DeepSeek V4 Flash scores 54.4% on DeepSWE — Alex Getman

Skip to content

Alex Getman @alexgetmancom · 31 Jul 2026, 08:23 UTC · Updated 31 Oct 2026, 20:59 UTC

⚡️ DeepSeek just dropped V4 Flash

It scores 54.4% on DeepSWE and 82.7% on TerminalBench 2.1, beating GLM 5.2 and V4 Pro while coming insanely close to Opus 4.8.

Pricing starts at just $0.28 per 1M input tokens and $0.87 per 1M output. The timing looks like a response to OpenAI cutting Luna and Terra prices.

Table showing DeepSeek V4 Flash benchmark scores: 82.7% on Terminal Bench 2.1 and 54.4% on DeepSWE, outperforming GLM 5.2 and V4 Pro, nearly matching Opus 4.8.

DeepSeek DeepSeek

Русская версия Markdown

Related posts

14 SepA DeepSeek engineer who wrote key compute kernels for V4.1 wrote an essay about what happens when AI starts replacing even the people building it, and what kind of future that could lead to.

DeepSeek DeepSeek

10 Sep🚨 DeepSeek V4.1 Flash is out • New architecture: it only uses the part of the model it actually needs, making it faster and cheaper. • New KV cache: long conversations and agent workflows are noticeably cheaper to run. • On coding and agent tasks, it’s already close to Opus 5…

GPT-5.6 Sol DeepSeek DeepSeek

23 AugTheo made his AI models tier list, so I made mine

Anthropic DeepSeek Google Meta Moonshot AI OpenAI xAI Z.ai Claude DeepSeek Gemini GLM GPT Grok Kimi Muse

17 Aug🚨 DeepSeek slashed OpenCode Go limits 17x DeepSeek V4 Flash limits dropped from 63,300 to 3,800, while V4 Pro fell from 3,450 to 1,050. This comes as DeepSeek API prices have jumped sharply.

DeepSeek DeepSeek OpenCode