OpenAI and Anthropic Cut Model Prices as Efficiency Race Intensifies

Ringg uses GPT-5.6 to power multilingual customer-service agents across voice, chat, WhatsApp, and the web, and OpenAI says those agents resolve up to 65% of customer calls. [1] OpenAI also says Ringg operates the GPT-5.6 system at 90% less cost than GPT-4.1, with the deployment covering all four listed customer-contact channels and automating as many as 65% of calls. [1]

OpenAI released GPT-6 Sol and GPT-6 Luna with per-token costs 50% below their GPT-5.6 predecessors, attributing the reduction to improvements in caching and inference. [2] Anthropic launched Claude Opus 5.5 with token prices 20% below Opus 5 and said it performs at Claude Fable 5.1’s level on most work while costing 40% less to run than Opus 5. [2] Forrester analyst Charlie Dai described a prolonged price-performance race driven by inference efficiency, caching, model optimization, competitive pressure, and converging capabilities, while noting that sovereignty, security, and intellectual-property requirements still support private or hybrid deployments. [2]

Taken together, the reports show cost efficiency operating at two layers: model vendors are lowering token prices, while an application provider reports lower operating costs and substantial automated call resolution. [1][2] OpenAI’s GPT-6 Sol and Luna use a 50% API-price cut against GPT-5.6, while Anthropic positions Opus 5.5 through a 20% token-price reduction and a 40% run-cost comparison with Opus 5. What the sources don't address: whether these lower prices will translate into comparable total-cost savings across enterprise workloads or how quickly customers will shift production traffic between the competing model families.