Heidelberg AI News

Heidelberg AI News

All stories

3 stories about GPT-6 Astra

Clear

Thursday 29 October

  • Latent Space

    ArticleLatent Space

    Can Frontier Models See Past a Decoy? Elorian Tests Visual Reasoning

    Elorian, which emerged from stealth in April with $55 million in seed funding co-led by Striker Venture Partners, Menlo Ventures and Altimeter Capital, launched its first visual reasoning model, Glimpse, described as a 'visual thinking model' that reasons about the physical world from what it sees. In partnership with AI data company Mercor, Elorian co-released DecoyCount, a benchmark testing whether models genuinely count objects or rely on visual shortcuts and substitution heuristics. According to a paper accompanying the release, frontier models scored below 50% accuracy, with GPT-6 Astra highest at 43.17%. CEO Andrew Dai said Glimpse scored 50.48% on 1,660 DecoyCount items in an internal test, making it the only model to clear 50%, though he noted the result came from a single run versus the frontier models' three-run averages. Dai, a former Google DeepMind researcher on GLaM MoE, PaLM 2 and Gemini, said the model uses visual thinking—generating and editing images as part of the reasoning trace—and distinguished Elorian's 'indirect action' approach from world models like World Labs' Atlas and Google's Genie 4.

  • ArticleThe Decoder

    OpenAI's Astra becomes the main target for Chinese AI labs after Anthropic tightens its defenses

    According to The Information, Chinese AI labs are now mainly tapping OpenAI's most powerful model, GPT-6 Astra, instead of Anthropic's Claude, to distill outputs for training their latest models. DeepSeek and Moonshot AI reportedly leaned heavily on Astra, including for Kimi K3.1 Pro and V4.1-Flash 1017. This kind of distillation requires the models to reveal their hidden reasoning steps, which OpenAI and Anthropic both ban. Claude used to be the go-to target, but Anthropic has tightened its defenses, saying its new methods cut the amount of data attackers can pull by 90 to 99 percent. In late September, OpenAI reported a coordinated distillation campaign it says involved people linked to Moonshot. Moonshot declined to comment and DeepSeek did not respond to a request for comment.

Tuesday 27 October