Ai2 rebuilds olmo-core 3 to train trillion-parameter AI 2.7x faster. Ai2 open-sources Olmo-core 3, a redesigned MoE training stack that hits 2.7x throughput and scales past a trillion parameters. · 4 hrs ago Read 5 min * Ai2 released Olmo-core 3, open MoE training infrastructure scaling to trillion-parameter models. * Delivers 2.7x throughput over Olmo-core 2 on a 47B MoE across 8 B300 GPUs. * Switches from FSDP to DDP, keeping experts resident on GPUs instead of regathering weights. * Benchmarked at 1.2T parameters on 512 GPUs hitting 858 TFLOP/s/GPU with random routing. * MXFP8 support lifts throughput ~21% versus BF16 while cutting peak memory from 103 to 95 GiB. * Tech report documents failures like token gerrymandering and misleading load-balance scores. Ai2 rebuilds Olmo's training stack for trillion-parameter MoEs. Ai2, the Allen Institute for AI, has released Olmo-core 3, a fully open training stack for large mixture-of-experts models. The organization plans to use it for the next generation of Olmo models. Mixture-of-experts models contain many feed-forward expert networks, while a router sends each token to only a small selection. This sparse activation lets parameter capacity grow without a proportional increase in computation. At cluster scale, token exchange, weight movement, optimizer state, and routing coordination can consume those savings. Olmo-core 3 is designed to keep throughput nearly flat as the expert pool expands. A 2.7x gain on eight B300s. Ai2 reports two preliminary benchmarks that show how the stack handles both implementation overhead and rising expert counts. Ai2's preliminary MoE benchmarks | Test | Configuration | Result | | Training stack | 47B-parameter MoE on eight NVIDIA B300 GPUs | 52,000 tokens per second per GPU, up from 19,400 with the earlier FSDP implementation, or about 2.7x higher throughput | | Expert scaling | Expert pool increased from 8 to 128, with four experts selected per token and about 3.2B active parameters | Total capacity increased from 4.6B to 47B parameters while throughput declined by less than 5% | Resident experts cut weight traffic. Ai2's earlier MoE implementation used fully sharded data parallelism, or FSDP, configured to gather and reshard expert weights for each small batch. Once experts outnumber GPUs, repeatedly moving those weights across the interconnect can leave compute units waiting for data. Olmo-core 3 combines a DDP-based design with expert parallelism. Experts remain resident on assigned GPUs, and token data travels to the devices holding the selected experts. This layout avoids gathering and resharding the full set of expert weights for every batch. The cluster is split three ways. Fitting a trillion-parameter model requires partitioning expert weights, model layers, and optimizer state across the cluster: * Expert parallelism assigns different experts to different GPUs, so each device stores only part of the expert pool. * Pipeline parallelism assigns groups of model layers to different GPU stages, reducing the weights held by each device. * A distributed optimizer shards optimizer state across GPUs, avoiding a full copy on every device. Routing work stays on the GPU. Communication speed also depends on how routed tokens are packed, transferred, and presented to each expert. Olmo-core 3 adds three optimizations for that path: * Rowwise expert parallelism writes routed token rows directly into expert input buffers, reducing rearrangement and copying. * GPU-resident routing keeps routing metadata on the device, allowing the CPU to queue work without waiting for a copy back from the GPU. * Grouped GEMM combines many small expert matrix multiplications into grouped operations that use the GPU more efficiently. MXFP8 reduces compute and memory costs. Olmo-core 3 supports MXFP8, an 8-bit floating-point format that applies separate scaling to small blocks of values. On four NVIDIA B300 GPUs with work distributed uniformly across experts, selective MXFP8 use increased training throughput by about 21% over a BF16 baseline. Peak active memory fell from 103 GiB to 95 GiB. Feed-forward computation and expert communication supplied most of the improvement, with a smaller contribution from attention. The measurement used uniform routing; learned routers can produce uneven workloads. Full-run convergence and final model quality remain separate validation questions. A 1.2-trillion-parameter systems test. Ai2 also tested a model with 1.2 trillion total parameters and 58.36 billion active parameters per token across 512 GPUs. The run reached a peak of 858 TFLOP/s/GPU, counting useful model computation. Random routing isolated system performance, leaving training quality outside the scope of the test. Using DeepEP v2 for expert communication, the team also reached 2.38 trillion total parameters. That figure comes from a short capacity test, with sustained-training behavior still unreported. Four failure modes from the report. The technical report documents several negative results that affect implementation and benchmarking: * A routing-balance score improved even as the actual workload became less balanced, a failure mode the team calls token gerrymandering. * Reducing expert learning rates to account for the smaller number of tokens processed by each expert did not improve results in the tested model family. * GPU execution times changed with input values even when matrix dimensions stayed fixed, so reliable performance comparisons require matching inputs as well as matching shapes. * Overlapping communication and computation on separate GPU streams sometimes increased end-to-end training time. Large expert pools are the best fit. Olmo-core 3 is most relevant when teams need to expand expert counts while keeping active parameters stable, combine pipeline and expert parallelism to fit a model, or inspect and modify the complete training implementation. Dense models and small MoEs that already fit comfortably on a cluster offer less reason to migrate from Megatron-Core or an established FSDP stack. The source code is available on GitHub. Ai2 has also published an interactive walkthrough showing how data, expert, and pipeline parallelism compose from one GPU to 512 GPUs. Ai2 is treating training infrastructure as part of its open-model release strategy. Publishing the stack gives researchers access to the implementation choices, benchmarks, and failed experiments behind its next Olmo generation. The release removes a software barrier for academic groups and smaller labs, while the hardware demands of the largest configurations remain substantial.

