DeepSeek V4.1-Flash-1017 benchmark breakdown: leads DeepSWE, trails Opus 5 on Terminal-Bench
Item-by-item: DeepSWE 77.0 versus 74.2 prior, Sol 73.0 and Opus 5 74.0; Terminal-Bench 4.0 42.3 versus 31.2 prior, Sol 39.9 and Opus 5 51.8; Automation-Bench 60.5 versus 54.8 prior, Sol 45.8 and Opus 5 50.3.
Pretty impressive: DeepSeek V4.1 Flash-1017 pulls ahead of GPT-5.6 Sol and Opus 5 on some coding and agent benchmarks, with much lower token prices. Open weights, an MIT license, and a 1M-token context window. Here’s the breakdown: - DeepSWE: 77.0, up from 74.2 for the previous Flash. GPT-5.6 Sol scores 73.0; Opus 5 scores 74.0. - Terminal-Bench 4.0: 42.3, up from 31.2. Ahead of GPT-5.6 Sol at 39.9, but behind Opus 5 at 51.8. - Automation-Bench: 60.5, up from 54.8. GPT-5.6 Sol scores 45.8; Opus 5 scores 50.3.
This post quotes another post: x.com/deepseek_ai/status/2111375640717033777