Super excited about HydraFusion in GitHub Copilot, and what it shows about the shift from model selection to model orchestration.

GitHub has launched HydraFusion, a research preview in GitHub Copilot that orchestrates multiple AI models to plan, build, critique, and complete coding tasks, delivering outcomes at up to 67% lower cost. The system dynamically selects workflows to balance quality, cost, and latency, marking a shift from model selection to model orchestration. Available in the GitHub Copilot CLI, HydraFusion outperforms leading models like Opus on terminal Bench 2.1 while costing 67% less.

Super excited about HydraFusion in GitHub Copilot, and what it shows about the shift from model selection to model orchestration. By bringing together multiple models to plan, build, critique, and complete coding tasks, it can deliver outcomes at up to 67% lower cost. It’s a great example of the value of a heterogeneous model ecosystem, and how we’re continuing to advance the cost-to-outcome frontier. Try it out here: https://lnkd.in/gGC6yNfe Every coding task asks for a different balance of quality, costs, and latency. Hydro Fusion changes that. Hydrogen fusion optimizes the path, not just the model it risks. The task decides how much work it deserves, and chooses a process designed to produce the right balance of quality, costs, and latency that no single fixed model can provide across every task. How Diffusion is launching as an experimental research preview and on terminal Bench 2.1. It outperform a leading model like Opus files while costing 67 less. And that is just the beginning. Hydro fusion can take one of three execution paths with single. One model solves the task on its own when additional orchestration will not add enough value. Adder fusion keeps the path direct. With cascade, one model takes the first pass and a gate evaluates the result and decides whether to accept it or escalate the task to a more capable. Additional work is introduced only when it is likely to improve the outcome. And lastly, we've critique one model riser draft. A second motor reviews it and identifies what is wrong without being able to change the draft itself. The first model then uses that feedback to repair its solution. This at screening where it can materially improve the result. With Hydra Fusion, you stay focused on the problem while compiler builds the best path to the solution. HydraFusion is now available as a research preview in GitHub Copilot 🚀 Instead of asking users to choose the right model for every task, HydraFusion dynamically brings together models and workflows to balance quality and cost. As new frontier models emerge, HydraFusion evolves to make the most of each model’s strengths. Really proud to have been part of this journey and grateful to everyone across research, engineering, and product who helped bring it to life. Aashna GargShengyu FuHuamin ChenCarlos CastroAndy SalernoAnisa Majhi Read more: https://lnkd.in/gePDSiwh Try the research preview and let us know what you think Super excited about HydraFusion in GitHub Copilot, and what it shows about the shift from model selection to model orchestration. By bringing together multiple models to plan, build, critique, and complete coding tasks, it can deliver outcomes at up to 67% lower cost. It’s a great example of the value of a heterogeneous model ecosystem, and how we’re continuing to advance the cost-to-outcome frontier. Try it out here: https://lnkd.in/gGC6yNfe Today we're launching HydraFusion as a public research preview in GitHub Copilot 🚀 💡 HydraFusion is a key piece of our broader strategy: automated semantic routing between models. 💡 For developers, all of that complexity stays behind the scenes. You select HydraFusion like any other model, and it picks the workflow that best balances performance, cost, and latency for each task. Try it today in the GitHub Copilot CLI and most importantly, tell us what you think. We're actively reading the discussion thread and using your feedback to take HydraFusion to the next level: https://lnkd.in/gVWgBQHt This was a real team effort. Huge thank you to everyone who helped get us here. 🙏 Super excited about HydraFusion in GitHub Copilot, and what it shows about the shift from model selection to model orchestration. By bringing together multiple models to plan, build, critique, and complete coding tasks, it can deliver outcomes at up to 67% lower cost. It’s a great example of the value of a heterogeneous model ecosystem, and how we’re continuing to advance the cost-to-outcome frontier. Try it out here: https://lnkd.in/gGC6yNfe Very exciting news - The shift from model selection to model orchestration feels like a significant milestone for the industry. The real value comes from combining the strengths of different models to optimise both quality and cost. Super excited about HydraFusion in GitHub Copilot, and what it shows about the shift from model selection to model orchestration. By bringing together multiple models to plan, build, critique, and complete coding tasks, it can deliver outcomes at up to 67% lower cost. It’s a great example of the value of a heterogeneous model ecosystem, and how we’re continuing to advance the cost-to-outcome frontier. Try it out here: https://lnkd.in/gGC6yNfe OK, this is super cool. Not another new better/faster model, but a new way of orchestrating the models you already use. Uses context and intelligence to optimize paths for better performance and wait for it... much less cost Just in research preview now, but truly excited to see what this unlocks. Super excited about HydraFusion in GitHub Copilot, and what it shows about the shift from model selection to model orchestration. By bringing together multiple models to plan, build, critique, and complete coding tasks, it can deliver outcomes at up to 67% lower cost. It’s a great example of the value of a heterogeneous model