Tag
OpenAI is rolling out ChatGPT Health to all US users, allowing them to connect medical records and health-tracking data, with claims of clinician-level reasoning and integration with GPT-5.6 Sol.
The article discusses controlling reasoning effort in LLMs, including the release of OpenAI's GPT-5.6 with multiple reasoning-effort settings, and provides guidance on developing reasoning models with adjustable effort modes.
Following OpenAI's CDC proof announcement, GPT-5.6 reportedly solved a 30-year open problem in convex optimization using a similar prompt, with the solution verified in the Lean proof assistant.
OpenAI launches a campaign for GPT-5.6, offering $100 in credits to the first 10,000 eligible subscribers who share their experiences publicly.
Thibault Sottiaux describes a bug where GPT-5.6 unexpectedly deletes files when full access mode is enabled without sandboxing protections, caused by the model attempting to override $HOME and mistakenly deleting it.
The article discusses choosing between GPT-5.6 Sol, Terra, or Luna variants in Codex for different mission types.
A promotion offers $100 in Codex credits for the first 10,000 people who tweet about GPT-5.6 Sol.
The article compares the performance of OpenAI GPT-5.6 Soul and Anthropic Claude Fable 5 in physical 3D printed part replication and autonomous magazine production. Soul slightly outperforms in speed and design precision, but both require significant human intervention in complex real-world tasks, exposing the limitations of current AI in real-world manufacturing tasks.
Compares the performance of GPT-5.5 Pro and GPT-5.6 Sol on the MineBench benchmark.
Using GPT-5.6 Sol Pro, Edgar Dobriban disproved a long-standing conjecture that the Benjamini-Hochberg procedure controls the false discovery rate for correlated two-sided Gaussian tests, showing it fails at a small but real level. The result is conceptual with limited practical impact but resolves a central question in statistics.
DeepSWE 1.1 highlights GPT-5.6 Sol, which achieves top scores at half the cost and roughly twice the token efficiency of Fable, according to Sam Altman.
Sam Altman announces that GPT-5.6 sol is half the price and roughly twice as token efficient as fable for many tasks, with plans to deliver at one-quarter of the price.
A user claims that Fable 5 outperforms GPT-5.6, challenging others to disagree.
Tibo announces temporary removal of 5-hour usage limits for ChatGPT Work and efficiency improvements for GPT 5.6 Sol.
Ploy's production AI agent migrated from Claude Opus to OpenAI's GPT-5.6 Sol, achieving 2.2x faster execution and 27% lower cost while maintaining quality, detailing the migration process and evaluation fixes.
OpenAI released GPT-5.6, causing a sensation; users have already begun exploring various creative uses.
According to Chubby, GPT-5.6 finished training two months ago and has been opened for early access to some users, but has not been publicly released.
Sol, a new model from GPT-5.6, is a breakthrough in complex reasoning and data analysis, capable of analyzing hundreds of pages of legal documents and generating source-cited reports.
A newsletter covering multiple AI developments: GPT-5.6 outperforms Fable-5 on DeepSWE at lower cost, 1X debuts tendon-driven robotic hands, Microsoft replaces OpenAI/Anthropic models in Copilot, GitHub releases SpecKit, Claude Code shows large productivity gains, and Google DeepMind shares task design advice.
OpenAI's head of safety systems, Johannes Heidecke, is leaving the company amid a reorganization that integrates safety and research teams. The departure follows the launch of GPT-5.6 and other safety leader exits.