List of AI News about reasoning
| Time | Details |
|---|---|
|
2026-11-10 08:06 UTC |
Google launches Gemini 4 Carbon to AI Ultra users
According to Google via X, Gemini 4 Carbon is rolling out to Google AI Ultra subscribers on desktop, mobile, and web with deeper reasoning, long-horizon task handling, and stronger spatial understanding for 3D and UI design. |
|
2026-11-09 10:57 UTC |
Claude Fable 5.2 and Mythos 5.2 Boost Coding and Vision
According to Claude on X, Anthropic introduced Claude Fable 5.2 and Claude Mythos 5.2 with improvements in coding, reasoning, and vision for complex, multi‑field projects. |
|
2026-11-02 12:17 UTC |
MIT CSAIL: GPT-6 Astra rewrites weights to boost Qwen math
According to The Rundown AI citing MIT CSAIL’s Rikiya Takehi, GPT-6 Astra edited 0.016% of Qwen3.5-0.8B’s weights to raise a math score to 48.1% from 7.9%, a method termed Weight-Space Programming. |
|
2026-10-21 11:05 UTC |
Claude Sonnet 5.5 shows lower mergeable code at max, per test
According to DeepLearning.AI, Claude Sonnet 5.5 produced less mergeable code at its max reasoning setting than at xhigh while consuming more tokens, underscoring the need to benchmark settings before deployment. |
|
2026-10-02 12:53 UTC |
Claude Opus 5.5 Builds working watermill showcase
According to @claudeai, creators used Opus 5.5 to design a functional watermill, highlighting rapid prototyping and physics reasoning in real projects. |
|
2026-09-25 16:26 UTC |
Reasoning models Challenge 'black box' myth
According to @timnitGebru, Nate Silver’s link misreads modern reasoning models, which are more robust than next-token prediction, per Tech Policy Press. |
|
2026-09-16 02:30 UTC |
ChatGPT Co‑Creator Unveils novel AI, 3 big moves
According to TheRundownAI, a ChatGPT co-creator launched a new AI, Salesforce trained a reasoning model, OpenRouter expanded coding access, and China detailed a frontier system. |
|
2026-09-15 09:05 UTC |
Gemini 3.8 Live Debuts with Extended Thinking
According to GoogleDeepMind, Gemini 3.8 Live adds real time voice, background tasking, and Extended Thinking for complex reasoning, boosting workflow speed. |
|
2026-09-10 10:35 UTC |
GPT6 Astra Powers Financial Services ChatGPT
According to @OpenAI, ChatGPT for Financial Services blends built-in market data with GPT6 Astra reasoning for research, modeling, and client materials. |
|
2026-09-06 21:14 UTC |
OpenAI Math Breakthrough Faces Misconduct Claims
According to @timnitGebru, experts allege OpenAI’s math paper committed research misconduct, raising credit, citation, and reproducibility concerns. |
|
2026-08-06 10:35 UTC |
GPT-5.6 Sol boosts ChatGPT accuracy
According to @OpenAI, GPT-5.6 Sol powers Instant and deep reasoning for Plus and Pro, while GPT-5.6 Luna brings unlimited text chats to Free and Go. |
|
2026-08-01 20:35 UTC |
GPT 5.6 Showcases real-time build power
According to gdb, GPT 5.6 helped a first-time user build a fynbos webpage with streamed reasoning and polished HTML, highlighting accessible web dev. |
|
2026-07-29 16:41 UTC |
GPT5.6 Sol Tops ARC AGI-3 with 2 Tweaks
According to Sam Altman, GPT‑5.6 Sol leads ARC‑AGI‑3 after two setting changes enabling multi-window reasoning and canonical compaction, per OpenAI. |
|
2026-07-29 15:55 UTC |
GPT5.6 Sol Tops ARC-AGI-3 With Harness Tweaks
According to emollick, GPT-5.6 Sol became SoTA on ARC-AGI-3 via two harness settings enabling multi-window reasoning and compaction, per OpenAI analysis. |
|
2026-07-24 17:36 UTC |
Claude Updates trim chain of thought
According to @emollick, Anthropic reduced visible reasoning traces in Claude, impacting error diagnosis and interpretability, as noted by users on X. |
|
2026-07-18 21:52 UTC |
Kimi K3 Reveals 32 page CoT Loops
According to @emollick, Kimi K3 produced a 32 page chain of thought with loops and dead ends when asked to pick two poems. |
|
2026-07-15 06:59 UTC |
Claude3.5 Showcases nuanced reasoning in new video
According to TheRundownAI, Claude highlights nuanced reasoning and hope in hard questions in a new video, signaling richer enterprise use cases. |
|
2026-07-14 13:09 UTC |
Meta AI model aces APhO exam with perfect 30
According to AIatMeta, Meta’s model scored 30/30 on the Asian Physics Olympiad theoretical exam, tying top 3 students, showcasing advanced multimodal reasoning. |
|
2026-07-13 19:56 UTC |
GPT5.6 Launches on Bedrock with 3 Tiers
According to AWSNewsroom, OpenAI GPT-5.6 Sol, Terra, and Luna are now GA on Amazon Bedrock, offering flagship reasoning to fast inference tiers. |
|
2026-07-10 23:27 UTC |
GPT5.6 Sol powers complex deal analysis
According to gdb, GPT-5.6 Sol analyzes lending deals, reconciles terms, flags risks, and saves a cited report to Box for faster diligence. |