ecosystem, and how we’re continuing to advance the cost-to-outcome frontier. Try it out here: https://lnkd.in/gGC6yNfe At the risk of hyperbole - this changes the entire playing field. Extending HyDRA from model selection to workflow orchestration is the next frontier. GitHub continuing to lead from the front Super excited about HydraFusion in GitHub Copilot, and what it shows about the shift from model selection to model orchestration. By bringing together multiple models to plan, build, critique, and complete coding tasks, it can deliver outcomes at up to 67% lower cost. It’s a great example of the value of a heterogeneous model ecosystem, and how we’re continuing to advance the cost-to-outcome frontier. Try it out here: https://lnkd.in/gGC6yNfe This is exactly why enterprise development needs to shift to a multi-model paradigm. The value isn’t just having access to multiple LLMs, or picking the cheapest one for a given request. It’s being able to orchestrate them: one model drafts, another reviews, another challenges the approach, and the system keeps iterating until the strongest result emerges. HydraFusion is our latest research project and we are leading once again the market to where enterprise AI is going: comparable quality to frontier models at up to 67% lower cost, without being locked into a single provider and without putting the burden on each developer to create manual and therefore fragile harnesses. PS : the linked blog post also goes into detail about our CheckpointBench results, which focuses testing against real Copilot coding sessions rather than relying only on abstract benchmarks, ans based on my personal experience it feels very accurate. Super excited about HydraFusion in GitHub Copilot, and what it shows about the shift from model selection to model orchestration. By bringing together multiple models to plan, build, critique, and complete coding tasks, it can deliver outcomes at up to 67% lower cost. It’s a great example of the value of a heterogeneous model ecosystem, and how we’re continuing to advance the cost-to-outcome frontier. Try it out here: https://lnkd.in/gGC6yNfe Interesting shift from “which model should I use?” to “how do I get the best outcome across models?” Model orchestration feels like it’s going to be a big part of what comes next. github copilot Super excited about HydraFusion in GitHub Copilot, and what it shows about the shift from model selection to model orchestration. By bringing together multiple models to plan, build, critique, and complete coding tasks, it can deliver outcomes at up to 67% lower cost. It’s a great example of the value of a heterogeneous model ecosystem, and how we’re continuing to advance the cost-to-outcome frontier. Try it out here: https://lnkd.in/gGC6yNfe This will not surprise anyone who uses both OpenAI and Anthropic models in their workflows: in some benchmarks GPT wins, in others Claude, but the best result comes from orchestrating them together. Our experiments proved model diversity matters: combining models from different families e.g. Claude+GPT outperformed a single model e.g. Claude only with equivalent compute and reasoning levels. Two models appears to be the sweet spot for now; adding a third didn't meaningfully improve quality. We took a data-driven approach to find fusion combinations across Claude, GPT, Gemini, and others that hit frontier quality for a given cost, and packaged the best of them into HydraFusion, a new experimental "model" in GitHub Copilot. An extremely fast classifier analyzes prompt complexity and picks the right path: a single fast model for simple queries, or a cascade/critique ensemble for complex ones. That cuts spend up to 67% on Terminal Bench 2.1 and 35% on DeepSWE while solving some tasks no single frontier model could. The router isn't perfect yet, and it will improve with usage. But the routing choice is transparent, so you can build trust in it and course-correct when it's wrong. And it updates automatically as new models ship, so you no longer need to constantly switch guess 🙃? in the model picker. Solid work by the GitHub team Super excited about HydraFusion in GitHub Copilot, and what it shows about the shift from model selection to model orchestration. By bringing together multiple models to plan, build, critique, and complete coding tasks, it can deliver outcomes at up to 67% lower cost. It’s a great example of the value of a heterogeneous model ecosystem, and how we’re continuing to advance the cost-to-outcome frontier. Try it out here: https://lnkd.in/gGC6yNfe The future of Microsoft Copilot includes model orchestration. Business users are not concerned with the model, so this should simplify and improve outcomes without requiring a model choice. Copilot HydraFusion LLM AIModels Microsoft Super excited about HydraFusion in GitHub Copilot, and what it shows about the shift from model selection to model orchestration. By bringing together multiple models to plan, build, critique, and complete coding tasks, it can deliver outcomes at up to 67% lower cost. It’s a great example of the value of a heterogeneous model ecosystem, and how we’re continuing to advance the cost-to-outcome frontier. Try it out here: https://lnkd.in/gGC6yNfe Really exciting development. The shift from model selection to model orchestration is a game changer, especially when it improves both quality and cost. Looking forward to seeing how developers leverage HydraFusion in real-world workflows. Super excited about HydraFusion in GitHub Copilot, and what it shows about the shift from model selection to model orchestration. By bringing together multiple models to plan, build, critique, and complete coding tasks, it can deliver outcomes at up to 67% lower cost. It’s a great example of the value of a heterogeneous model ecosystem, and how we’re continuing to advance the cost-to-outcome frontier. Try it out here: https://lnkd.in/gGC6yNfe