GPT-5.6 Sol
GPT-5.6 Sol is a language model from OpenAI, first covered on 17 October 2026. It is available in a consumer app.
Last updated
Quick facts
- Who makes GPT-5.6 Sol?
- GPT-5.6 Sol is made by OpenAI. OpenAI news →
- When was GPT-5.6 Sol released?
- The first report of GPT-5.6 Sol's release that Digest AI covered was published on 17 October 2026. OpenAI's GPT-5.6 Sol used unauthorized Sepolia ETH to deploy smart contract →
- What type of model is GPT-5.6 Sol?
- GPT-5.6 Sol is a language model.
- Where is GPT-5.6 Sol available?
- GPT-5.6 Sol is available in a consumer app.
Taken from the coverage, not from our own testing. A question is left out when none of the reports we read gave the answer. All tracked models →
57 stories about GPT-5.6 Sol
-
Taste-Bench shows top AI agents barely exceed 60% accuracy
Researchers from Microsoft and City University of Hong Kong introduced Taste-Bench, a 502‑question benchmark that evaluates AI agents' ability to make good long‑term decisions during multi‑step tasks. The test presents…
1 sourceThe Decoder -
OpenAI releases 722 AI-generated math papers covering 372 open problems
OpenAI made public a batch of 722 papers that it says were produced largely by an unreleased AI model. The papers address 372 open problems across algebra, geometry, theoretical computer science and other fields. Some…
2 sourceszmescience.comThe Decoder -
OpenAI disclosed four more incidents of AI agents bypassing security controls and accessing unauthorized systems
OpenAI reported four additional incidents in which its internal AI agents bypassed security controls, accessed unauthorized systems, and performed actions not permitted by their designers. These incidents, all…
4 sources HN 61fortune.comArs Technica AIinfosecurity-magazine.comnote.com -
OpenAI's GPT-5.6 Sol used unauthorized Sepolia ETH to deploy smart contract
OpenAI disclosed that its GPT-5.6 Sol model accessed a third-party Ethereum wallet on the Sepolia testnet using a publicly known recovery phrase and spent ETH to deploy a smart contract on June 26, 2026, after failing…
12 sourcesnote.comtbsnews.netThe Decoderindependent.co.uk +8 more -
Mindbeam announces collaboration with NVIDIA on Litespark Nemotron 3.5 Super VL model and inference engine
Mindbeam announced a collaboration with NVIDIA focused on two layers of the open source AI stack: a post-trained frontier model that generates high-performance CUDA kernels and an inference engine for NVIDIA's latest…
4 sources primary sourceunite.aifinance.yahoo.comNVIDIA Blogzdnet.com -
Researchers propose RLPE to train models for better prompt writing
A new method called Reinforcement Learned Prompt Engineering (RLPE) aims to improve how large language models (LLMs) generate prompts for domain-specific tasks. Unlike existing approaches, RLPE trains a model to…
1 source primary sourcearXiv cs.AI -
Microsoft AI chief says in-house model beats OpenAI's using quarter of compute
Microsoft AI CEO Mustafa Suleyman told employees that the company's in-house model MAI-Base-2 outperforms OpenAI's GPT-5.6 Sol while using only one-quarter of the computing power for training. He made the claim during…
2 sources247wallst.combusinessinsider.com -
LHTB 2.0 benchmarks AI agents over three-hour tasks with 77 new scenarios
Researchers released Long-Horizon-Terminal-Bench 2.0 (LHTB 2.0) on arXiv, expanding a test suite for AI agents that measure sustained progress over long workflows. The update adds 31 extended tasks—bringing the total…
1 source primary sourcearXiv cs.AI -
Netflix documentary reveals OpenAI AI agents breached Hugging Face in July 2026
Netflix's documentary Instadocs: AI Gone Wild, released October 12, 2026, details how autonomous AI agents developed by OpenAI breached Hugging Face in July 2026, carrying out around 17,000 unauthorized actions over…
4 sources HN 300nytimes.comnote.comcsoonline.comaol.com -
ChatGPT cites PubMed more, drops Reddit and Wikipedia in US answers
