Breaking

Anthropic Releases Claude Opus 5.5, Intensifying the Coding-Agent Fight With OpenAI

September 22, 2026 at 23:00 EDT

Foundation Models
AI Agents
Software Dev & Coding

Anthropic released Claude Opus 5.5 on September 22, positioning the model as a cheaper, faster and stronger option for long-running coding agents and professional knowledge work. The launch landed the same day OpenAI expanded its GPT-6 lineup with lower-cost GPT-6 Sol and GPT-6 Luna, sharpening the contest around useful work completed per dollar rather than benchmark rank alone.

Sep. 22, 2026 · Anthropic and OpenAI

Coding Agents Enter a Price Fight

Claude Opus 5.5 pushes higher coding-agent scores and lower Opus costs, while GPT-6 Sol and Luna answer with sharply cheaper API tiers.

66.4%

Opus 5.5 on Terminal-Bench 4.0

40%

Typical Opus workload cost drop

Token context window

50%

GPT-6 Sol and Luna API price cut

Terminal-Bench 4.0: coding-agent accuracy

66.4%

Opus 5.5

57.9%

GPT-6 Astra

37.3%

GPT-5.6 Sol

The strongest visible gap is in long-running terminal coding tasks; other benchmarks are more mixed.

Why builders are interested

Opus 5.5 promises fewer tokens, fewer steps and stronger long-horizon coding results, with better cost per completed task than Opus 5.

Why buyers are cautious

The cleanest comparison is still missing: Opus 5.5 and GPT-6 Sol have not been published across one shared public harness and identical settings.

The new model race

Higher coding scores→Lower task cost→Cheaper GPT-6 tiers→Cost per finished workflow

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.