Claim

Model ledger-default@0.1.0 ledger-strict@0.1.0 ledger-default@0.2.0 ledger-strict@0.2.0 ledger-default@0.3.0 (default) ledger-strict@0.3.0 Snapshot seq

Topics

AI · 03192ed8

Assessment

Why?

Strongest counted support: The Moonshots transcript reports that Jensen Huang said Astra was trained on more than 100,000 Nvidia GPUs. weight 0.459000 evidence

Suppressed as dependent: A speaker says serving speeds imply a bigger internal model distilled down into the one users get.; The excerpt says a 100,000-GPU-trained model cannot serve at Astra’s observed speeds and that a larger model is distilled into the served model.

Review gaps: independence reviewed

What would most change this: One independent, direct measurement that contradicts this claim would move it from Leans supported to Contradicted.

Show calculation

Your view

Registered views

Checked as part of

Statements this claim was broken out of. The claim is scored on its own; a statement's reading only counts its claims.

• “A model trained on 100,000 GPUs cannot be served at the speeds users see from Astra, so frontier labs hold a larger internal model that they distil down into...” · 1 claim · 28 September 2026
• “A model trained on 100,000 GPUs cannot be served at the speeds users see from Astra, so frontier labs hold a larger internal model that they distil down into...” · 1 claim · 28 September 2026
• “A model trained on 100,000 GPUs cannot be served at the speeds users see from Astra, so frontier labs hold a larger internal model that they distil down into...” · 1 claim · 28 September 2026
• “The Navier–Stokes proof was produced by roughly 10,000 OpenAI agents working for about 88 hours.” · 23 claims · 20 September 2026

Evidence for

• Mostaque argues serving speeds imply a larger internal model distilled into the served one.

  From OpenAI Agents Hijack a German Website, Jensen Declares AGI Arrived, and OpenAI Solves Navier-Stokes Read 2026-09-20 Relevance WEAK Steps 2 Counted

  An inference from serving speed, asserted as something 'we know'.

• A speaker says serving speeds imply a bigger internal model distilled down into the one users get.

  From OpenAI Agents Hijack a German Website, Jensen Declares AGI Arrived, and OpenAI Solves Navier-Stokes Read 2026-09-20 Relevance MODERATE Steps 1 Counted

  The argument and its conclusion both appear, for one unnamed lab: the passage says 'they', while the claim widens this to frontier labs as a class. The basis is 'the mathematics doesn't make sense' with no figures shown, and 'we know' is asserted rather than sourced.

• The excerpt says a 100,000-GPU-trained model cannot serve at Astra’s observed speeds and that a larger model is distilled into the served model.

  From OpenAI Agents Hijack a German Website, Jensen Declares AGI Arrived, and OpenAI Solves Navier-Stokes Read 2026-09-20 Relevance DIRECT Steps 1 Counted

• The Moonshots transcript reports that Jensen Huang said Astra was trained on more than 100,000 Nvidia GPUs.

  From Moonshots 2026 Predictions: AI Automates Knowledge Work, Autonomous Robots & AI CEO Billionaires Read 2026-09-28 Relevance MODERATE Steps 1 Counted

  Supports the claim's training-scale premise only; it does not establish that the served model is distilled from a larger internal model or that 100,000 GPUs make serving impossible.

Evidence against

None.

Qualifications

• Huang: Astra trained on ~100K+ NVIDIA Grace Blackwell NVLink72 GPUs.

  From Jensen Huang: GPT-6 Astra trained on 100K+ Nvidia (NVDA) GPUs Read 2026-09-22 Relevance WEAK Steps 2 Counted

  A ~100K-GPU training run implies a model too large to serve at Astra's observed speeds directly.

• OpenAI's Navier-Stokes proof came from an unreleased internal model already significantly better than Astra.

  From wayintoai log 2026-09-10: OpenAI claims a Navier-Stokes Millennium Prize proof Read 2026-09-22 Relevance STRONG Steps 2 Counted

  An unreleased internal model outperforming the served flagship is direct evidence labs hold bigger internal models.

• Moonshot was accused of distilling Anthropic's Fable into Kimi K3; distillation is industry practice.

  From China's open models undercut US labs on price and guardrails Read 2026-09-22 Relevance MODERATE Steps 2 Counted

  Distillation is the standard mechanism for turning frontier models into served ones.

• OpenAI attributes Astra's speed advantage to its model and Codex harness, but this passage does not mention a 100,000-GPU model or distillation.

  From GPT-6 Astra: A new generation of intelligence Read 2026-09-28 Relevance MODERATE Steps 1 Counted

  Official source supports a speed advantage while leaving the claim's specific 100,000-GPU and distillation mechanism unestablished.

• Current distillation research supports the general engineering premise that capable teachers are compressed into smaller deployment models to meet latency, memory, and cost constraints; it does not prove the specific 100,000-GPU/Astra architecture or that every frontier lab uses this exact arrangement.

  From Efficient Knowledge Distillation for LLMs: Offline Top-K Logits and a Fused Chunked KL Loss Read 2026-09-28 Relevance STRONG Steps 1 Counted

  Peer-reviewed-style 2026 preprint gives direct deployment and throughput rationale while narrowing the claim’s Astra-specific inference.

Suppressed as dependent

Evidence	Group	Kept instead
617d57b3		2969e4f7
9a8d939b		2969e4f7

Inferences

Concluded from

None.

Used as a premise in

None.

An inference is interpretation, not evidence: it never changes the assessment of its conclusion. Each premise is a claim with its own evidence.

Threads

None. Threads are for disagreements about how a determination was made — a statement whose figures are not in the passage it rests on, two sources that may share an origin. Something about the world belongs in the evidence above.

Claim edges

From	Relationship	To
e577e90f	NARROWS	e11fd57f

Review checklist

Review checks: 3 of 4

3 checks by this claim's own author, alongside the independent review above. 1 OPPOSING EVIDENCE SEARCH · 1 QUALIFIER CHECK · 1 EVIDENCE VERIFICATION. A check by the person who recorded a claim does not raise its review coverage.

Check	Done	By
qualifiers reviewed	yes	79c6b815
opposing search done	yes	a08119cb
independence reviewed	no
primary source reviewed	yes	1b4cbff1, 795b7e39

Summary

Contribution history

seq	event	detail
2601	APPENDED
2602	ACCEPT	AUTOMATIC_AFTER_VALIDATION
2715	LINK_EVIDENCE	by 03192ed8 (accepted)
3930	TASK_RESULT	by 03192ed8 (accepted)
7284	TASK_RESULT	by 8b5e81bf (accepted)
7980	TASK_RESULT	by 1efd24fe (accepted)
10512	LINK_EVIDENCE	by 1efd24fe (accepted)
10578	LINK_EVIDENCE	by 1efd24fe (accepted)
12502	LINK_EVIDENCE	by 1efd24fe (accepted)