From 0ed4489abee139bd8fd39216807a82727e74897a Mon Sep 17 00:00:00 2001
From: Florian Heuer <florian.heuer@digits.schwarz>
Date: Fri, 5 Jun 2026 17:48:09 +0200
Subject: [PATCH 1/6] chore: remove deprecated neuralmagic models

---
 .../Meta-Llama-3.1-8B-Instruct-FP8.toml       | 22 -------------------
 .../Mistral-Nemo-Instruct-2407-FP8.toml       | 22 -------------------
 2 files changed, 44 deletions(-)
 delete mode 100644 providers/stackit/models/neuralmagic/Meta-Llama-3.1-8B-Instruct-FP8.toml
 delete mode 100644 providers/stackit/models/neuralmagic/Mistral-Nemo-Instruct-2407-FP8.toml

diff --git a/providers/stackit/models/neuralmagic/Meta-Llama-3.1-8B-Instruct-FP8.toml b/providers/stackit/models/neuralmagic/Meta-Llama-3.1-8B-Instruct-FP8.toml
deleted file mode 100644
index bdfbda001..000000000
--- a/providers/stackit/models/neuralmagic/Meta-Llama-3.1-8B-Instruct-FP8.toml
+++ /dev/null
@@ -1,22 +0,0 @@
-name = "Llama 3.1 8B"
-family = "llama"
-release_date = "2024-07-23"
-last_updated = "2024-07-23"
-attachment = false
-reasoning = false
-temperature = true
-tool_call = true
-structured_output = true
-open_weights = true
-
-[cost]
-input = 0.16
-output = 0.27
-
-[limit]
-context = 128_000
-output = 8_192
-
-[modalities]
-input = ["text"]
-output = ["text"]
diff --git a/providers/stackit/models/neuralmagic/Mistral-Nemo-Instruct-2407-FP8.toml b/providers/stackit/models/neuralmagic/Mistral-Nemo-Instruct-2407-FP8.toml
deleted file mode 100644
index 2f35759aa..000000000
--- a/providers/stackit/models/neuralmagic/Mistral-Nemo-Instruct-2407-FP8.toml
+++ /dev/null
@@ -1,22 +0,0 @@
-name = "Mistral Nemo"
-family = "mistral"
-release_date = "2024-07-01"
-last_updated = "2024-07-01"
-attachment = false
-reasoning = false
-temperature = true
-tool_call = true
-structured_output = false
-open_weights = true
-
-[cost]
-input = 0.49
-output = 0.71
-
-[limit]
-context = 128_000
-output = 8_192
-
-[modalities]
-input = ["text"]
-output = ["text"]

From cb4884c3ec849fa891d8495f8969946af1989e19 Mon Sep 17 00:00:00 2001
From: Florian Heuer <florian.heuer@digits.schwarz>
Date: Fri, 5 Jun 2026 18:13:31 +0200
Subject: [PATCH 2/6] chore: adjust prices (conversion rate of 6m avg used
 (1.17))

---
 .../stackit/models/Qwen/Qwen3-VL-235B-A22B-Instruct-FP8.toml  | 4 ++--
 .../models/cortecs/Llama-3.3-70B-Instruct-FP8-Dynamic.toml    | 4 ++--
 providers/stackit/models/google/gemma-3-27b-it.toml           | 4 ++--
 providers/stackit/models/openai/gpt-oss-120b.toml             | 4 ++--
 4 files changed, 8 insertions(+), 8 deletions(-)

diff --git a/providers/stackit/models/Qwen/Qwen3-VL-235B-A22B-Instruct-FP8.toml b/providers/stackit/models/Qwen/Qwen3-VL-235B-A22B-Instruct-FP8.toml
index 4df523123..5e28c8908 100644
--- a/providers/stackit/models/Qwen/Qwen3-VL-235B-A22B-Instruct-FP8.toml
+++ b/providers/stackit/models/Qwen/Qwen3-VL-235B-A22B-Instruct-FP8.toml
@@ -10,8 +10,8 @@ structured_output = false
 open_weights = true
 
 [cost]
-input = 1.64
-output = 1.91
+input = 1.76
+output = 2.05
 
 [limit]
 context = 218_000
diff --git a/providers/stackit/models/cortecs/Llama-3.3-70B-Instruct-FP8-Dynamic.toml b/providers/stackit/models/cortecs/Llama-3.3-70B-Instruct-FP8-Dynamic.toml
index 468f3da31..c9ac96651 100644
--- a/providers/stackit/models/cortecs/Llama-3.3-70B-Instruct-FP8-Dynamic.toml
+++ b/providers/stackit/models/cortecs/Llama-3.3-70B-Instruct-FP8-Dynamic.toml
@@ -10,8 +10,8 @@ structured_output = false
 open_weights = true
 
 [cost]
-input = 0.49
-output = 0.71
+input = 0.53
+output = 0.76
 
 [limit]
 context = 128_000
diff --git a/providers/stackit/models/google/gemma-3-27b-it.toml b/providers/stackit/models/google/gemma-3-27b-it.toml
index c25df2ab6..7d23a1cc5 100644
--- a/providers/stackit/models/google/gemma-3-27b-it.toml
+++ b/providers/stackit/models/google/gemma-3-27b-it.toml
@@ -10,8 +10,8 @@ structured_output = false
 open_weights = true
 
