Anthropic and OpenAI release new models (Claude Opus 5.5, GPT-6 Sol, GPT-6 Luna) with steep price cuts, intensifying an AI model price war

Anthropic released Claude Opus 5.5 with a price cut to $4/M input and $20/M output tokens (down from $5/$25), plus a 60% cut to cached input pricing. On the same day, OpenAI released GPT-6 Sol and GPT-6 Luna, both priced at roughly half their GPT-5.6 predecessors, with GPT-6 Luna coming in at $0.10/M input and $0.50/M output. This is reported by Simon Willison based on vendor announcements and his own testing, not linked pricing pages, so treat the exact figures as needmassle secondary reporting rather than primary-sourced. Willison also notes Claude Opus 5.5 at maximum thinking level failed his standard pelican-SVG test twice, hitting the 128,000 output token limit while still reasoning.

Anyone routing API workloads across these vendors gets an immediate, concrete reason to re-run cost comparisons, particularly for agentic or high-cache-read workloads where the 60% cache cut matters most. The pricing convergence across Claude Opus 5.5, GPT-6 Sol/Luna and Grok 4.7 suggests the mid-tier model market is being squeezed on price, which could pressure smaller or higher-priced offerings like Haiku. That said, this is one commentator's same-day read; it says nothing yet about actual usage shifts, retention, or whether cheaper pricing changes anyone's vendor choice in practice.

The framing of this as "hard to overstate how competitive" the market is is analyst flourish, not evidence; a price cut announcement is not proof of a structural shift until usage data follows. The failed pelican-SVG test is a single anecdote about one prompt on one model configuration, not a reliable capability signal, and should not be read as evidence Opus 5.5 is worse than its predecessor. No primary sources (vendor pricing pages, official blog posts) are linked, so the specific dollar figures should be verified independently before being used in any budgeting decision.

1. 01Check Anthropic's and OpenAI's official pricing pages directly to confirm the $4/$20 and $0.10/$0.50 figures cited here
2. 02Watch for any usage or market-share data over the next one to two months showing whether customers actually shift spend toward the cheaper models
3. 03Track whether Anthropic discontinues or re-prices Haiku in response to mid-tier price pressure, which would confirm the squeeze narrative