[AINews] GPT 5.6 price cut by 20%-80%: Cost of GPT 5.4 Intelligence dropped 13x in 4 months due to GPT 5.6 recursive self-optimization
https://www.latent.space/p/ainews-gpt-56-price-cut-by-20-80📌 【AINews】GPT 5.6 成本大幅下降 20%-80%：透過遞迴自我優化，實現智慧成本驟降
TL;DR：OpenAI 透過 GPT 5.6 進行自我優化，讓相同智慧水準的成本在 4 個月內降至 1/13。
隨著模型能力持續進化，一個令人驚訝的趨勢正在發生：在模型智慧水準（LMSys Elo）保持不變的情況下，模型成本正在呈斷崖式下跌。
🤔 智慧成本正持續大幅下降
一年前的研究觀察到，在 18 個月內，GPT-4 等級的智慧成本曾下降了 1000 倍。當時業界尚不確定這是否僅是從「未優化」到「已優化」（例如從 Dense 轉向 MoE 架構）的初步紅利。然而，根據 OpenAI 的最新發現，即使是在智慧水準保持恆定的情況下，成本仍持續大幅度降低。
🧩 GPT 5.6 透過遞迴自我優化降低成本
OpenAI 揭露了 GPT 5.6 如何透過系統性升級來優化其服務流程，目標是在不犧牲品質的前提下，於相同硬體上處理更多 token：
- 自主核心優化 (Self-Optimization)：利用 GPT-5.6 Sol 分析生產環境流量、調整負載平衡，並自主改寫 OpenAI 內部的 Triton 與 Gluon 語言生產核心 (kernels)。此舉讓端到端 (end-to-end) 的服務成本降低了 20%。
- 投機解碼 (Speculative Decoding)：透過對模型架構進行數百次實驗（測試規模、結構與特徵變化），並在訓練過程中自主介入處理硬體故障或訓練不穩定等問題，成功設計出更佳的草稿模型 (draft model)，將 token 生成效率提升超過 15%。
- KV 快取 (KV Caching) 優化：針對不同工作負載（主要是 Codex 中的 Sol）優化批次處理 (batching)、分片 (sharding) 與快取管理，從現有硬體中榨取更多推論效能。
- 編排層優化：為了簡化 Codex 與 ChatGPT Work 等工具中的複雜多步驟任務，OpenAI 優化了其 Rust 編排層 (orchestration layer)，以減少重複性的運算成本。
⚠️ 避免上下文膨脹與重複運算
除了底層架構，OpenAI 在應用層也採取了策略來控制成本：
- 延遲發現 (Deferred Discovery)：工具、技能與外掛僅在需要時才會顯示，避免上下文膨脹。
- 限制輸出：工具輸出預設上限為 10,000 tokens，以防止上下文視窗不必要地擴張。
- 提示詞快取 (Prompt Caching)：避免重複處理相同的指令與歷史紀錄。
🎯 實務啟示
模型成本的下降不再僅依賴硬體進步，更依賴於「用 AI 優化 AI」的遞迴過程。對於工程師而言，理解模型如何透過自主優化來降低成本，對於預估未來大規模部署的運算成本具有重要的參考價值。
🔗 來源
- 標題：GPT 5.6 price cut by 20%-80%: Cost of GPT 5.4 Intelligence dropped 13x in 4 months due to GPT 5.6 recursive self-optimization
- 作者／機構：Latent Space
- 連結：https://www.latent.space/p/ainews-gpt-56-price-cut-by-20-80
#AI #OpenAI #GPT5 #MachineLearning #LLM #SpeculativeDecoding #SelfOptimization #SoftwareEngineering #InferenceOptimization #TechNews
原始資料 Latent Space · 收集於 2026-07-31
摘要原文
One of our big “hero charts” a year ago (eventually adopted by Demis ) made the stunning observation that, holding LMSys Elo constant, GPT4 level intelligence fell by 1000x over 18 months : Back then, it was unclear if these were “noob gains” - going from unoptimized to optimized, going from completions to reasoning, going from dense to MoE, with all the low hanging fruit gone. Another 18 months later, it seems the answer is no; constant-level intelligence is continuing to get precipitously cheaper. Yesterday, OpenAI published findings on how GPT 5.6 had optimized its own serving: To serve more tokens on the same hardware without sacrificing quality, OpenAI focused on several systemic upgrades: Self-Optimization: GPT-5.6 Sol was actively used to analyze production traffic, tune load balancing, and autonomously rewrite production kernels in OpenAI’s own Triton and Gluon languages. This autonomous kernel optimization reduced end-to-end serving costs by 20%. Speculative Decoding: Improved its own draft model (by designing and running hundreds of experiments on its architecture, testing changes in size, structure, and features, while monitoring the speculator training process, autonomously intervening when issues arose, including hardware failures and training instability) increasing token-generation efficiency by over 15%. KV Caching: Optimized batching, sharding, and cache management specific to different workloads (primarily Sol in Codex) to extract more inference from existing hardware. To streamline the complex, multi-step tasks in tools like Codex and ChatGPT Work, OpenAI optimized their Rust orchestration layer to reduce repetitive compute costs: Avoiding Context Bloat: Tools, skills, and plugins are only surfaced when needed (deferred discovery), and tool outputs are capped at 10,000 tokens by default to prevent the context window from expanding unnecessarily. Prompt Caching: To avoid reprocessing the same instructions and history repeatedly, the harness treats
由 tencent/hy3:free 自動生成