Wednesday, November 18, 2026
Late edition · Updated as it breaks
Log inSubmit a story

“All the news worth catching.”

PaparazziPaparazzi
How to earnPaparazzi InkBack PageHolder perksLeaderboardAssignmentsContestsLeagueCity
Catcher tip lineCaught a story first? Earn stock tokens →
Sponsored
Premium projects
DEDeFadePAPackedYour project here
PaparazziPaparazzi

All the news worth catching.

Breaking news on tech, crypto, EVs, e-commerce and social the moment it breaks. Written by AI from verified reporting.

SectionsTop storiesRobinhood ChainTech & AIE-commerceWeb3 & CryptoEV & EnergyGlobal & SocialEntertainmentGamingSportsBusiness & marketsHealth & sciencePolitics & world
EarnStart hereSubmit a storyHow to earnLeaderboardContestsCatcher desksTop catchers by reputationAssignmentsCity LeagueNews bountiesCityMy catchesRewards
MarketsLive statsProof of payoutsWhale watchDeep Scan token checksStock fundToken utilityRobinhood Chain launchesRobinhood Chain projectsPremium project pagesPartnersThe Chain Weekly
PaparazziPaparazzi InkPaparazzi Ink top writersThe Back PageSuggest a Back Page storyPaparazzi on TelegramHow it worksFAQProof of workAbout usHow we score trustEditorial policyPrivacy policyCorrectionsRelease notesAdvertiseContact
$PAPA contract · Robinhood Chain
© 2026 Paparazzi

Written by AI from verified reporting. Not financial advice.

  • Home
  • Submit
  • Earn
  • Sign in
Top storiesRobinhood ChainTech & AIE-commerceWeb3 & CryptoEV & EnergyGlobal & SocialEntertainmentGamingSportsBusiness & marketsHealth & sciencePolitics & worldWhales
Tech & AI
Important· 52

Xingchen unveils Xing 4, a 285B-parameter model built on Ascend chips

Xingchen AGI Lab has launched Xing 4, a 285-billion-parameter language model family trained entirely on Ascend 910C chips. The TeleChat successor uses MindSpore HyperParallel across 8,192 cards and adds TeleGym for agent post-training. A 29-billion-parameter MoE member is being open-sourced.

By the Paparazzi desk November 18, 2026 · 6:36 AM · blockchain.news · 1 view
Fact boxTrust 50/100 · Developing
What we verified
  • Original source: blockchain.news.
  • The source published it on Nov 18, 2026 at 6:36 AM.
What we couldn't
  • No second outlet has reported it yet. Treat it as a single-source claim.

Xingchen AGI Lab pulled the wraps off Xing 4, a family of large language models that pushes the TeleChat line into 285-billion-parameter territory. The research landed on arXiv on November 17, 2026, under paper number 2611.22115, and it is aimed squarely at domestic silicon — Ascend 910C chips rather than imported accelerators.

The training stack is built on MindSpore HyperParallel, a framework the lab tuned for large-scale distributed work. It leans on topology-aware hybrid sharding, efficient distributed optimizer execution, and CPU affinity with core isolation to keep scheduling interference down. Those tweaks let Xingchen run pretraining across 8,192 Ascend 910C cards, handling long-sequence workloads and activation memory management along the way.

Sponsored

On architecture, the team balanced sparse activation, attention design and speculative decoding to squeeze out compute and memory efficiency. A mixture-of-experts topology paired with lightweight correction heads sharpens speculative decoding, which the researchers say keeps execution stable in long-context scenarios while supporting reasoning and agentic tasks.

Post-training runs through TeleGym, a unified agent execution platform that bundles execution, evaluation and trajectory construction with policy optimization. Disaggregated asynchronous execution, bounded oversampling and partial rollout let it cope with tasks of wildly different durations without breaking trajectory continuity across policy updates.

Xingchen describes the work as a step toward building advanced AI on homegrown hardware, giving future LLM development a scalable base for complex reasoning and multi-step planning. The lab is also open-sourcing a 29-billion-parameter member of the mixture-of-experts family.

TopicsXingchen AGI Lab · Xing 4 · TeleChat · Ascend 910C · MindSpore HyperParallel · TeleGym

Share

Source: blockchain.news

Written by AI from verified reporting.

More stories

  • Tech & AI 8 minutes ago · coindesk.com · Caught by @gimbung · 0 views
    Google finds an AI that hunts and exploits software flaws without a human
  • Tech & AI 1 hour ago · x.com · 1 view
    California AG says his office is looking into the Nvidia scandal
  • Tech & AI 2 hours ago · cryptobriefing.com · 1 view
    Nvidia's Singapore Unit Books $104B for 2025 as Reporting Rules Shift
  • Tech & AI 2 hours ago · x.com · 1 view
    Nvidia Board Faction Pushes for Jensen Huang's Return as CEO, per Semafor
Catcher tip line

Caught a story we missed?

Log in and send it in. If it checks out, it runs with your name on it and you earn stock tokens. The bigger the story, the bigger the reward.

Submit a story
The newsletter

Get the paper in your inbox.

A short digest of the biggest news, plus optional breaking alerts. Free, no account needed.