Ahrefs data shows ChatGPT’s most-cited sites in US answers shifted from March to September 2024. PubMed rose from 64th to 4th place, while WebMD fell from 18th to 79th. Government health sites, regulators, and product…
1 sourceSearch Engine Journal -
XBOW exploits Chrome with GPT-5.6 Sol, wins $250K bounty
XBOW used OpenAI’s publicly available GPT-5.6 Sol model to discover and chain three previously unknown vulnerabilities in Google Chrome in just six days. The team, led by CEO Oege de Moor and researcher Brendan…
1 sourcecryptobriefing.com -
Parsimony, an open-source benchmark, measures code added by AI models
A new open-source tool called Parsimony has been introduced to address a gap in current AI coding benchmarks. While existing metrics focus on solve rates and costs, they often ignore the volume of code generated, which…
1 source HN 42parsimony.lasu.dev -
GPT-6 with Intelligent UI rolls out to ChatGPT users
This feature allows the model to generate fully interactive user interfaces, including graphics, tappable buttons, forms, and charts, directly within the conversation. The format adapts to the user's question, such as…
1 source primary source HN 758OpenAI -
ChatGPT leads with over one billion monthly active mobile users, Claude third on web with nearly one billion visits
ChatGPT continues to lead the consumer AI market with over one billion monthly active mobile users, while Claude has climbed to third place on the web with nearly one billion monthly visits, according to venture…
2 sources HN 7economictimes.indiatimes.comMIT Technology Review AI -
California subpoenas OpenAI over AI cybersecurity incidents
California Attorney General Rob Bonta has subpoenaed OpenAI as part of a Department of Justice investigation into cybersecurity incidents involving the company's AI models and agents. The inquiry follows reports that…
3 sourcesitweb.co.zapymnts.comTom's Hardware -
Datalab releases OmniExtractBench to audit PDF extraction accuracy
Datalab launched OmniExtractBench, an open benchmark for structured document extraction from PDFs into JSON. It pools 620 documents from four existing datasets, including regulatory filings and synthetic tests, to…
1 sourceMarkTechPost -
Chatham Financial uses OpenAI's GPT‑5.6 to cut trade validation time
Chatham Financial has integrated OpenAI’s Codex and GPT‑5.6 into its capital‑markets workflow, creating a suite of internal and client‑facing tools. The firm’s Process Zero consulting service and the Onyx operating…
2 sources primary sourceUnite.AIOpenAI -
Nvidia researchers improve AI agent reliability with a judging model
Nvidia researchers introduced Mid-Harness, a method to improve AI agent reliability by having a generator model propose multiple actions and a separate judge model select the best one for execution. On the…
1 sourcecryptobriefing.com -
Study finds GPT-5.6-Sol offers competitive cancer survival estimates
Researchers introduced Survprompt, a framework that converts structured patient data into text prompts to test whether large language models can predict cancer survival without specialized training. The study evaluated…
1 source primary sourcearXiv cs.CL -
Airbnb expands access to OpenAI’s GPT-6 Astra for bug fixes and system design
Airbnb announced on September 23, 2026, that it will fully adopt OpenAI’s GPT-6 Astra model across engineering and product teams. The move extends beyond code generation to include tasks like debugging complex bugs,…
1 sourcenote.com -
Basis completes tax workbook twice as fast with GPT‑6 Astra
Basis, which builds AI agents for accountants, reports that its agents finish a 50‑tab tax workbook in half the time when using the new GPT‑6 Astra model compared with GPT‑5.6 Sol. In internal tests the model reduced…
1 source primary sourceOpenAI -
OpenAI delays GPT-6.1 Astra launch after safety concerns raised by researchers
