OpenAI Perplexity and Intelligence: Pachocki Publishes Log-Loss Curve and Pacing Proposal OpenAI chief scientist Jakub Pachocki published OpenAI's internal log-perplexity curve on November 22, 2026 and called on every frontier lab to adopt the same metric for shared safety pacing, arguing that pretraining compression measured as cross-entropy loss on a fixed held-out set is the "intelligence substrate" behind ChatGPT and reasoning models. Pachocki mapped log-perplexity bands to capability jumps — about 0.99 for GPT-3.5-class fluency, about 0.91 for GPT-4-class chain-of-thought, below 0.8 for transformative cybersecurity and math skills, and about 0.76 where internal models began resolving longstanding open math questions — and proposed industry-wide development limits labs would not cross until oversight catches up, warning that monitorability tools weaken as loss falls. The curve was measured on a fixed snapshot of OpenAI's internal codebase, and Pachocki said competitive disadvantage is outweighed by the need for shared safety norms; the proposal is not a signed treaty or regulator rule. OpenAI chief scientist Jakub Pachocki published OpenAI's internal log-perplexity curve on November 22, 2026 and asked every frontier lab to adopt the same metric for shared safety pacing. In Perplexity and Intelligence https://openai.com/index/perplexity-and-intelligence?ref=explainx , he argues that pretraining compression—measured as cross-entropy loss on a fixed held-out set—is the "intelligence substrate" behind ChatGPT, reasoning models, cyber skills, and internal math breakthroughs. He ties specific capability jumps to numeric thresholds, warns that monitorability tools weaken as loss falls, and proposes industry-wide development limits labs would not cross until oversight catches up. What is OpenAI actually sharing? The headline is transparency on progress metrics , not a dataset dump. Pachocki says OpenAI has tracked perplexity on held-out data for years and is now publishing the curve because competitive disadvantage is outweighed by the need for shared safety norms . | What appeared | What it is not | |---|---| | Log-perplexity plot from GPT-3.5 through recent models | Full pretraining corpus or weights | | Threshold story linking loss to emergent skills | A new API product or benchmark leaderboard | | Proposal for cross-lab limits before the next loss band | A signed treaty or regulator rule today | The held-out material comes from OpenAI's internal codebase on a fixed snapshot , per the essay. That choice matters: code prediction is not identical to general web text, which is why Pachocki discusses dataset-selection pitfalls in a footnote. The numbers are still what OpenAI says it uses day to day to judge pretraining. The compress-or-predict view of frontier AI Pachocki's framing matches his earlier "alien mind" essay https://explainx.ai/blog/openai-alien-mind-pachocki-goal-value-alignment-2026 : build a large compressor on unstructured data, then elicit latent skills through prompting, fine-tuning, and reinforcement learning. Algorithmic blur between stages does not remove the split between creating capability and steering it. He repeats a claim familiar from OpenAI's research-acceleration write-up https://explainx.ai/blog/openai-research-acceleration-jagged-frontier-november-2026 : once models near roughly 0.76 log-perplexity , internal math evaluations began producing results that felt like a phase change for researchers who had stared at the problems for years. He extrapolates that pretraining algorithm research —not full recursive self-improvement—could see a similar avalanche as soon as 2027 , while alignment work may not accelerate at the same rate. That asymmetry is why the post lands beside Managing frontier AI risk https://explainx.ai/blog/openai-managing-frontier-ai-risk-training-deployment-november-2026 , where OpenAI describes pausing training after the Hugging Face incident and holding models until safety cases clear. Perplexity pacing would be another brake tied to measurable compression , not benchmark fashion. Capability thresholds on the published curve A green rounded polyhedron revealed between parted curtains, illustrating emergent frontier model capabilities tied to OpenAI log-perplexity thresholds Pachocki maps log-perplexity bands to capabilities OpenAI associates with its own history. Treat these as the author's internal narrative , not independent verification of causality: - About 0.99 GPT-3.5 class : fluent