Cline says DeepSeek V4.1-Flash-1017 has been released and scores above GPT-5.6 Sol on Terminal-Bench while costing about 17 times less.
The post says the previous V4.1-Flash result improved by 35% with no algorithmic innovation, through RL training on multiple coding agent harnesses.
The attached chart, sourced to DeepSeek and vendor price lists, shows V4.1-Flash-1017 at 42.3 Pass@1 versus 39.9 for GPT-5.6 Sol, priced at $0.30/$1.20 per 1M tokens versus $4/$20.
DeepSeek V4.1-Flash-1017 has released, beating GPT 5.6 Sol on Terminal-Bench while being ~17x cheaper.
It improves the previous DeepSeek V4.1-Flash result by 35%, accomplished with "no algorithmic innovation" but instead by RL training it on multiple coding agent harnesses.
This story is featured in top picks: “DeepSeek upgrades V4.1 Flash to DeepSeek-V4.1-Flash-1017”
Source: Cline · x.comPublished · added here
