Field notes — week of September 21, 2026

Anthropic introduced Claude Opus 5.5 on Sept 22, 2026, a flagship model it says performs at the level of Claude Fable 5.1 on most work while costing 40% less to run than Opus 5 at default settings. OpenAI released the reasoning models GPT-6 Sol and GPT-6 Luna the same day through its Responses and Chat Completions APIs, priced at $2/$0.20/$10 and $0.10/$0.01/$0.50 per million input, cached-input and output tokens respectively. DeepL also became available inside Microsoft Copilot, ChatGPT and Claude on Sept 22 via a Model Context Protocol integration that applies a company's glossary and style rules without leaving the assistant.

Field notes — week of September 21, 2026 What changed in AI-support and AI-localization tooling this week. Every item was checked against its source before it entered this list. - Anthropic - introduced Claude Opus 5.5 https://www.anthropic.com/news/claude-opus-5-5 Sept 22 , a new flagship model that Anthropic says performs at the level of Claude Fable 5.1 on most work while costing 40% less to run than Opus 5 at default settings. - OpenAI - released GPT-6 Sol and GPT-6 Luna https://developers.openai.com/api/docs/changelog Sept 22 , two new reasoning models available through the Responses and Chat Completions APIs; pricing is $2 / $0.20 / $10 per million input / cached-input / output tokens for Sol and $0.10 / $0.01 / $0.50 for Luna. - Sierra - published Your agent, laid bare https://sierra.ai/blog/your-agent-laid-bare-and-why-it-matters Sept 22 , arguing for agent platforms that expose their logic and keep a complete, exportable record of agent actions rather than acting as an opaque black box. - Decagon - is expanding into the Benelux region https://decagon.ai/blog/growing-in-benelux Sept 21 , building a local team across Belgium, the Netherlands and Luxembourg; the post cites Rituals, an Amsterdam beauty brand, deploying Decagon's "Ray" concierge across 19 countries and 15 languages within two months, handling up to 9,000 customer cases a day at peak. - Intercom - two changes this week: custom metrics in reporting https://www.intercom.com/changes/en/158136-build-the-metric-you-need-across-your-full-history Sept 23 let teams combine built-in measurements and filters into their own metrics; call termination reasons https://www.intercom.com/changes/en/157743-see-why-an-outbound-call-didn-t-connect Sept 21 now show why an outbound call failed to connect. - DeepL - is now available inside Microsoft Copilot, ChatGPT and Claude https://www.deepl.com/en/blog/deepl-available-in-microsoft-copilot-chatgpt-claude-mcp Sept 22 through a Model Context Protocol integration that translates text and documents and applies a company's glossary and style rules without leaving the assistant.