Ai2 and Global Fishing Watch unite to bring AI agents to ocean monitoring. * Published September 23, 2026 Strategic partnership will bring next-generation AI and satellite analytics to ocean monitoring, advancing responsible AI through transparency and human oversight. NEW YORK, United States - Ai2, the nonprofit AI lab behind the Skylight maritime intelligence platform, and Global Fishing Watch, the international organization advancing ocean governance through transparency of human activity at sea, today announced a partnership to bring cutting-edge technology and AI agents to ocean monitoring and enforcement, giving authorities around the world new tools to detect, analyze and investigate activity at sea. Announced at Climate Week NYC 2026, the collaboration pairs Ai2's state-of-the-art AI models and Skylight's enforcement-tailored detection capabilities with Global Fishing Watch's mapping technology, ocean data and international experience supporting governments with fisheries monitoring and enforcement. Moving beyond complementary technologies to true co-development, the two organizations plan to integrate the latest satellite data, build new and real-time computer vision models and explore AI agents that fuse data sources to uncover patterns an analyst might otherwise miss. "AI is fundamentally changing how we access information about our ocean and the human activity that takes place on it," said Tony Long, chief executive officer of Global Fishing Watch. "But the technology's real value lies in its ability to get timely and actionable insights to those who can drive beneficial impact." Both organizations already build AI and work directly with maritime authorities. By combining forces, the two organizations are now seeking to scale their impacts, moving from distinct efforts to a unified detection pipeline, jointly developed AI models and a shared pool of maritime expertise. Transparency and human oversight will remain central to that work, with AI designed to support rather than replace human judgment. The technology can surface patterns and potential risks, while people determine what those signals mean and how to act on them. "We are at an inflection point in deciding how AI enters the ocean space," said Namrata Kolla, Head of Skylight. "Our maritime partners have made it clear that they care about having purpose-built AI where they own their own data, leverage open models, and decide what is the 'right' answer. Each step we take is pushing toward that vision." Scaling what works: from AI innovation to ocean action Global Fishing Watch and Ai2 have long collaborated, sharing data, building vessel registry resources together and working concertedly as part of the Joint Analytical Cell, a coalition that pools analysis to support fisheries enforcement. Most recently, in March 2026, Skylight's Visible Infrared Imaging Radiometer Suite, or VIIRS, went live on the Global Fishing Watch map, showing near-real-time detections of vessel lights at night and giving users a fuller picture of ocean use. The next iteration of the partnership will co-develop AI agents like Shippy, Skylight's AI agent, to support enhanced analysis delivery. Rather than manually navigating datasets and analytical tools, users will now be able to ask about a vessel's history, activity or unusual patterns around a marine protected area or within a particular region. After the prompt, the agent will search the relevant data and return vessels, events and insights for further investigation. "By combining Ai2's technological expertise with our data, platforms and partnerships around the world, we can accelerate the journey from detection to decision," Long continued. "This will give governments, civil society and ocean stewards faster, more powerful ways to understand what is happening at sea and protect the resources under their care." The technical collaboration also extends beyond Shippy. Ai2 will make OlmoEarth available to Global Fishing Watch to support data annotation and model development, while the organizations will jointly develop machine learning models that can extract vessel detections and other insights from satellite imagery. Together, these capabilities will hasten the transformation of enormous volumes of raw ocean data into useful intelligence and support Global Fishing Watch's ambition to map all human activity across the ocean by 2030. In addition, Global Fishing Watch super users will now be able to craft a more immediate picture of the ocean as real-time vessel detections generated through Ai2's Skylight platform are integrated into the Global Fishing Watch marine manager portal. Faster access to ocean intelligence will help accelerate the monitoring, control and surveillance of suspected illegal, unreported and unregulated fishing in marine protected areas and bolstering enforcement efforts. Building responsible AI for people and planet Against that backdrop, the two organizations will also inaugurate a series of in-depth conversations within the environment and technology communities to promote AI best practices. Rooted in principles of transparency and equitable access, and grounded in rigorous impact assessment, the conversations seek to ensure that open AI standards and shared tools remain trustworthy, accountable, and designed for the global public good. "Our goal isn't AI for AI's sake. Our focus is on using technology to make the ocean more visible and give decision-makers the critical information they need to ultimately enable better decisions and drive impact," added Namrata Kolla. " If we can make the extraordinary advances taking place in AI work for the ocean in an open and responsible way and with human oversight and at global scale, we can fundamentally change our ability to protect our planet's most valuable resource."

