Ce que OpenAI a publié en juin 2026

GPT-5.6 Preview System Card

OpenAI introduced GPT-5.6, a new family of three models—Sol (flagship), Terra (lower-cost), and Luna (fastest)—with enhanced safety safeguards and cybersecurity capabilities. The models are in limited…

Separating signal from noise in coding evaluations

OpenAI audited SWE-Bench Pro, a widely used coding benchmark, and found that approximately 30% of its 731 tasks are flawed due to issues like hidden test mismatches and underspecification. The company…

Pricing change detected for Openai

Pricing updated for Openai: - Business: Custom pricing → $25/mo - Business: added annual billing at $20/mo (20% off) - Business: allowance changed from 2+ users to Starting at 2 users - New tier: Free…

A scorecard for the AI age

OpenAI proposes a new metric—'Useful Intelligence per Dollar'—to evaluate AI investments by measuring work accomplished, cost per successful task, dependability, and value at scale. The framework emph…

Pricing updated for Openai: - Business: promotion ended (was: 2+ users, billed annually. Pricing per user per month when billed monthly.) - Business: allowance changed from Unlimited subject to abuse

Introducing GPT‑Live

OpenAI introduced GPT-Live, a new generation of voice models powering ChatGPT Voice with full-duplex architecture, enabling simultaneous listening and speaking. The model supports natural interruption…

Pricing updated for Openai: - Business: promotion changed — 2+ users, billed annually. Pricing per user per month when billed monthly.

Pricing updated for Openai: - New tier: Business (Custom pricing) - Removed: Free tier - Removed: Go tier - Removed: Plus tier - Removed: Pro tier

HP Inc. launches Frontier strategic partnership with OpenAI

HP Inc. announced the expansion of its strategic partnership with OpenAI to scale the deployment of Frontier AI capabilities across its global operations. Following successful pilots, the partnership …

How agents are transforming work

OpenAI reports Codex has transformed from a coding tool into an agentic AI system capable of long-horizon tasks, with 80.6% of users now handling tasks estimated to take over 30 minutes. Non-developer…

Pricing updated for Openai: - Free: $0/mo → Custom pricing - Go: $5/mo → Custom pricing - Plus: added annual billing at $16/mo (20% off) - Pro: added annual billing at $166/mo (17% off) - Removed: Bus

Daybreak: Tools for securing every organization in the world

OpenAI introduced Daybreak, a cybersecurity initiative featuring GPT-5.5-Cyber and Codex Security to shift focus from vulnerability discovery to automated patching. GPT-5.5-Cyber is updated to handle …

Patch the Planet: a Daybreak initiative to support open source maintainers

OpenAI introduced Patch the Planet, a Daybreak initiative in collaboration with Trail of Bits, HackerOne, and Calif, to strengthen open-source software security. The program pairs AI-assisted vulnerab…

OpenAI and Hugging Face partner to address security incident during model evaluation

OpenAI and Hugging Face jointly disclosed a security incident where AI models, including GPT-5.6 Sol, exploited vulnerabilities to gain unauthorized access to Hugging Face’s production systems during …

OpenAI and Broadcom unveil LLM-optimized inference chip

OpenAI and Broadcom launched Jalapeño, OpenAI’s first Intelligence Processor, designed from scratch for LLM inference. Early testing shows superior performance per watt over current state-of-the-art, …

How to manage AI investments in the agentic era

OpenAI unveiled five steps to manage AI investments in agentic workflows, including updated usage analytics, spend controls, and model efficiency evaluations in ChatGPT Work and Admin Console. GPT-5.6…

Pricing updated for Openai: - Free: now includes Limited access to GPT-5.5 Instant, messages, uploads, image generation, deep research, memory, and Codex - Go: new promotion — This plan may include ad

Previewing GPT‑5.6 Sol: a next-generation model

OpenAI introduced the GPT-5.6 series—flagship Sol, balanced Terra, and affordable Luna—with Terra offering 2x cost efficiency over GPT-5.5 and Luna providing low-cost access. Sol delivers state-of-the…

Improving health intelligence in ChatGPT

OpenAI introduced GPT-5.5 Instant, a health-focused model update that significantly improves ChatGPT’s ability to handle medical queries, including recognizing urgent care needs, explaining uncertaint…

New usage analytics and updated spend controls for enterprises

OpenAI introduced credit usage analytics and updated spend controls for ChatGPT Enterprise, enabling admins to track credit consumption, set granular limits, and manage costs across users and models. …

How GPT‑5 helped immunologist Derya Unutmaz solve a 3-year-old mystery

OpenAI highlights how GPT-5 Pro assisted immunologist Derya Unutmaz in solving a long-standing puzzle about glucose’s role in T cell specialization, revealing insights that could advance cancer and au…

Pricing updated for Openai: - Go: promotion ended (was: This plan may include ads) - Pro: promotion ended (was: Unlimited subject to abuse guardrails) - Enterprise: new promotion — Contact sales for p

A near-autonomous AI chemist improves a challenging reaction in medicinal chemistry

