Anthropic and OpenAI released new AI model offerings and cut token pricing within about an hour of each other
Model release ● Confirmed 78% confidence first seen
Anthropic released Claude Opus 5.5, while OpenAI released two cheaper GPT-6 models, GPT-6 Sol and GPT-6 Luna, shortly afterward. Both vendors introduced pricing reductions for token usage (including cache-read discounts cited for Opus 5.5), intensifying a cost-focused competitive “price war” between the new model releases.
Anthropic
Anthropic PBC
Claude Opus 5.5
GPT-6 Luna
GPT-6 Sol
Luna
OpenAI
OpenAI Group PBC
Sol
GPT-5.6
GPT-6
Grok 4.7
MiMo v2.6 Flash
MiMo v2.6 Pro
SiliconANGLE
Decision brief
- What changed
- Anthropic launched Claude Opus 5.5, and about an hour later OpenAI launched two lower-cost GPT-6 variants, Sol and Luna. Both vendors also reduced token-related pricing, including Anthropic’s cited 60% cache-read cut to $0.20 per million tokens, sharpening direct price competition around inference usage.
- Why it matters
- Leaders buying or embedding foundation models now face a faster-moving cost curve, where model selection decisions may need to be revisited more frequently for common inference-heavy workloads. The combination of new model releases and lower token prices can change unit economics for AI features, margin assumptions, and vendor positioning even when model quality differences are still being evaluated. For organizations already using Anthropic or OpenAI, this creates an immediate procurement and architecture review point rather than a purely technical model-update event.
- Evidence
- All three cited pieces consistently report that Anthropic and OpenAI released new models within roughly an hour of each other and framed the moves as an intensifying price war. SiliconANGLE AI provides the most specific pricing details cited here, while Simon Willison and TLDR independently reinforce the timing, lower-cost positioning, and competitive pricing dynamic.
- What remains uncertain
- The coverage does not establish like-for-like performance, reliability, or total cost of ownership across the new models, so any claim of better business value depends on workload-specific testing. It is also unclear whether these prices are durable, promotional, region-specific, or accompanied by usage limits and other terms that could affect realized savings.
- Monitor next
- Watch for official vendor benchmark disclosures and customer-side workload tests that compare real-world cost and performance between Claude Opus 5.5, GPT-6 Sol, and GPT-6 Luna.
Analytical support, not advice — assumptions and open questions stated above.