Geo News
Community curated by people like you
LatestAICryptoHealthWorld AffairsUS Politics

Reasoning models stories

Oct 4, 2026

Google restricts free Gemini access to Flash-Lite model starting October 9

Google will limit free Gemini users to only the Flash-Lite model starting October 9, 2026, removing access to the more powerful Flash and Pro models. The $5/month AI Plus tier will also lose access to Pro, while AI Pro subscribers gain the new Deep Think reasoning feature.

Oct 4, 2026·4 sources
00
Sep 30, 2026

OpenAI accuses Moonshot of coordinated campaign to extract model reasoning

OpenAI says it disrupted a campaign involving over 15,000 accounts attempting to copy hidden reasoning from its models, linking part of the activity to people connected to Chinese AI lab Moonshot.

Sep 30, 2026·3 sources
00
Sep 15, 2026

Salesforce and Nvidia launch Koa reasoning model

Salesforce announced Koa, its first reasoning AI model built on Nvidia's open-weight Llama architecture and specialized for CRM data. The model represents a partnership between the two companies at Dreamforce conference.

Sep 15, 2026·2 sources
00
Aug 12, 2026

Pathway raises funding at $500 million valuation with 150M-parameter AI model

Pathway has secured funding at a $500 million valuation, backed by Id4 Ventures, TQ Ventures, Red Bridge Ventures, Kadmos Capital, WS Investment Co, and Databricks chief AI scientist Jonathan Frankle. The company is challenging AI reasoning cost curves with its 150-million-parameter model.

Aug 12, 2026·2 sources
00
Aug 7, 2026

Study finds next-generation AI medical models still perpetuate racial and gender stereotypes

Flinders University researchers evaluated two advanced reasoning large language models—o3-mini and DeepSeek-R1—and discovered they continue to reproduce racial and gender stereotypes when describing fictional patients with common medical conditions, despite being newer generation AI systems.

Aug 7, 2026·1 source
00
May 21, 2026

OpenAI model claims to disprove Erdős conjecture

OpenAI announced that its reasoning model has produced an original mathematical proof disproving a famous unsolved conjecture about planar unit distances, first posed by mathematician Paul Erdős in 1946.

May 21, 2026·2 sources
00
May 10, 2026

NVIDIA Releases Star Elastic: Single Checkpoint Contains Multiple Reasoning Models

NVIDIA researchers have introduced Star Elastic, a post-training method that embeds multiple nested reasoning models at 30B, 23B, and 12B parameter scales inside a single checkpoint, enabling zero-shot model slicing without retraining.

May 10, 2026·1 source
00
May 7, 2026

Zyphra Releases ZAYA1-8B Reasoning Model Trained on AMD Hardware

Zyphra launched ZAYA1-8B, a Mixture of Experts reasoning model with only 760M active parameters that reportedly outperforms much larger open-weight models on math and coding benchmarks, trained entirely on AMD hardware.

May 7, 2026·1 source
00
Apr 12, 2026

Arcee AI Spends Half Its Venture Capital to Build Open Reasoning Model Rivaling Claude Opus

US startup Arcee AI invested approximately half of its total venture capital to train Trinity-Large-Thinking, a 400 billion parameter open reasoning model designed to compete with Anthropic's Claude Opus in agent tasks.

Apr 12, 2026·1 source
00
Apr 5, 2026

Anthropic Leaked 'Mythos' Model Shows Advanced Cybersecurity Capabilities

A leaked draft blog post reveals Anthropic's unreleased 'Mythos' model, a compute-intensive LLM with advanced reasoning capabilities specifically designed for enterprise cybersecurity use cases.

Apr 5, 2026·2 sources
00
Mar 13, 2026

Anthropic Launches Claude 3.7 Sonnet Amid Infrastructure Struggles and Pricing Controversies

Anthropic released Claude 3.7 Sonnet, its first hybrid reasoning model, while simultaneously experiencing its tenth service outage in 12 days and pausing token-based billing for its Agent SDK due to unexpectedly high costs for developers.

Mar 13, 2026·9 sources
00

Top claims

  • ▪In October 2026, Google plans to introduce low, medium, and high computational effort settings for Gemini models, which will allow users to adjust reasoning depth at the expense of their weekly usage limits.
  • ▪Google has begun rolling out its new Gemini 4 Argon model, but access is initially limited to AI Ultra subscribers, U.S. government agencies, and a small group of trusted partners.
  • ▪Starting October 9, 2026, Google AI Plus subscribers, who pay $4.99 per month, will lose access to the Gemini Pro model, leaving them with access to only the Flash-Lite and Flash models.

People involved

Jonathan Frankle

Subtopics

AI foundation models4AI research & benchmarks4AI startups4Compute, chips & AI infrastructure3AI agents2AI security2AI tools & products2Business & enterprise AI2Large language models (LLMs)2Open-source AI2AI assistants & chatbots1AI bias & fairness1AI breakthroughs1AI copyright & training data1AI ethics1AI for developers1AI in healthcare1AI math benchmarks1AI misuse1AI safety & social impact1China1Chinese copying of American AI models1Drug Approval and Clinical Trials1Drug Policy, Pricing & Approval1

Related timelines

Iran War

209 stories

Russia-Ukraine war

136 stories

Strait of Hormuz blockage

148 stories

Trump administration

161 stories

Congress

110 stories