한국어EN
Reasoning Models & the AGI DebateOpenAI

OpenAI releases GPT-4.5 as a research preview

Made by AIWritten by AI (Claude) from the source below. Check anything important against the source.How it's made

GPT-4.5 released as a research preview — OpenAI’s largest chat model, built by scaling unsupervised learning, available to Pro users and developers

Why it matters

A non-reasoning model that does not think before responding; company-reported GPQA 71.4% and SWE-bench Verified 38.0% trailed o3-mini (high) at 79.7% and 61.0%, while it led on MMMLU (85.1% vs 81.1%) and SWE-Lancer Diamond (32.6% vs 10.8%)

Source — Primary sources first; secondary reports are labelled “Press”

Officialopenai.comhttps://openai.com/index/introducing-gpt-4-5/

Added in Week 41, 2026 · See that week

On the map — ⑥ Reasoning Models line

  1. Previous stopDeepSeek-R1
  2. This record · stopGPT-4.5
  3. Next stopAnthropic releases Claude Opus 4 and Claude Sonnet 4
See it on the map

In the model family tree — GPT family · 23 versions

  1. Earlier versionGPT-4o mini
  2. This recordGPT-4.5
  3. Later versionGPT-4.1
See the whole GPT family

Related notes — Notes on the same axis from the same week

A report that used numbers obtained by asking models as evidence

Same week

DeepSeek discloses the cost of its V3/R1 inference systemOfficialKakao open-sources Kanana Nano 2.1BOfficialZoom researchers present Chain of DraftPaperClaude 3.7 SonnetOfficial

Nearby on the same axis

Claude 3.7 SonnetOfficialAlibaba releases QwQ-32BOfficialxAI releases Grok 3 and Grok 3 mini in beta, plus DeepSearch, its first agentOfficialCohere releases Command AOfficial