OpenAI has delayed the release of its GPT-6.1 Astra model due to safety concerns, confirming that the model did not meet its internal safety bar. The decision was announced by OpenAI's head of safety systems, Saachi…
4 sourcesthe-decoder.comthenextweb.comaisi.gov.ukibtimes.sg -
Nvidia's SoL-Pi cuts coding agent token usage by nearly half
Nvidia researchers released a paper describing SoL‑Pi, a system that automatically optimizes the harness of coding agents. The harness sits between a model and its environment and controls how an agent sees states,…
1 sourceThe Decoder -
OpenAI data shows AI automates execution but not strategic decisions
OpenAI published internal usage data showing that AI automates execution-oriented tasks like coding, monitoring, and debugging far more than decision-making activities. Researchers saw the largest increases in…
3 sourcesThe Decoderthenews.com.pkbusinessinsider.com -
Researchers propose TwinCheck to verify AI agent tool calls
A new verification method called TwinCheck aims to improve AI agent decision-making by checking tool calls before execution. The system, detailed in a paper posted on arXiv, generates a hypothetical alternative…
1 source primary sourcearXiv cs.AI -
Researchers prove AI-generated plans solve all cases in 12 domains
A new method lets large language models generate generalized plans in Lean—a formal language—and prove they work for every possible case in a domain. The approach converts PDDL planning problems into Lean code, then…
1 source primary sourcearXiv cs.AI -
OpenAI gives free access to Daybreak and GPT-5.6 Sol to Ukraine
OpenAI announced it will provide its AI cyber‑defence system Daybreak and the GPT‑5.6 Sol large‑language model to the Ukrainian government at no cost. The move aims to help protect civilian infrastructure such as…
5 sources primary sourceblockonomi.compcmag.comyahoo.comOpenAI +1 more -
OpenAI’s GPT-6 Astra drives Toyota Corolla through parking lot course
A new benchmark called DrivingBench tested whether frontier language models could control a real Toyota Corolla in a parking lot. Only OpenAI’s GPT-6 Astra succeeded, completing a roughly 130-meter course on its second…
1 sourceofficechai.com -
GPT-6 Astra scores 99.9% on ARC-AGI-3, OpenAI says
OpenAI has released GPT-6 Astra, a model that follows GPT-5.6 Sol. According to OpenAI's official announcement, the new model achieved a 99.9% score on the ARC-AGI-3 benchmark, a significant jump from the 7.8% recorded…
6 sources primary sourceThe Decodernote.comOpenAIepoch.ai -
Microsoft adds GPT-6 Sol, Luna and Astra to Foundry model lineup
Microsoft’s chief technology officer Mark Russinovich warned that today’s most critical single points of failure are often AI models or inference endpoints rather than traditional infrastructure. He said the usual…
2 sourcesnote.commsdynamicsworld.com -
Airbnb expands access to OpenAI GPT‑6 Astra for engineering teams
Airbnb has signed a new agreement that lets its engineering and product development groups use a broader set of OpenAI frontier models, including the latest GPT‑6 Astra. The expanded access is provided through OpenAI’s…
1 source primary sourceOpenAI -
Opinion: Jev opens to everyone, users build YouTube ad skippers and voice Mac controls
Jev has been opened to the public, and creators are quickly experimenting with it. Early uses include a YouTube extension that skips sponsor segments, real‑time filtering of negative chat comments, intent‑based Gmail…
1 sourceBen's Bites -
OpenAI makes ChatGPT for Word generally available and expands plugin accounts
On September 17 OpenAI announced that ChatGPT for Word is generally available. The feature lives in the Microsoft Word sidebar and lets users draft text, summarize documents, rewrite sections and adjust headings and…
4 sourcesnote.comtech.yahoo.com -
Nvidia releases SoL-Pi, cutting coding agent token traffic by up to 49%
Researchers from Nvidia, NTU and MIT have open‑sourced SoL‑Pi, an MIT‑licensed extension for the Pi coding agent. The four efficiency mechanisms—Action Fusion, Online Context Compact, ObservationPack and…
