Over the past few months, OpenAI and Anthropic have repeatedly cut LLM pricing—up to 80% in July for GPT-5.6 Luna (from $1/$6/$0.10 to $0.20/$1.20/$0.02) and ~20% for GPT-5.6 Sol ($5/$30/$0.50 to $4/$20/$0.40)—in a tit-for-tat response. Anthropic followed with Claude Sonnet 5 at $2/$10/$0.20 (later made permanent), Opus 5 at $5/$25/$0.50, and Opus 5.5 at $4/$20/$0.20 (a 40–60% total cost reduction versus Opus 5), while Fable 5.1 cut cache read costs by 75% (reducing effective cost 70–75%). OpenAI countered with GPT-6 Sol at $2/$10/$0.20 and GPT-6 Luna at $0.10/$0.50/$0.01, then GPT-6.1 Sol at $2/$10/$0.10, making it cumulatively 75% cheaper than prior families. The drivers include open-weight competition (DeepSeek Flash V4.1 at $0.15/$0.60/$0.003 across 29 inference providers on OpenRouter; Kimi K3.1 Pro at $2/$10/$0.20 from Moonshot AI), commoditization of model choice via harnesses like Codex and Claude Code, and improved compute efficiency. Reports suggest profitable inference margins of ~80%, though subscription tiers saw usage allowance cuts (e.g., Codex $200/month tier reduced by 50%), complicating the narrative around IPO pressure. Ultimately, price declines reflect genuine market dynamics: low switching costs, competitive inference providers, and compute efficiency gains, rather than pure loss-leading.