Turing

Hosting PhD's at Harvard University: asking the hard questions about AI Last week at Harvard, we hosted a PhD Happy Hour to connect with researchers across disciplines, from biology to physics. Questions weren't just about what domain experts currently do when working at Turing, but also about how they can use AI in the future to help write model breaking prompts even more effectively going forward. These conversations are just the beginning. More events are coming soon to bring exceptional PhD talent closer to the frontier of AI, starting with our PhD lunch at the Conference on Language Modeling next week.

COLM researchers: join us for an invite-only evening with peers driving frontier AI forward. At the Happy Hour, you’ll hear how Turing collaborates with leading AI Labs to accelerate ASI research with: • LLM reasoning, coding, and multimodal evaluation • Human-in-the-loop workflows for SFT, RLHF, and DPO • Benchmark-quality RL, multimodal, vision, robotics and STEM data packs This isn’t just networking. It’s a behind-the-scenes look at the systems and experiments defining next-generation ASI. Space is limited. RSVP now to secure your spot: https://luma.com/891ox4ar #COLM2025 #TuringAGIAdvancement #ResearcherEvent

In 1950, Alan Turing asked: "Can a machine think like a human?" That question sparked a movement 💡 Today, we measure intelligence differently. At Turing, AI isn’t evaluated by imitation, it’s assessed by real-world outcomes: - Is it safe? - Is it scalable? - Does it perform? From post-training alignment to embedded systems, our work redefines what it means to 'pass the test'. ➡️ Explore how Turing delivers AI that performs: https://www.turing.com/ #TuringTest75 #AlanTuring #LLMs #AGI #SuperIntelligence

Many businesses skip straight to customizing AI and then wonder why it underperforms. In this blog, Brent Blum, AI Solutions Lead at Turing, argues that without robust data pipelines, feedback loops, system context, and observability, even the best models fail to deliver. If your AI experiments aren’t producing consistent ROI, this is the missing piece. Read the full blog: https://lnkd.in/e2eBhg_k #TuringIntelligence #AIForward #LLMs

Thank you, Columbia University! Last week, 40+ Columbia PhD's and Masters students joined our happy hour in NYC. From Physics to Finance, they asked: “What do experts actually do at Turing?” We spoke about the longer version of the answer at the event, but in short: write model breaking prompts, build datasets based off domain knowledge, evaluate outputs, all in collaboration with leading researchers at the top AI labs. We're having a ton of fun going to campuses to share about work with Turing that pays well, applies expertise, and allows subject domain experts to collaborate directly with leading AI labs. #TuringTalent #Turing

ServiceNow just dropped Apriel-1.5-15B-Thinker. On par with DeepSeek-R1-0528 at 1/40th the size 🚀 ➡️ Runs on a single GPU with frontier-level reasoning capabilities ➡️ Benchmarks: Artificial Analysis 52 (beats Gemini-2.5-Flash, GLM-4.5), IFBench 62, Tau 68 ➡️ Mid-training alone yields major gains (no RL applied yet) ➡️ Turing × ServiceNow: high-quality SFT data across code, instruction tuning, agentic tasks and complex reasoning. Grateful to our partners at ServiceNow and to Sathwik Tejaswi Madhusudan and Srinivas Sunkara, for driving this collaboration forward alongside our team, led by Sudarshan Sivaraman, GM of AGI Advancement at Turing. Together, we’re accelerating model innovation with real-world impact. Explore the model: https://lnkd.in/exmZC4J6 Test the chat demo: https://lnkd.in/eC_vx8zr #AI #LLM #Reasoning #GenAI #EnterpriseAI #Efficiency #ServiceNow #Turing

Congratulations to the team at Anthropic on the release of Claude Sonnet 4.5! It's officially Anthropic’s most capable coding model to date! It now leads on SWE-bench, surpassing other models and delivers improved reasoning on HumanEval, RAG, and long-horizon agentic tasks. With 30 hours of autonomous execution, it’s built to assist real developers, not just pass benchmarks. At Turing, we build around this kind of progress, embedding agents into coding pipelines that prioritize accuracy, alignment, and real-world maintainability. https://lnkd.in/eEN3MDfc #ClaudeSonnet4.5 #Anthropic #LLMs #Claude

We’re live from Groceryshop, connecting with retail and CPG leaders shaping the next wave of AI-driven transformation. Stop by to talk Superintelligence, meet the Turing team, and hear how brands are moving from instinct to impact with AI. Thanks to our sales team for capturing the energy on the ground. To learn more about what we do, visit: https://lnkd.in/eBa7-dSY #TuringIntelligence #RetailAI #AgenticAI #Groceryshop

AI that doesn’t integrate won’t scale. In this strategic piece, James Raybould, SVP and GM of Turing Intelligence, breaks down why so many enterprise AI pilots stall, and what to do about it. The shift from generic AI to ‘Proprietary Intelligence’ isn’t about bigger models. It’s about embedding feedback, domain memory, and human oversight to turn AI into a long-term business advantage. ➡️ Read how leading organizations are driving measurable gains in efficiency, decision speed, and system resilience: https://lnkd.in/dERpCaEX #TuringIntelligence #AICopilots #AIForward

Most enterprises stall in the handoff between research and execution. Turing connects both: Turing AGI Advancement → frontier model innovation Turing Intelligence → operationalizes pricing, personalization, & logistics Find us at Groceryshop to see how we move from roadmap to retail-ready systems. DM us to book a 1:1 today with our strategists: https://lnkd.in/eDi6kxjP #TuringIntelligence #RetailAI #AgenticAI #Groceryshop