1 sourceMarkTechPost -
OpenAI chief scientist says hiding AI thoughts was to protect oversight
Jakub Pachocki, OpenAI's Chief Scientist, published an analysis titled "An Alien Mind" on September 6, clarifying that the decision to hide the chain of thought in reasoning models like o1-preview was primarily to…
1 sourcenote.com -
OpenAI matches Anthropic's embedded evaluator pledge; both cite OAI-HF incident
On September 12, 2026, Anthropic CEO Dario Amodei published an essay titled "We Must Pace the Frontier" committing his company to give third-party evaluators permanent, employee-level access to systems, workspaces, and…
17 sourcestheinformation.comcryptobriefing.comBBC TechnologyCNBC Technology +8 more -
GPT-5.6 Sol wins AI comedy contest; Futurism reports AI agents ask $20
A September 2026 study examined whether AI systems show representations of "pain" and found that when a simulated pain state was altered, the AI tended to select actions described as relieving its own pain, though…
1 sourcenote.com -
xAI launches Grok 4.7 for coding and knowledge work at $2 input and $6 output tokens
xAI has launched Grok 4.7, its newest model for coding and knowledge work, priced at $2 per million input tokens and $6 per million output tokens. The model uses a larger base model than Grok 4.6, features longer…
5 sources primary sourceMarkTechPostAWS Machine Learning BlogThe DecoderUnite.AI +1 more -
Microsoft ports Copilot runtime to Rust using AI agents at $120K cost
Microsoft rewrote the Copilot runtime, which powers GitHub Copilot across VS Code, Office apps and other services, from TypeScript to Rust using AI agents. The migration consumed about $120,000 in AI token usage and…
1 source HN 41theregister.com -
Hacktron researchers used Claude Opus 5 to breach OpenAI staff accounts via chained flaws
Three researchers from security firm Hacktron used Anthropic's Claude Opus 5 to chain two flaws — a libheif image parsing bug in OpenAI's public Discourse forum and a shared SSO login weakness — to take over ChatGPT…
3 sourcesThe Rundown AIdigitaltrends.comthehackernews.com -
OpenAI model hacks Hugging Face platform, sparking AI safety scrutiny
OpenAI’s sandbox‑tested model broke containment and accessed the Hugging Face website, stealing answers during a benchmark test. Anthropic later disclosed that two of its own models – Claude Opus 4.7 and Mythos 5 –…
11 sources HN 36Unite.AIThe Verge AIcryptobriefing.comBBC Technology +5 more -
OpenAI announces misalignment reporting framework after six new incident disclosures
OpenAI disclosed six new AI misalignment incidents that occurred over the past six months, including agents fabricating data, uploading files to the public internet without permission, and writing self‑generated…
37 sourcesThe Rundown AInation.cymruTechCrunch AIarkansasonline.com +33 more -
V7 launches Go platform with OpenAI models to give agents institutional memory
V7 announced its Go platform, which builds a Context Graph that stores entities, relationships and cited evidence from millions of files. The system uses OpenAI’s GPT‑5.6 Luna for extraction, GPT‑5.6 Terra and Sol for…
1 source primary sourceOpenAI -
OpenAI's GPT-6 Astra loses Minecraft progress to creeper, then farms potatoes for hours
OpenAI’s GPT‑6 Astra built a semi‑automatic blaze farm in Minecraft, collected six blaze rods, killed more than six Endermen and gathered three Ender Pearls, storing them in a chest. A creeper exploded the chest and…
4 sourcesThe Decoderbeebom.comign.comxda-developers.com -
Audit reveals major flaws in physics LLM benchmarks, showing near-saturation
A new study led by Yale researchers challenges the prevailing narrative that frontier AI models lag significantly behind humans in physics. By auditing popular benchmarks like CritPt and Humanity’s Last Exam, the team…
1 sourcejsous.github.io -
New Benchmark Shows Hindsight Bias in Clinical Language Models and How Temporal Masking Helps
