# DeepSeek-V4 Flash 0731 Vs GPT-5.6 Luna on DeepSWE: Cost and Coding

> **Platform:** [SPIDITS AI](https://spidits.com/) — Real-Time AI News & Market Intelligence  
> **Published:** 2026-08-06T00:00:00.000Z  
> **Category:** PRODUCT_LAUNCH  
> **Impact Score:** 130/100  
> **Primary Source:** [Together AI Blog](https://www.together.ai/blog/deepseek-v4-flash-0731-vs-gpt-5-6-luna-on-deepswe-cost-and-coding)  
> **Canonical Citation:** [https://spidits.com/timeline/deepseek-v4-flash-0731-vs-gpt-5-6-luna-on-deepswe-cost-and-coding](https://spidits.com/timeline/deepseek-v4-flash-0731-vs-gpt-5-6-luna-on-deepswe-cost-and-coding)

## Executive Summary
We ran 900 DeepSWE rollouts on DeepSeek-V4 Flash and GPT-5.6 Luna. Luna leads pass@1 by 14 points; DeepSeek delivers 4.8x the solves per dollar.

## Why It Matters (Strategic Analysis)
Substantially cheaper open-weight models allow multi-candidate sampling cascades that exceed flagship single-shot performance at lower total inference costs. Moreover, reduced regression rates in low-cost models lower the verification overhead required for enterprise automated software engineering pipelines.

## Key Entities & Companies
- **OpenAI**

## Referenced Coverage & Sources
- **[Together AI Blog](https://www.together.ai/blog/deepseek-v4-flash-0731-vs-gpt-5-6-luna-on-deepswe-cost-and-coding)**: Untitled — _We ran 900 DeepSWE rollouts on DeepSeek-V4 Flash and GPT-5.6 Luna. Luna leads pass@1 by 14 points; DeepSeek delivers 4.8x the solves per dollar._

---
*Synthesized by SPIDITS AI Market Intelligence Desk. Track live AI news, model releases, and funding: [https://spidits.com](https://spidits.com)*