I Ran Flint, the CRAZY New 40B Coding Model, WITHOUT a Mac. Did It Fix My Buggy Repo?

Ollama says Recursive's new Flint coding model only runs on macOS for now. I ran it anyway, on a rented 48 GB GPU with llama.cpp, and asked it to fix two buggy Python repos with no hints. Result in my tests: 6 of 6 clean runs passed, no test files edited, 15 of 15 hidden checks passed on the harder repo. The limits are in the video: two small repos I wrote, my own agent harness, 4-bit quantization, llama.cpp instead of Ollama. Chapters 0:00 Flint, the Ollama catch, and my test 0:37 What Flint is (model card, MoE, benchmarks) 2:30 Ollama page and the macOS-only note 3:12 Why my laptop can't run it, and picking a RunPod GPU 4:24 Getting Flint running on the pod 5:08 What an agent loop is 5:52 The test repo and the rules 6:49 Run 1: my harness mistake 7:26 Three clean runs 9:02 How it worked (tool-call logs) 10:00 A harder repo, hidden checks 11:54 One run in real time 13:20 What this can and can't tell you 14:23 Try it yourself 15:03 Pod shut down, close Sources (all shown on screen) Model card: https://huggingface.co/RecursiveSI/flint Ollama page: https://ollama.com/library/flint Ollama v0.40.8 release: https://github.com/ollama/ollama/releases/tag/v0.40.8 GGUF files (unsloth): https://huggingface.co/unsloth/flint-GGUF llama.cpp Flint support: https://github.com/ggml-org/llama.cpp/pull/30928 Try it yourself (what I ran) hf download unsloth/flint-GGUF --include "*UD-Q4_K_XL*" --local-dir models cmake -B build -DGGML_CUDA=ON && cmake --build build -j --target llama-server llama-server -m models/flint-UD-Q4_K_XL.gguf -a flint -ngl 99 -c 65536 --jinja --port 8080 Pod: RTX 6000 Ada 48 GB, RunPod secure cloud, $0.99/hr, about 19 minutes, about $0.32 total. If you are looking for cheap and reliable GPUs, I use RunPod: https://get.runpod.io/pe48 (affiliate link) Benchmarks quoted (54.9% Terminal-Bench 2.1, 74.2% SWE-bench Verified, 89.7% SWE-bench Pro v2) are Recursive's own reported numbers, not my tests. The voice and the on-screen avatar are AI-generated from my own voice and likeness. The test repos and logs are mine. Get GPUs Runpod: https://get.runpod.io/pe48 Get Hostinger: https://www.hostg.xyz/SHJiu Coupon Code "PROMPT" 10% discount 🔗 Connect with me: Email → [email protected] Ko-fi → https://ko-fi.com/promptengineer Patreon → https://www.patreon.com/PromptEngineer975 Book a Call → https://calendly.com/prompt-engineer48/call GitHub → https://github.com/PromptEngineer48 Twitter/X → https://twitter.com/prompt48 Website → https://promptengineer48.com/