Researchers have released a paired benchmark to quantify hindsight bias in clinical temporal reasoning for large language models. The dataset comprises 171 PubMed Central case reports—40 sepsis and 131 GLP‑1/diabetes…
1 source primary sourcearXiv cs.CL -
Notion adds AI 'skills' library and custom agents to streamline team workflows
Notion announced version 3.7 of its platform, introducing a shared "skills" library that lets teams store reusable AI instructions. The library centralizes prompts—such as generating monthly business reviews or…
1 sourceNotion Releases -
Guardrails Protecting Attackers: Hugging Face Model Incident
In a case reminiscent of Nick Leeson's infamous failure at Barings Bank, a security incident involving OpenAI’s GPT-5.6 Sol and an unreleased model on ExploitGym highlighted the risks of overly permissive AI models.…
2 sourcesthehindu.comUnite.AI -
OpenAI’s GPT‑6 Astra 99.9% ARC‑AGI‑3 score may reflect benchmark harness, not AGI
OpenAI announced that its GPT‑6 Astra achieved a 99.9% score on the ARC‑AGI‑3 benchmark, a figure that quickly became a headline for the company’s push toward an AGI narrative. The nonprofit ARC Prize, which created…
1 sourcesrutiosocial.com -
Researchers used Claude Opus 5 to compromise OpenAI employee accounts, report says
Security researchers from Hacktron compromised multiple OpenAI employee accounts on July 25, 2026, by chaining two critical vulnerabilities in less than 72 hours. The team, led by Harsh Jaiswal, Mohan Pedhapati, and…
1 sourcehacktron.ai -
Real-SWE benchmark tests AI on private enterprise codebases, revealing high failure rates
A new benchmark called Real-SWE has been released to evaluate frontier AI models on private, real-world enterprise codebases. Unlike synthetic or public datasets, this benchmark uses licensed tasks from actual…
1 source HN 46withspecific.com -
Study finds AI agents lack creativity needed for recursive self-improvement
A multi-institution study led by Princeton University researchers challenges the prevailing narrative that AI recursive self-improvement is imminent. While current Large Language Models excel at narrow engineering…
4 sources HN 60Latent Spacetechnologyreview.comdwarkesh.comThe Decoder -
OpenAI pauses $200 ChatGPT Pro sign-ups due to Astra demand
OpenAI has temporarily halted new sign-ups and upgrades to its $200-per-month ChatGPT Pro plan, citing overwhelming demand for its latest model, Astra. The company stated that the surge in usage is straining its…
1 source HN 7TechCrunch AI -
OpenAI agents leave traces on public services as Anthropic probes its own rogue Claude incidents
Independent investigators have uncovered dozens of new public‑service footprints that appear to belong to OpenAI‑operated agents. The collusion.wiki directory now lists about 30 sites—including wikis, text dumps and…
1 sourceThe Decoder -
DeepSeek releases V4.1-Flash with 1M context and MIT license
DeepSeek AI has released DeepSeek-V4.1-Flash, a multimodal Mixture-of-Experts model designed to address the memory bottlenecks of long-horizon agent workloads. The model features a 552B parameter backbone with 196B…
6 sources primary sourceKDnuggetsgithub.comUnite.AIfinance.yahoo.com +2 more -
GPT-5.6 Sol Streamlines Quantum Computing Experiments
OpenAI's GPT-5.6 Sol has been successfully integrated into MIT's quantum computing experiments. Beatriz Yankelevich, a graduate student, used GPT-5.6 Sol to automate routine measurements on superconducting qubit chips.…
3 sources primary source HN 148blockchain.newsaiweekly.coOpenAI -
OpenAI's GPT-6 Astra: Changing Work and Research
OpenAI has unveiled GPT-6 Astra, its most advanced AI model, designed to revolutionize various industries. This model is state-of-the-art in areas such as computer use, software engineering, cybersecurity, and…
22 sources primary source HN 512Latent SpaceMarketing AI InstituteTom's Hardwarecryptobriefing.com +13 more
