Together AI has released DeepSWE, a state-of-the-art, fully open-sourced software engineering agent that is trained entirely through reinforcement learning (RL). Built on top of the Qwen3-32B language model, DeepSWE achieves 59% accuracy on the SWEBench-Verified benchmark and 42.2% Pass@1, topping the leaderboard among open-weight models. This launch represents a significant shift for Together AI, from […] The post Together AI Releases DeepSWE: A Fully Open-Source RL-Trained Coding Agent Based on Qwen3-32B and Achieves 59% on SWEBench appeared first on MarkTechPost.

Anthropic has released Claude Opus 4.5, claiming an industry-leading 80.9% coding score and introducing "Tool Search" with a promise to reduce agent costs by 85%.

Coding agents are the latest promising Artificial Intelligence (AI) tool, and an impressive step up from LLMs. This article is a deep dive into them, with the creators of SWE-bench and SWE-agent.

Today, AWS announces the update of Amazon Q Developer’s software development agent. This new agent achieves state-of-the-art performance on industry benchmark SWTBench Verified (49%) and sits among the top ranking models on SWEBench Verified (66%). The agent has access to tools for planning and reasoning that use the capacity of advanced models to their fullest. By running in a dedicated environment with built-in access to all the functionalities of a modern IDE, the agent is now able to generate multiple candidate solutions for a given problem, select the most promising one, and return higher quality code to the developer. With this new agent, developers can further accelerate their development team velocity. The update to the agent translates to more reliable suggestions and reduced debugging time for developers. This allows developers to focus on higher-level design and innovation, while the agent handles more routine coding tasks with increased accuracy. The new software development agent for Amazon Q Developer is available in all AWS Regions where Amazon Q is supported. Getting started with the software development agent is simple. Developers can begin using it immediately by typing '/dev' in the Q chat window in Visual Studio Code or JetBrains integrated development environment (IDE) where the Amazon Q Developer plugin is installed. To learn more about Amazon Q, visit the Amazon Q product page or refer to the agent documentation.

I predicted future AI models would someday learn to cheat on SWE-bench. They were already doing it.

A post for developers about the new Claude 3.5 Sonnet and the SWE-bench eval

Searching code is an important part of every developer's workflow. We're trying to make it better.

The edge shifts to practical local RAG while legal and dependency risks mount.

Together AI wprowadza DeepSWE, otwartego agenta oprogramowania, który dzięki uczeniu wzmocnionemu (RL) osiąga 59% skuteczności w teście SWEBench, wyznaczając nowy kierunek rozwoju agentów AI.

Aide is now SOTA on swebench-verified, solving 62.2% of benchmark issues. We do this by scaling our inference agent at test time and relearning the bitter lesson. > The biggest lesson to be learned from 70 years of AI research is that general methods that use computation are the most effective, and by a large margin. In the midst of this exploration, we also developed a…