AI firms unveil lower-cost GPT-6 and Opus models

OpenAI and Anthropic unveiled new models on Sept. 23, 2026, with companies saying per‑million output token prices now run as low as $0.50 and $20, a step they framed as reducing cloud AI operating costs.

OpenAI expands GPT-6 family

OpenAI provided pricing that it said reflects lower inference costs versus its 5.6 series: Sol at $2 per 1 million input tokens and $10 per 1 million output tokens; Luna at $0.10 per 1 million input tokens and $0.50 per 1 million output tokens. The company said internal tests on real‑world user reports showed materially lower error rates for Sol, reaching Astra‑level reliability at a lower price tier.

Anthropic positioned Claude Opus 5.5 as a mid‑to‑high performer between its Sonnet and Haiku models and said Opus 5.5 matches or beats larger models like Fable on many benchmarks. Anthropic said Opus 5.5’s output token cost is $20 per million versus $25 on the prior model and that throughput and response style were simplified to surface key information earlier in messages.

Watch for initial user feedback and early adoption signals by Sept. 25, 2026, after two days of public availability.

OpenAI said it added GPT-6 Sol and GPT-6 Luna to the GPT-6 line after launching GPT-6 Astra earlier this month. The company said Sol targets complex tasks and coding workflows while Luna is aimed at high‑volume, low‑latency jobs such as document summarization and quick Q&A.