Anthropic cuts Opus prices and ships a cheaper coding model
Reviewed by Helen Jones
On 22 September 2026 Anthropic launched Claude Opus 5.5, the first model in its Claude 5.5 line. It says the model matches Claude Fable 5.1 on most work and costs about 40 percent less than Opus 5 on typical billed workloads. API list prices are $4 per million input tokens and $20 per million output tokens, 20 percent below Opus 5. Cache reads are $0.20 per million tokens, 60 percent cheaper. The model has a 1 million token context window and 128,000 max output tokens. Thinking is always on and cannot be turned off. Reuters reported Anthropic claiming Opus 5.5 beat OpenAI GPT-5.6 Sol on a software development benchmark at roughly one third the run cost. Fast mode is $8 and $40 per million tokens. Forced tool_choice types any and tool now return 400 errors.
Until this week, the expensive Opus tier was the default for long coding sessions, big context dumps, and agent loops. Founders paid a premium and could still switch thinking off or force a tool when a workflow needed a cheap, deterministic path. The new economics make those long loops cheaper on paper, especially with cache reads. The catch is that thinking is no longer optional, so token bills can still climb, and any prompt chain that forced a tool will break. Cheaper coding is not a distribution advantage. It is a cheaper way to hide in the product.
Analysis
Treat this as a cost and breakage check, not a reason to rebuild. Pin your current model ID, then run one real workflow (support draft, codebase change, or long prompt chain) against Opus 5.5 and compare billed tokens plus any 400 errors from forced tools before you change production.
Source note
Pulse published by Collab365 Spaces, reviewed by Helen Jones on . Cite as "Anthropic cuts Opus prices and ships a cheaper coding model", Collab365 Spaces. 5 sources referenced.