DeepSeek-V4 Flash 0731 vs GPT-5.6 Luna on DeepSWE: Cost and Coding
Together AI BlogAug 64 min read
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.
Read the full story at Together AI Blog →
More in AI
TN
Meta lets Claude and Codex configure WhatsApp Business via MCP
AIThe New Stack AIPaul Sawers3h
R
W
Inside OpenAI’s Ad Rush
AIWSJ3h
RR
ARR review
AIReddit r/LanguageTechnology/u/Hedgehog_Dapper3h
F
FT
OpenAI weighs funding round at $1.2tn valuation before IPO
AIFinancial Times3h