Changelog

Major changes to AI Foundation Services — new models, feature updates, and deprecations.

v1.2.2 — 2026-09

Aligns the documentation with Service Description v1.23 (31 August 2026), which is the binding source for everything below.

Added

New models on the Essential, Professional and Agentic plans: GPT-5.6 Terra (gpt-5.6-terra) at €3.52 / €15.82 per million input / output tokens, and GPT-5.6 Luna (gpt-5.6-luna) at €0.20 / €1.20. Both are served from Azure Sweden with a 1050K context window.
Gemini 3.1 Pro is now priced by context length, matching the Service Description: ≤200k input tokens at €1.80 / €10.80, >200k at €3.60 / €16.20. Previously a single row.
Cached input token pricing. Prompt tokens served from the prefix cache are billed at 15% of the model’s normal input token price on T-Cloud-hosted models. The per-model price appears in the new Cached Input Token Price column on each rate plan page and the Cached €/M column of the models comparison. Cached-token pricing for models on external providers (Azure, GCP) will be added later. See Prefix Caching.

Changed

The availability target is 99.5%, not the 99.9% previously published. Corrected on Service Levels, Rate Limits, Dedicated LLM Serving and Shared vs Dedicated.
GLM 5.2 token prices reduced on all plans: input €4.00 → €1.50, output €8.00 → €3.50 per million tokens. Cached input tokens are therefore €0.225 per million.
GLM 5.2 moved from the premium to the standard model tier on all plans.
GPT-Image-2 is €4.39 / €26.33 per million tokens (previously published as €9.55 / €38.20).
Prefix caching is on by default on T-Cloud-hosted models and can be switched off per request with save_cache: false. The guide previously described caching as opt-in.
Server location and data processing for the Gemini 3 family — Gemini 3 Pro (≤200k and >200k), Gemini 3 Flash, Gemini 3.5 Flash, Gemini 3.1 Pro and Gemini 3 Pro Image — now read Worldwide instead of Europe. Google runs these models on its own infrastructure and may process data outside the EU/EEA; see Enterprise Trust.
The Anthropic models are labelled GCP / Azure, since a request may be served from either cloud.
The Service Description shipped with the docs is now v1.23, replacing v1.20. It is published in German only, so both language versions link to the same document. Readers on an archived docs version still get the v1.20 document that version was written against.
Enterprise Trust category 1 now lists the models the Service Description names — Gemma 4 31B, Mistral Small 4, Qwen 3.6 35B, Nemotron 3 Super, GLM 5.2, GPT-OSS-120B, Whisper, and the embedding models.

Removed

Llama 3.3 70B, Claude 4.5 Sonnet (both context tiers) and Claude 4.5 Opus reached their 2026-09-01 retirement date and have left the catalog. See Retired Models for replacements.

Fixed

OpenCode setup: corrected the authentication URL in step 2 of the OpenCode plugin instructions to https://plugins.ai.t-systems.net/external/.
/category/model-serving, the API documentation URL printed in § 2.2 of the Service Description, now redirects to the guides instead of returning a 404.

v1.2.1 — 2026-08

Added

New preview model in the Test catalog: Qwen 3.8 27B FP8 (Qwen3.8-27B-FP8-Preview) from Alibaba Cloud — a natively multimodal model (text and image input) with a 256K context window, Apache-2.0 licensed. Free to try on any active plan (Essential / Professional / Agentic). See rate limits for the preview-model policy.

v1.2.0 — 2026-07

Added

New models across the Essential, Professional, and Agentic plans: GPT-5.4, GPT-5.4 mini, GPT-5.5, and GPT-Image-2 (OpenAI, via Azure); Gemini 3.1 Pro and Gemini 3.5 Flash (Google). Claude 4.8 Opus (Anthropic) is available on Professional and Agentic. See Plans & Pricing for per-plan rate limits and pricing.
Mistral Small 4 (Mistral-Small-4-119B-2603) and Gemma 4 (gemma-4-31B-it-FP8) are now generally available (GA) on the Essential, Professional, and Agentic plans, with fixed RPM/TPM limits and standard pricing. They were previously preview-only in the Test catalog. See Plans & Pricing for per-plan rate limits and pricing.
New preview model in the Test catalog: GLM 5.2 (GLM-5.2-Preview) from Zhipu AI — free to try on any active plan (Essential / Professional / Agentic). See rate limits for the preview-model policy.

Deprecated

The following models are scheduled for retirement on 2026-08-01. If you use any of them, migrate to the suggested replacement before that date. See Scheduled for Retirement for the full table.

Mistral Small 3 → Mistral Small 4
Qwen3-VL, Qwen3 Coder, Qwen3-Next → Qwen3.6 35B
GPT-4.1 / GPT-4.1 mini / GPT-4.1 nano, GPT-5 mini → GPT-5.4 mini
GPT-4o / GPT-4o mini, GPT-5 / GPT-5.2, o1 / o1-mini / o3 / o3-mini / o4-mini → GPT-5.4
GPT Image 1 → GPT-Image-2
Claude 3.7 Sonnet, Claude 4 Sonnet → Claude 4.6 Sonnet
Claude 4.5 Opus / Claude 4.6 Opus → Claude 4.8 Opus
Gemini 2.5 Flash / Flash Image / Pro → Gemini 3.5 Flash / Gemini 3 Pro Image / Gemini 3.1 Pro
Gemini 3 Pro / Gemini 3 Flash → Gemini 3.1 Pro / Gemini 3.5 Flash

v1.1.0 — 2026-05

Archived snapshot: v1.1.0.

Added

New top-level Models section with a comparison page covering every hosted model — context window, pricing, cloud, modalities, plan availability, and OpenAI-compatible endpoint reference. Image-capable models also surface on the dedicated Vision page via a capability filter.
Changelog page (you are reading it).
Two new preview models in the Test catalog: Mistral Small 4 (Mistral-Small-4-119B-2603-Preview) and Gemma 4 (gemma-4-31B-it-FP8-preview). See rate limits for the preview-model policy.

Removed

Fine-Tuning API documentation. Requests to /guides/fine-tuning now redirect to the optional services. Superseded: fine-tuning is offered as an optional service on request — see § 3.4.2 of the Service Description. The self-service API documentation remains withdrawn.

v1.0.0

Archived snapshot: v1.0.0 — the initial released state of the docs prior to the changelog being introduced.