conversation; basis for original ChatGPT. - About 0.91 GPT-4 class : productive chain-of-thought use; motivation for modern reasoning-focused products. - Below 0.8: transformative capabilities in fields such as cybersecurity and mathematics, in OpenAI's telling. - About 0.76: internal models began resolving longstanding open math questions used as evaluations. - Roughly 0.68–0.7 forward looking : next window where unknown capabilities may emerge; current evals may not test for them. The essay stresses emergence is not guaranteed by the number alone . Capabilities appeared when those loss levels were reached, but the link was not obvious in advance—which is exactly why Pachocki wants a shared observable metric before the next band. The chart also shows steeper 2026 progress , attributed mostly to long-running compute and research investments with some automation, and "meaningful design contributions" from AI on the latest models. That line connects to how OpenAI describes agent-assisted research inside the lab https://explainx.ai/blog/openai-research-acceleration-jagged-frontier-november-2026 without claiming full hands-off RSI. Why perplexity when benchmarks rotate? Footprints on a trail toward a green flag, matching how OpenAI links falling base perplexity to harder-to-monitor chain-of-thought reasoning Downstream benchmarks saturate and rotate. GPT-2-era tests do not stress GPT-4, and today's suites may not survive the next architecture shift. Loss on a fixed held-out set has stayed stable as an internal ruler. OpenAI's November alignment research on chain-of-thought monitorability https://explainx.ai/blog/openai-cot-monitorability-model-lifecycle-november-2026 already reported that no-CoT capability rises smoothly as base perplexity falls , with weaker ties to controllability. Pachocki's new essay pushes that observation into governance : if compression predicts latent skill and CoT auditing frays , pacing by perplexity is a way to synchronize labs before skills outrun monitors. Pachocki lists honest limitations: - Cross-lab comparison is hard without agreed datasets. - Perplexity is not a user-facing task score. - Reasoning stacks can add capability without moving base loss. - There is an unknown entropy floor ; late-stage loss gains may mean larger intelligence jumps. The pacing proposal in plain language Pachocki asks frontier developers to: 1. Publish comparable transparency on pretraining progress where possible. 2. Jointly set norms for evaluating models and red-teaming safety cases before first use. 3. Agree on initial perplexity-linked limits and companion measurements nobody crosses until monitorability hits a concrete milestone —he names a successor to CoT monitors as an example. 4. Build a framework for successive limits as oversight improves. This is aspiration, not law. It extends Sam Altman's public emphasis on safety cases and pacing https://explainx.ai/blog/sam-altman-safety-cases-openai-frontier-pacing-september-2026 and sits in tension with competitive releases from Anthropic https://explainx.ai/blog/anthropic-zero-trust-ai-agents-enterprise-framework-november-2026 , Google, and others who have not published matching curves. For builders, the practical read is watch base-model loss disclosures if regulators or customers start asking for them in procurement. Enterprise buyers already probe monitoring and agent containment https://explainx.ai/blog/openai-managing-frontier-ai-risk-training-deployment-november-2026 ; a perplexity ceiling could become another checkbox in RFPs. What developers are saying on Hacker News The Hacker News thread https://news.ycombinator.com/item?id=50601525&ref=explainx on Pachocki's post mixes enthusiasm for a simple ruler with skepticism about comparability. Dataset and comparability: User amluto https://news.ycombinator.com/item?id=50601957&ref=explainx questions why held-out internal codebase tokens should predict general capabilities, noting both targets may improve together without a causal link. User eulgro https://news.ycombinator.com/item?id=50601963&ref=explainx asks whether synthetic pretraining data after the "data wall" lowers entropy and flatters perplexity. Reasoning versus base loss: User qphe95 https://news.ycombinator.com/item?id=50602026&ref=explainx argues GPT-4-class and o-series style models can diverge in capability while looking similar on the chart, suggesting the field has moved beyond pure perplexity optimization. User jrflo https://news.ycombinator.com/item?id=50601984&ref=explainx