EasyCommand: open models, training data, and lessons from teaching small models to write shell commands.
Teaching a 1.5B model to generate bash commands at gpt-4o level using 400k synthetic examples — Dirac Skip to content DIRAC Field manual Benchmarks Docs Posts Labs Market Share Discord View on GitHub Toggle navigation Posts / EasyCommand On this page Summary How it started The good stuff The training loop More examples ≠ better Debugging the benchmark Try it, or keep training Field report Teaching a 1.5B model to generate bash commands at gpt-4o level using 400k synthetic examples October 05, 2026 Small-model training 5 min read Max Trivedi Tl;Dr: Fine-tuned Qwen2.5-Coder-1.5B-Instruct and Qwe
EasyCommand: open models, training data, and lessons from teaching small models to write shell commands.
Teaching a 1.5B model to generate bash commands at gpt-4o level using 400k synthetic examples — Dirac Skip to content DIRAC Field manual Benchmarks Docs Posts Labs Market Share Discord View on GitHub Toggle navigation Posts / EasyCommand On this page Summary How it started The good stuff The training loop More examples ≠ better Debugging the benchmark Try it, or keep training Field report Teaching a 1.5B model to generate bash commands at gpt-4o level using 400k synthetic examples October 05, 2026 Small-model training 5 min read Max Trivedi Tl;Dr: Fine-tuned Qwen2.5-Coder-1.5B-Instruct and Qwe
