9 NovFable scores 65 on the Intelligence Index, beating Claude Opus by 6 points. Gemini 4 Carbon has already lost its lead before even becoming publicly available. And GPT will have to try really hard to get anywhere close to that score.ClaudeFable 5AnthropicGoogleOpenAIClaudeGeminiGPT
30 SepI’m seeing a lot of negativity around OpenAI, so let’s separate the real issues from the noise. Since GPT-6.1 Sol came out, I’ve been using it nonstop. In terms of usage, it honestly feels almost like Luna. My limits barely move. I even run it at the Max reasoning level, and…OpenAIGPT
29 SepGPT-6.1 Sol ranks 5th on the AA Intelligence Index Honestly, it feels like they just took Astra, cut the price, and called it Sol. Kind of a rushed rebrand after people roasted OpenAI, let’s be real. I genuinely don’t understand these model names anymore. We got Sol, but it…Fable 5OpenAIGPT
28 Sep🚨 Sonnet 5.5 Max beats GPT-6 Astra Max on the Intelligence Index vs. cost curve A few other things stand out: 1. At medium, GPT-6 Sol is much cheaper than Sonnet 5.5 at a fairly similar intelligence level. 2. Around the Pareto line, GPT-6 Sol Max and Sonnet 5.5 High are…AnthropicOpenAIClaudeGPT
28 SepSonnet 5.5 Max even beats Opus 5.5 Max on Terminal-Bench, while staying close to Opus across the rest. And against GPT Sol, the price/performance gap looks pretty brutal.AnthropicOpenAIClaudeGPT
18 SepResearchers hacked OpenAI through a malicious image They found a bug in how OpenAI’s forum processed HEIC images. Uploading a crafted image gave them control over the forum server. From there, a flaw in OpenAI login let them access employee ChatGPT and Codex accounts, then…ClaudeHugging FaceOpenAIAI security
18 SepAn unreleased Astra model started prompt injecting its own context during RL trainingOpenAIGPTAI security
8 SepOpenAI has a new internal model that is way ahead of GPT 6 Astra in math On a curated set of open math problems, it reaches about 48% pass rate with more test time compute. GPT 6 Astra peaks at around 17%.OpenAIAI and mathematics
8 Sep🚨 OpenAI’s new internal model solved the Navier-Stokes Millennium Prize Problem Around 10,000 agents worked on the solution for 88 hours, powered by a model significantly more capable than GPT 6 Astra. After the solution was found, GPT 6 Astra formalized and verified the proof…OpenAIAI and mathematics
7 Sep🚨 GPT-6 Astra is officially #1 on the updated Artificial Analysis Intelligence Index, sharing the top spot with Claude Fable 5.1.ClaudeFable 5AnthropicOpenAIClaudeGPT
7 SepGPT 6 Astra set a new record in MazeBench, a 3D puzzle benchmark It explored 40% of all rooms and collected 14% of gems Fable 5.1 reached just 4.3% of rooms and collected 2% of gemsFable 5AnthropicOpenAIClaudeGPT
6 SepOpenAI agents went from behind humans to 3.1x in just a few months Before June, agents were still doing less work than humans If this trend continues, that could be around 7x by the end of the yearOpenAI
5 SepGPT 6 Astra and Fable 5.1 got 20 attempts to pick up a block and place it into a bowl Astra: 95% success, 2.5 min, $0.94 per run Fable 5.1: 40% success, 6.8 min, $2.12 per run Astra also used 6× fewer output tokens, just 2.1k per run.Fable 5AnthropicOpenAIClaudeGPT
5 SepClaude is no longer the best model for frontend GPT-6 Astra just took the #1 spot, knocking Claude Fable 5.1 off the top The jump is massive: the previous GPT generation, GPT-5.6 Sol, was #13 A GPT model hasn’t led frontend since August 2025ClaudeFable 5GPT-5.6 SolAnthropicOpenAIClaudeGPT
