior” stance echoes the network-IP view bunnie Huang documented in Shenzhen hardware.
  • technological-singularity: Liang’s “singularity” is explicitly gradual and continual-learning-gated — a concrete lab-level operationalization of a usually abstract concept.
  • paperclip-maximizer: An AGI-aiming lab whose stated philosophy is taking less — a live counterpoint to pure instrumental-convergence pessimism, worth tracking as it scales.
  • specification-gaming-openai-hf-incident / anthropic-cybersecurity-eval-incidents: The frontier-lab cluster DeepSeek now belongs to structurally, whether or not it shares the incidents.
  • china-ai-datacenter-buildout: the physical layer under the Ascend pivot — DeepSeek is siting its own facilities in Ulanqab, and the chip gate its investors call a “resources gap” is the same export-control regime that makes the national buildout chip-gated rather than power-gated.

Caveats

  • The investor-call transcript is a leak (widely corroborated, not officially confirmed); funding/IPO details are partly rumor; benchmark numbers mix official self-reports and third-party evals.

Sources

See Also

Footnotes

  1. 2026 — DeepSeek founder Liang Wenfeng in His Own Words: 64 Quotes from DeepSeek’s Investor Call ↩

  2. 2026 — Change Log ↩

  3. 2026 — Deepseek V4 Tech Report ↩

  4. 2026 — DeepSeek founder Liang Wenfeng in His Own Words: 64 Quotes from DeepSeek’s Investor Call ↩

  5. 2026 — DeepSeek founder Liang Wenfeng in His Own Words: 64 Quotes from DeepSeek’s Investor Call ↩

  6. U.S. Department of Homeland Security 2026 — Notice of Order Regarding Exclusion and Removal of DeepSeek Products From Federal Civilian Executive Branch Information Technology Systems and Procurements ↩

  7. Swati Khandelwal 2026 — U.S. Orders DeepSeek Removed From Federal Civilian Systems, Bans New Purchases ↩

  8. 2026 — DeepSeek founder Liang Wenfeng in His Own Words: 64 Quotes from DeepSeek’s Investor Call ↩

Backlinks

Recent notes

See 577 more →

Graph View