Ai2 and Global Fishing Watch unite to bring AI agents to ocean monitoring. Strategic partnership brings next-generation AI and satellite analytics to ocean monitoring, advancing responsible AI through transparency and human oversight AI is fundamentally changing how Today in New York access information about its ocean. Its real value lies in its ability to get timely and actionable insights to those who can drive beneficial impact." - Tony Long NEW YORK, NY, UNITED STATES, September 23, 2026 / EINPresswire.com / - Ai2, the nonprofit AI lab behind the Skylight maritime intelligence platform, and Global Fishing Watch, the international organization advancing ocean governance through transparency of human activity at sea, today announced a partnership to bring cutting-edge technology and AI agents to ocean monitoring and enforcement, giving authorities around the world new tools to detect, analyze and investigate activity at sea. Announced at Climate Week NYC 2026, the collaboration pairs Ai2's state-of-the-art AI models and Skylight's enforcement-tailored detection capabilities with Global Fishing Watch's mapping technology, ocean data and international experience supporting governments with fisheries monitoring and enforcement. Moving beyond complementary technologies to true co-development, the two organizations plan to integrate the latest satellite data, build new and real-time computer vision models and explore AI agents that fuse data sources to uncover patterns an analyst might otherwise miss. "AI is fundamentally changing how we access information about our ocean and the human activity that takes place on it," said Tony Long, chief executive officer of Global Fishing Watch. "But the technology's real value lies in its ability to get timely and actionable insights to those who can drive beneficial impact." Both organizations already build AI and work directly with maritime authorities. By combining forces, the two organizations are now seeking to scale their impacts, moving from distinct efforts to a unified detection pipeline, jointly developed AI models and a shared pool of maritime expertise. Transparency and human oversight will remain central to that work, with AI designed to support rather than replace human judgment. The technology can surface patterns and potential risks, while people determine what those signals mean and how to act on them. "We are at an inflection point in deciding how AI enters the ocean space," said Namrata Kolla, Head of Skylight. "Our maritime partners have made it clear that they care about having purpose-built AI where they own their own data, leverage open models, and decide what is the 'right' answer. Each step we take is pushing toward that vision." Scaling what works: from AI innovation to ocean action Global Fishing Watch and Ai2 have long collaborated, sharing data, building vessel registry resources together and working concertedly as part of the Joint Analytical Cell, a coalition that pools analysis to support fisheries enforcement. Most recently, in March 2026, Skylight's Visible Infrared Imaging Radiometer Suite, or VIIRS, went live on the Global Fishing Watch map, showing near-real-time detections of vessel lights at night and giving users a fuller picture of ocean use. The next iteration of the partnership will co-develop AI agents like Shippy, Skylight's AI agent, to support enhanced analysis delivery. Rather than manually navigating datasets and analytical tools, users will now be able to ask about a vessel's history, activity or unusual patterns around a marine protected area or within a particular region. After the prompt, the agent will search the relevant data and return vessels, events and insights for further investigation. "By combining Ai2's technological expertise with our data, platforms and partnerships around the world, we can accelerate the journey from detection to decision," Long continued. "This will give governments, civil society and ocean stewards faster, more powerful ways to understand what is happening at sea and protect the resources under their care." The technical collaboration also extends beyond Shippy. Ai2 will make OlmoEarth available to Global Fishing Watch to support data annotation and model development, while the organizations will jointly develop machine learning models that can extract vessel detections and other insights from satellite imagery. Together, these capabilities will hasten the transformation of enormous volumes of raw ocean data into useful intelligence and support Global Fishing Watch's ambition to map all human activity across the ocean by 2030. In addition, Global Fishing Watch super users will now be able to craft a more immediate picture of the ocean as real-time vessel detections generated through Ai2's Skylight platform are integrated into the Global Fishing Watch marine manager portal. Faster access to ocean intelligence will help accelerate the monitoring, control and surveillance of suspected illegal, unreported and unregulated fishing in marine protected areas and bolstering enforcement efforts. Building responsible AI for people and planet Against that backdrop, the two organizations will also inaugurate a series of in-depth conversations within the environment and technology communities to promote AI best practices. Rooted in principles of transparency and equitable access, and grounded in rigorous impact assessment, the conversations seek to ensure that open AI standards and shared tools remain trustworthy, accountable, and designed for the global public good. "Our goal isn't AI for AI's sake. Our focus is on using technology to make the ocean more visible and give decision-makers the critical information they need to ultimately enable better decisions and drive impact," added Namrata Kolla. " If we can make the extraordinary advances taking place in AI work for the ocean in an open and responsible way and with human oversight and at global scale, we can fundamentally change our ability to protect our planet's most valuable resource." Legal Disclaimer: EIN Presswire provides this news content "as is" without warranty of any kind. Today in New York do not accept any responsibility or liability for the accuracy, content, images, videos, licenses, completeness, legality, or reliability of the information contained in this article. If you have any complaints or copyright issues related to this article, kindly contact the author above.