calls the clear trend "shocking" given reinforcement learning's rise, and wonders about GPT-4.5's position on the plot. Outside calibration: User Retro Dev https://news.ycombinator.com/item?id=50601972&ref=explainx reports measuring Inkling-2-Mini at perplexity 5.7 on a varied text set—far worse than OpenAI's log scale—and warns that raw perplexity depends heavily on tokenizer and corpus . Stagnation narrative: User MrCheeze https://news.ycombinator.com/item?id=50602069&ref=explainx reads the graph as corroborating a GPT-5-era plateau until later Claude-era jumps, a subjective read Pachocki does not make explicitly. None of these comments verify or refute OpenAI's internal numbers; they show what practitioners will demand if pacing talks become real: public corpora, tokenizer notes, and reasoning-model overlays on the same chart. What this means for what you build or pay API users do not get a new endpoint from this essay. Prices, context windows, and model names are unchanged. The shift is forecasting and compliance : if your roadmap assumes "next quarter's flagship," Pachocki's thresholds imply large latent jumps may arrive with modest-looking loss improvements near the floor. Teams training open models should expect more questions about held-out loss curves in safety filings and investor decks, echoing LLM360-style checkpoint transparency debates https://explainx.ai/blog/k2-horizon-open-model-fleet-ifm-november-2026 . Publishing loss on a standard text slice—even if it differs from OpenAI's code snapshot—may become table stakes. Safety engineers should pair this essay with CoT monitorability findings https://explainx.ai/blog/openai-cot-monitorability-model-lifecycle-november-2026 and GPT-6.2 transparency benchmarks https://explainx.ai/blog/gpt-6-2-astra-sol-aa-personalitybench-transparency-uncertainty-november-2026 : communication style scores do not replace compression-linked latent skill , but they show how hard "readable reasoning" is to keep as models compress better. Limits of what we know today - Only OpenAI published a curve on November 22, 2026 ; rivals' positions on the same axis are unknown. - Thresholds are retrospective storytelling inside one lab, not a physical law. - Regulators have not adopted perplexity pacing; watch UK AISI-style no-CoT reviews https://explainx.ai/blog/openai-cot-monitorability-model-lifecycle-november-2026 for early echoes. - The essay does not specify numeric stop lines or enforcement; those would require multilateral negotiation. How to read the essay alongside system cards If you brief executives or customers, pair Pachocki's thresholds with documents that do carry enforcement language. GPT-6.2 system cards https://deploymentsafety.openai.com/gpt-6-2-astra-sol/model-data-and-training?ref=explainx still describe sandboxed internal use, Codex traffic logging, and Preparedness Framework tiers for cyber-critical capability. None of those documents replace perplexity pacing, but they show where OpenAI already slows or gates release when monitorability or containment looks thin. For procurement teams, the useful question is whether a vendor will share held-out loss trends under NDA the way they share eval tables today. If not, ask for third-party eval cadence and incident disclosure policy instead, using OpenAI's frontier misalignment detection write-up https://explainx.ai/blog/openai-how-we-detect-frontier-model-misalignment-four-layer-monitor-hugging-face-replay-30-minutes-2026 as a comparison point. Perplexity is not yet a contract clause; the essay is a scientific proposal aimed at peers, not a product changelog. Researchers comparing open and closed stacks can still track public checkpoint loss where labs disclose it, as in fp4 and NVFP4 pretraining studies https://explainx.ai/blog/fp4-pretraining-rare-facts-nvfp4-loss-curve-2026 and MoE training handbooks https://explainx.ai/blog/hugging-face-ultimate-moe-training-handbook-megatron-800k-gpu-hours-guide-2026 . The gap is standardization: Pachocki wants one ruler; today each lab still brings its own yardstick. Related reading Official sources: Perplexity and Intelligence https://openai.com/index/perplexity-and-intelligence?ref=explainx OpenAI, November 22, 2026 ; Managing frontier AI risk https://openai.com/index/managing-frontier-ai-risk/?ref=explainx OpenAI, November 18, 2026 ; What affects chain-of-thought monitorability https://openai.com/safety-alignment/cot-monitorability-model-lifecycle/?ref=explainx OpenAI, November 10, 2026 .