AI Model Price War Pressures Open-Source Providers

The AI model price war intensified in 2026 as OpenAI and Anthropic cut API costs and launched cheaper models. OpenAI reduced GPT-5.6 Luna pricing by 80% to $0.20 per million input tokens and $1.20 per million output tokens. It later introduced GPT-6 Sol at $2/$10 and GPT-6 Luna at $0.10/$0.50 per million input and output tokens. GPT-6 Sol reportedly delivered a 33.2% AutomationBench score at an average task cost of $0.27, while Luna targets high-volume automation and API workloads. Anthropic made Claude Sonnet 5 introductory pricing permanent at $2/$10 and launched Claude Opus 5.5 at $4/$20 per million tokens. Opus 5.5 offers a 1-million-token context window and a 128,000-token output limit. Anthropic says it is more than 30% faster and about 40% cheaper than its predecessor, with strong coding benchmark results, although it remains less suitable for the most complex tasks. DeepSeek V4 Flash reportedly charges about $0.14/$0.28 per million tokens, adding pressure on US model providers. The AI model price war may shift routine enterprise workloads such as classification, summarisation and customer support towards cheaper proprietary models. This could weaken demand for self-hosted open-source AI, although open-source models remain widely used through platforms such as OpenRouter. Lower prices could increase AI adoption and demand for data-centre, cloud and semiconductor infrastructure, but they may also compress margins ahead of potential OpenAI and Anthropic listings. OpenAI research into self-replicating prompt injections was conducted only in simulated environments, with no reported real-world attacks. For crypto traders, the AI model price war has no direct token catalyst. Its immediate cryptocurrency-market impact is likely neutral, with relevance mainly to technology-sector sentiment and AI infrastructure equities.