Know what changed before it reaches production.

Google made Gemini 4 Argon generally available on October 13, 2026 at an introductory $2 / $10 per million tokens — a rate that becomes $4 / $20 on December 1, with cached input moving from $0.10 to $0.20. The model also ships a different request contract: temperature, top_p and top_k are accepted and ignored, thinking_budget returns a 400, explicit context caching is gone, and one 1,048,576-token window covers input and output. From November 13, a long completion that times out is still billable.

Anthropic released Claude Haiku 5.5 on October 7, 2026 at $0.10 / $0.50 per million tokens — but with the newer Claude 4.7+ tokenizer, so the same text is roughly 30% more tokens than on Claude Haiku 4.5, and prompts over 100,000 tokens step to a rate five times the base. The same day it halved Claude Sonnet 5.5 cache reads to $0.10. OpenAI added Ultrafast mode to GPT-6.1 Sol with EU data residency and shipped the Decisions API, a /v1/decisions beta surface that bills input tokens only. The cheap end of the stack is where routing defaults are actually set — and it moved in a single week.

Anthropic released Claude Sonnet 5.5 on September 28, 2026 at $2 / $10 per million tokens with five breaking changes, and put Claude Sonnet 4.5 on a November 30, 2026 retirement clock. OpenAI released GPT-6.1 Sol on September 29 at $2 / $10 with cache reads cut to $0.10 and a $2.50 cache-write line. The workhorse tier that coding and agent traffic routes to is now price-symmetric — but the request contracts and cache economics are not.

OpenAI released GPT-6 Sol at $2.00 / $10.00 and GPT-6 Luna at $0.10 / $0.50 per million tokens on September 22, 2026, and Anthropic released Claude Opus 5.5 the same day at $4 / $20 — below Claude Opus 5's $5 / $25. Azure's retail meter API already prices the new 6-sol meters while Microsoft's own pricing page still lists GPT-6 Sol and Luna as "in processing".

OpenAI documents GPT-5.6 Sol's promotional rate — $4 / $20 per million tokens — as available "at least through November 21, 2026". Azure's published meters price the same model at $4 / $20 Global and $4.40 / $22 in a Data Zone, while the registry's rate-card source still carries $5 / $30 for the gpt-5.6 alias.

OpenAI shipped a voice endpoint billed per minute and a managed agent harness billed per container on September 10, 2026. Neither is token-metered — and neither fits a gateway route most teams already proxy.

Claude Fable 5.1, Gemini 3.8 Flash and GPT-6 Astra all shipped between September 1 and 3, 2026. What each launch changes for routing, fallbacks and the registry snapshot.

Which Azure AI model versions actually retire in October 2026 — and where the live registry currently disagrees with Microsoft's schedule.

A live watchlist of the provider endpoints ALLM has observed in the last seven days.

See concrete deployment retirement dates before they become a migration incident.

A current, deployment-level answer instead of a model-family assumption.

Compare one shared canonical model across the direct and Azure AI surfaces.