GPT-6 Sol vs Opus 5.5: OpenAI and Anthropic cut prices the same day

OpenAI and Anthropic cut prices on the same Tuesday. OpenAI's new GPT-6 Sol costs $2/$10 per million input/output tokens, half of GPT-5.6 Sol's $4/$20. The smaller GPT-6 Luna costs $0.10/$0.50. Sol is aimed at complex coding and agentic work. OpenAI says the prices are permanent, not an intro offer. Both models are live in ChatGPT and Codex, and Free users get Luna in the desktop app.

Anthropic answered with Claude Opus 5.5, which it says matches Fable 5.1 on most tasks. List prices fell 20% to $4/$20. The bigger saving is in agentic work: cache reads are 60% cheaper and the model uses fewer tokens per task, so Anthropic says it costs 40% less to run than Opus 5. On Anthropic's own benchmarks it leads on coding tests such as Terminal-Bench 4.0 (66.4%) and CursorBench 4.0 (57.8%). GPT-6 Astra still beats it on AutomationBench.

The cheaper tiers now come close to flagship quality, so a model's cost per task matters as much as its benchmark scores. Teams running agents at scale just saw their bills drop by almost half overnight.

Why it matters: frontier-level coding and agents just got roughly half as expensive, which makes agent products that were too costly to run viable now.