MindPattern
← The Wire

Public story · 2026-03-04 · source-backed

DeepSeek V4 Imminent

Confidence
source-backed
Sources
1
Redaction
passed

Story

Expected this week, timed to China's Two Sessions. Specs: ~1T parameters, ~32B active per token (MoE), native multimodal, 1M-token context. Designed for Huawei Ascend chips with zero NVIDIA dependency. Leaked benchmarks claim 90% HumanEval and 80%+ SWE-bench (unverified). Consumer tier: dual RTX 4090s or single RTX 5090. (TechNode)

Related stories

Each link below shares sources, entities, or timing with this story.

  1. Shared entities / Shared topic / What happened next

    DeepSeek V4 Drops: MIT-Licensed Frontier Model at 1/6th the Cost, Running on Chinese Chips

    Both cover MoE, NVIDIA, SWE; overlapping topics (active, chip, context, deepseek, huawei); picks up the MoE thread on 2026-04-26.

  2. Shared entities / What happened next / Tension

    Kimi K2.6 Beats GPT-5.4 on SWE-Bench Pro. It's Open-Weight. It Costs 88% Less.

    Both cover Huawei Ascend, MoE, NVIDIA, SWE; picks up the Huawei Ascend thread on 2026-05-05; pushes against this story (vs).

  3. Shared entities / Shared topic / What happened next

    DeepSeek V4 Preview Drops: 1.6T Parameters, 1M Context, Apache 2.0, and a 35x Price Gap That Should Make Every Builder Rethink Their Inference Budget

    Both cover MoE, NVIDIA, SWE; overlapping topics (benchmark, context, deepseek); picks up the MoE thread on 2026-04-24.

  4. Shared entities / Shared topic / Earlier coverage

    DeepSeek V4: Day 6, Still No Official Launch

    Both cover HumanEval, SWE; overlapping topics (benchmark, claim, context, deepseek, dual); earlier HumanEval coverage from 2026-02-23.

  5. Shared entities / Shared topic

    GLM-5: 744B Open-Source, Free on NVIDIA NIM

    Both cover Huawei Ascend, MoE, SWE; overlapping topics (ascend, benchmark, chip, context).

  6. Shared entities / Shared topic / Earlier coverage

    DeepSeek V4: Day 7+, Still Not Officially Launched. V4 Lite Leaks.

    Both cover HumanEval, Leaked, SWE; overlapping topics (benchmark, context, deepseek); earlier HumanEval coverage from 2026-02-24.

  7. Shared entities / What happened next / Tension

    A 27B Model Just Matched the 400B One. It Fits on Your Laptop.

    Both cover MoE, RTX, SWE; picks up the MoE thread on 2026-04-23; pushes against this story (versus).

  8. Shared entity: RTX / Shared topic / Earlier coverage

    DeepSeek V4 Expected Mid-February

    Both cover RTX; overlapping topics (benchmark, consumer, context, deepseek, designed); earlier RTX coverage from 2026-02-12.

Source trail

TechNode

Entities

ChinaConsumerDeepSeek V4 Imminent ExpectedDesignedHuawei AscendHumanEvalLeakedMoENVIDIARTXSpecsSWETechNode

Provenance

Canonical issue
Ramsay Research Agent — 2026-03-04
AI generated
no
Story unit
2026-03-04-deepseek-v4-imminent
Labels
source-backed, canonical briefing excerpt