DeepSeek V4-Flash brings frontier agent work to bargain pricing
DeepSeek ● Covered by 6 sources
DeepSeek released V4-Flash, an API model with enhanced agent capabilities that scores 82.7 on Terminal Bench 2.1 and 54.2 on NL2Repo, surpassing the V4-Pro-Preview version. The model maintains the same architecture as the preview version but was re-post-trained to improve performance. Developers can now access stronger agent capabilities at the lower pricing tier by switching their API model parameter to deepseek-v4-flash.
Why it matters
DeepSeek upgraded V4-Flash into a far stronger coding and agent model while keeping the same bargain API rates at $0.28 per million output tokens. The model scored 50 on Artificial Analysis' Intelligence Index while remaining dramatically cheaper than frontier alternatives.
Related stories
DeepSeek-V4-Flash Outshines Pro, The Biggest GitHub Crawl Yet, Engineering System Prompts for Safer Code
The Batch ·
43
DeepSeek debuts multimodal language model competitive with Opus 4.8
SiliconANGLE · 3 weeks ago ·
34