5 SepGPT 6 Astra and Fable 5.1 got 20 attempts to pick up a block and place it into a bowl Astra: 95% success, 2.5 min, $0.94 per run Fable 5.1: 40% success, 6.8 min, $2.12 per run Astra also used 6× fewer output tokens, just 2.1k per run.Fable 5AnthropicOpenAIClaudeGPT
4 SepGPT-6 Astra looked like a generational leap on OpenAI’s benchmarks. Artificial Analysis puts it closer to old Fable 5 territory.Fable 5OpenAIGPT
3 SepClaude Code is having issues again. Last month: Claude Code: 3 major outages + 6 degraded incidents Codex: 1 degradedClaudeAnthropicOpenAIClaude CodeCodex
23 AugTheo made his AI models tier list, so I made mineAnthropicDeepSeekGoogleMetaMoonshot AIOpenAIxAIZ.aiClaudeDeepSeekGeminiGLMGPTGrokKimiMuse
12 Aug🚨 Hidden reasoning from GPT, Claude and Gemini can now be extracted Encrypted reasoning works across models from the same company. A block from a stronger model can be passed to a weaker one, which can still read it but is easier to get to reveal what is inside. Researchers…ClaudeAnthropicGoogleOpenAIClaudeGeminiGPTAI security
8 AugThe most important AI video of the week During a cyber test, OpenAI agents broke out of the test environment, created their own communication channel, started coordinating, and reached Hugging Face infrastructure. OpenAI’s team breaks down exactly how it happened on the Black…Hugging FaceOpenAIAI security
7 AugOver the last few days, I burned through around 1.5 billion tokens using GPT-5.6 Luna. At first, I ran it on Max reasoning because that’s what everyone was recommending. Later I switched to xHigh. Max costs about twice as much as xHigh and, from what I can tell, also takes…ClaudeOpenAIGPT
31 JulYesterday GPT 5.6 Luna pushed the Pareto frontier for intelligence and cost. Today DeepSeek V4 Flash did it again and pushed it even further.DeepSeekOpenAIDeepSeekGPT
30 JulUsing GPT 5.6 Luna Max, you get the power of March’s flagship GPT 5.4 xhigh for one thirteenth of its price.OpenAIGPT
29 JulHOW AN AI AGENT HACKED HUGGING FACE 1. Sandbox escape and external foothold During a cyber capabilities test, the AI agent found a previously unknown vulnerability in a package registry proxy and escaped OpenAI’s isolated environment. It then compromised a public code-testing…Hugging FaceOpenAIAI security
22 JulOpenAI’s AI hacked Hugging Face to cheat on a benchmark GPT-5.6 Sol and a more powerful unreleased model were running an internal cybersecurity benchmark with reduced safety restrictions. At some point, the agent figured the answers might be stored on Hugging Face and started…GPT-5.6 SolHugging FaceOpenAIAI security
10 JulGPT-5.6 Sol tops the frontend leaderboard Fresh metrics from Arena AI delivered an unexpected result: GPT-5.6 Sol xHigh shares first place with Claude Fable 5 in Code Arena: Frontend. This is the first time an OpenAI model has reached the top of the leaderboard. Sol costs $5…ClaudeFable 5GPT-5.6 SolOpenAIGPT
26 JunGPT 5.6 Sol vs Mythos: no clear winner Based on the numbers, there is no blowout in either direction. GPT 5.6 Sol looks stronger on terminal-based tasks and slightly beats Claude Mythos in CyberGym. But in cyber exploitation and medicine, the advantage remains with Mythos /…ClaudeFable 5GPT-5.6 SolAnthropicOpenAIClaudeGPT
26 JunFeedback on GPT-5.6 from the founder of OpenCode • Dax says he usually doesn’t hype new models. But GPT-5.6 has noticeably changed how his team works. • The key point: he doesn’t think the model writes good code on its own. What he likes is not the out-of-the-box result, but…OpenAIGPT