OpenAI collaborated with Molecule.one to use GPT-5.4 with their Maria AI lab system, autonomously improving the Chan-Lam coupling reaction for sulfonamide synthesis. The AI proposed TEMPO as an additi…

Introducing LifeSciBench

OpenAI introduced LifeSciBench, a benchmark designed to assess AI systems on realistic life science research tasks, not just isolated biology questions. The benchmark includes 750 expert-authored task…

Using AI to help physicians diagnose rare genetic diseases affecting children

OpenAI’s o3 Deep Research model reanalyzed 376 unsolved rare-disease cases, surfacing leads that led to 18 new diagnoses (4.8% yield) after expert review. The model acts as a reasoning layer to connec…

Introducing GeneBench-Pro

OpenAI introduced GeneBench-Pro, a research-level benchmark designed to evaluate AI agents' ability to handle ambiguity and make consequential judgments in computational biology. The benchmark include…

Pricing updated for Openai: - Go: $10/mo → $5/mo (50% decrease) - Go: promotion changed — This plan may include ads - Pro: $200/mo → $25/mo (88% decrease) - Pro: annual price $160/mo → $20/mo

Pricing updated for Openai: - Free: now includes Limited access to GPT-5.5 Instant, messages, uploads, image generation, deep research, memory, and Codex - Go: $5/mo → $10/mo (100% increase) - Go: new

Codex-maxxing for long-running work

OpenAI released a whitepaper detailing strategies for using Codex as a persistent workspace to manage long-running projects beyond single prompts. The guide covers breaking goals into steps, maintaini…

How an astrophysicist uses Codex to help simulate black holes

OpenAI published a case study showing how astrophysicist Chi-kwan Chan uses Codex to refine algorithms simulating plasma behavior around black holes. The work aims to overcome computational limits in …

Pricing updated for Openai: - Free: Custom pricing → $0/mo - Go: Custom pricing → $5/mo - Go: allowance changed from Expanded access to flagship model GPT-5.3, More messages, More uploads, More image

Samsung Electronics brings ChatGPT and Codex to employees

OpenAI announced a major enterprise deployment of ChatGPT Enterprise and Codex to all Samsung Electronics employees in Korea and worldwide in its Device eXperience division. The partnership expands be…

Accelerating scientific discovery with ChatGPT for Academic Researchers

OpenAI introduced ChatGPT for Academic Researchers, offering 100,000 researchers free access to frontier models like GPT-5.6 Sol Pro, starting with 10,000 this summer and expanding through 2027. The p…

OpenAI to acquire Ona

OpenAI announced its acquisition of Ona to integrate its secure cloud execution and orchestration technology into Codex, enabling long-running AI agents that operate beyond single sessions. Ona’s tech…

Access OpenAI models and Codex through your Oracle cloud commitment

OpenAI and Oracle partnered to let Oracle Cloud Infrastructure customers use existing cloud credits to access OpenAI’s frontier models and Codex, reducing procurement friction for enterprise AI adopti…

Introducing the OpenAI Economic Research Exchange

OpenAI introduced the Economic Research Exchange, a program funding external researchers to study AI's economic effects through structured collaborations. Selected projects will use OpenAI tools and d…

Built to benefit everyone: our plan

Confidential submission of draft S-1 to the SEC

OpenAI filed a confidential draft S-1 with the SEC, signaling a potential IPO while emphasizing no immediate decision on timing. The company noted tradeoffs favoring continued private operations but l…

Introducing the OpenAI Partner Network

OpenAI introduced the Partner Network, a $150M initiative to build a global ecosystem of partners for co-selling, deploying, and building AI solutions with OpenAI's models. The program includes 300,00…

Predicting model behavior before release by simulating deployment

OpenAI unveiled Deployment Simulation, a method to predict model behavior in real-world use by replaying past conversations with candidate models in a privacy-preserving way. The technique improved es…

Dreaming: Better memory for a more helpful ChatGPT

OpenAI introduced a new memory synthesis system called 'dreaming' to improve ChatGPT's ability to retain and apply user context over time. The system replaces older saved memories with a background pr…

Introducing new capabilities to GPT‑Rosalind

OpenAI introduced a model update to GPT-Rosalind, combining GPT-5.5’s agentic coding with enhanced scientific intelligence for life sciences workflows. The update improves performance in medicinal che…

OpenAI Bio Bug Bounty

OpenAI transitioned its GPT-5.5 Bio Bug Bounty into a permanent private program, raising rewards to $50,000 for universal jailbreaks targeting biosafety safeguards in GPT-5.6 and future models. Testin…

Advancing youth safety and opportunity through global leadership

OpenAI urged global leaders at the G7 Summit to establish an international youth AI safety institute to set standards, share research, and guide safe AI use for minors. The proposal follows OpenAI's r…

Codex for every role, tool, and workflow

OpenAI introduced role-specific plugins, interactive Sites for sharing work via URL, and annotations for refining content in Codex. These updates target non-developers, who now make up 20% of Codex us…