Randomly switching between LMs at every step boosts SWE-bench score
What if your agent uses a different LM at every turn? We let mini-SWE-agent randomly switch between GPT-5 and Sonnet 4 and it scored higher on SWE-bench than with either model separately. GPT-5 by itself gets 65.0%, Sonnet 4 64.8%, but randomly switching at every step gets us 67.2% This result came pretty surprising to us. There's a few more experiments in the blog post.
Does the same job
all alternatives →
- MSMini-swe-agent achieves 65% on SWE-bench in 100 lines of python2025 · github.com · ▲7
- BABenchmark AI on your actual code (GPT-5, Claude, Grok, Gemini, o3)Oct 2025 · codelens.ai · ▲7
I built CodeLens.AI - a tool that compares how 6 top LLMs (GPT-5, Claude Opus 4.1, Claude Sonnet 4.5, Grok 4, Gemini 2.5 Pro, o3) handle your actual code tasks. How it works: - Upload code + describe task (refactoring, security review, architecture, etc.) - All 6 models run in parallel (~2-5 min) - See side-by-side comparison with AI judge scores - Community votes on winners (blind voting) - Each evaluation gets reflected in the overall AI model leaderboard, showing us best ones Why I built this: Existing benchmarks (HumanEval, SWE-Bench) don't reflect real-world developer tasks. I wanted to…
- WMWe made GPT-4.1-mini beat 4.1 at Tic-Tac-Toe using dynamic context2025 · github.com · ▲5
We wanted to test if a smaller model like GPT-4.1-mini could beat its bigger brother 4.1 at the game Tic-Tac-Toe using only context engineering. We put them in a 100-game tournament. For the smaller model, we gave it a few examples of winning moves from past games right before it made its own move. The results were clear. Without the examples, the smaller model struggled against GPT-4.1. With the examples, its effectiveness increased by nearly 200%, and it consistently won. It's a simple demonstration, but it shows that a smaller, faster model with good, timely examples can outperform a more…
- ARA registry of agent benchmarks (including many OSS agent trajectories)2024 · explorer.invariantlabs.ai · ▲6
If you're interested in exploring what LLM-based agent systems these days actually do to solve certain benchmarks such as SWEBench or WebArena, we created a small leaderboard with our team, that allows to view a lot of public and OSS agent results including all the runtime traces (the step-by-step reasoning behind the scenes). Looking at traces is actually quite interesting, as they reveal a lot about the inner working and shortcomings of current agent system, e.g. see https://explorer.invariantlabs.ai/u/invariant/webarena--SteP... for an example trace.
More ai this month
the category →
I trained a 125M-parameter transformer to autocomplete piano performances in real time (~108 notes/sec on an iPhone 15). The idea is basically GitHub Copilot or Tabnine, except instead of prompting it with code, you prompt it by playing a few notes on a MIDI piano. The model then continues what you played, entirely on-device. The app is free if anyone wants to try it. Happy to answer questions about the model, training, Core ML, or the many things that didn't work.
AI · 17d ago · simedw.com
Astute▲585Automate your B2B brand going viral, with new media creators
AI · 18d ago · company-app.joinastute.com
Hey HN, Henry from Cactus here! We previously released Cactus Needle, a 14MB agentic LLM for tool call, device use, and structured extraction for phones, wearables, smart homes, small robots and microcontrollers. We got really great feedback here, and have now incorporated the suggestions to release Needle 2. The whole model is a single 14MB binary that runs a full session in 28MB of RAM; 45m parameters at 2bit compression. Needle hits 500 tokens/sec decode speed on a Raspberry Pi 5, sits between 400-1,500 tokens/sec on VR devices like Meta Quest 3S and Apple Vision Pro, and ranges…
AI · 26d ago · cactuscompute.com
Launched alongside, August 2025
the whole month →
- IS
I built the world's most impractical 1000-pixel display and anyone in the world can draw on it. It draws a single pixel at a time and takes 30-60 minutes to complete a single image. Anyone can participate in the project by voting for the next image to be drawn, and submitting images. https://kilopx.com/
Work · 2025 · benholmen.com
- KT
Kitten TTS is an open-source series of tiny and expressive text-to-speech models for on-device applications. We are excited to launch a preview of our smallest model, which is less than 25 MB. This model has 15M parameters. This release supports English text-to-speech applications in eight voices: four male and four female. The model is quantized to int8 + fp16, and it uses onnx for runtime. The model is designed to run literally anywhere eg. raspberry pi, low-end smartphones, wearables, browsers etc. No GPU required! We're releasing this to give early users a sense of the latency and voices…
Dev tools · 2025 · github.com
- IW
I was wondering how I can arrange objects along a spherical helix path, and read some articles on it. I ended up learning about parametric equations again, and make this visualization to document what I learned: https://visualrambling.space/moving-objects-in-3d/ feel free to visit and let me know what you think!
Life & fun · 2025 · visualrambling.space
- TC
For HTML Day 2025 [1], I made a web service that displays the current sky at your approximate location as a CSS gradient. Colours are simulated on-demand using atmospheric absorption and scattering coefficients. Updates every minute, without the use of client-side JavaScript. Source code and additional information is available on GitHub: https://github.com/dnlzro/horizon [1] https://html.energy/html-day/2025/index.html
Dev tools · 2025 · sky.dlazaro.ca