Google Gemini 3.8 Leads DeepSWE Benchmark at Low Cost

This is very interesting. New release from Google DeepMind, Gemini 3.8 Flash now leads the DeepSWE benchmark (widely considered the most reliable benchmark for measuring a model's coding capability) at a very low cost (~5x lower than Opus 5, ~10x lower than Fable 5, and ~3x lower than GPT 5.6 Sol). Has Google really cooked with this new release, or have they benchmaxxed?

Even if the model behaviour aligns with the benchmark score, Google still has a long way to go in terms of token efficiency or the number of steps to complete a task.