 [cost]
-input = 0.49
-output = 0.71
+input = 0.53
+output = 0.76
 
 [limit]
 context = 37_000
diff --git a/providers/stackit/models/openai/gpt-oss-120b.toml b/providers/stackit/models/openai/gpt-oss-120b.toml
index 33acefc69..f7da5ffbe 100644
--- a/providers/stackit/models/openai/gpt-oss-120b.toml
+++ b/providers/stackit/models/openai/gpt-oss-120b.toml
@@ -10,8 +10,8 @@ structured_output = false
 open_weights = true
 
 [cost]
-input = 0.49
-output = 0.71
+input = 0.53
+output = 0.76
 
 [limit]
 context = 131_000

From 19edd83bad07783ee3877ad44311f70cc5852747 Mon Sep 17 00:00:00 2001
From: Florian Heuer <florian.heuer@digits.schwarz>
Date: Fri, 5 Jun 2026 20:02:06 +0200
Subject: [PATCH 3/6] chore: adjust limits according to specs

---
 .../stackit/models/Qwen/Qwen3-VL-235B-A22B-Instruct-FP8.toml    | 2 +-
 .../models/cortecs/Llama-3.3-70B-Instruct-FP8-Dynamic.toml      | 2 +-
 providers/stackit/models/google/gemma-3-27b-it.toml             | 2 +-
 3 files changed, 3 insertions(+), 3 deletions(-)

diff --git a/providers/stackit/models/Qwen/Qwen3-VL-235B-A22B-Instruct-FP8.toml b/providers/stackit/models/Qwen/Qwen3-VL-235B-A22B-Instruct-FP8.toml
index 5e28c8908..80144afa5 100644
--- a/providers/stackit/models/Qwen/Qwen3-VL-235B-A22B-Instruct-FP8.toml
+++ b/providers/stackit/models/Qwen/Qwen3-VL-235B-A22B-Instruct-FP8.toml
@@ -15,7 +15,7 @@ output = 2.05
 
 [limit]
 context = 218_000
-output = 8_192
+output = 16_384
 
 [modalities]
 input = ["text", "image"]
diff --git a/providers/stackit/models/cortecs/Llama-3.3-70B-Instruct-FP8-Dynamic.toml b/providers/stackit/models/cortecs/Llama-3.3-70B-Instruct-FP8-Dynamic.toml
index c9ac96651..cd31de457 100644
--- a/providers/stackit/models/cortecs/Llama-3.3-70B-Instruct-FP8-Dynamic.toml
+++ b/providers/stackit/models/cortecs/Llama-3.3-70B-Instruct-FP8-Dynamic.toml
@@ -15,7 +15,7 @@ output = 0.76
 
 [limit]
 context = 128_000
-output = 8_192
+output = 4_096
 
 [modalities]
 input = ["text"]
diff --git a/providers/stackit/models/google/gemma-3-27b-it.toml b/providers/stackit/models/google/gemma-3-27b-it.toml
index 7d23a1cc5..48c83c263 100644
--- a/providers/stackit/models/google/gemma-3-27b-it.toml
+++ b/providers/stackit/models/google/gemma-3-27b-it.toml
@@ -15,7 +15,7 @@ output = 0.76
 
 [limit]
 context = 37_000
-output = 8_192
+output = 4_096
 
 [modalities]
 input = ["text", "image"]

From 8234ea4855d9db0d02c986788331ee9c41febdad Mon Sep 17 00:00:00 2001
From: Florian Heuer <florian.heuer@digits.schwarz>
Date: Fri, 5 Jun 2026 20:02:44 +0200
Subject: [PATCH 4/6] feat: add model gpt-oss-20b to provider stackit

---
 .../stackit/models/openai/gpt-oss-20b.toml    | 22 +++++++++++++++++++
 1 file changed, 22 insertions(+)
 create mode 100644 providers/stackit/models/openai/gpt-oss-20b.toml

diff --git a/providers/stackit/models/openai/gpt-oss-20b.toml b/providers/stackit/models/openai/gpt-oss-20b.toml
new file mode 100644
index 000000000..ea765b84d
--- /dev/null
+++ b/providers/stackit/models/openai/gpt-oss-20b.toml
@@ -0,0 +1,22 @@
+name = "GPT-OSS 20B"
+family = "gpt"
+release_date = "2025-08-05"
+last_updated = "2025-08-05"
+attachment = false
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = false
+open_weights = true
+
+[cost]
+input = 0.18
+output = 0.29
+
+[limit]
+context = 131_000
+output = 8_192
+
+[modalities]
+input = ["text"]
+output = ["text"]

From a75c38133a27f9df01f5a181ef620a636d64d94d Mon Sep 17 00:00:00 2001
From: Florian Heuer <florian.heuer@digits.schwarz>
Date: Fri, 5 Jun 2026 20:03:16 +0200
Subject: [PATCH 5/6] feat: add model qwen3.6-27b to provider stackit

---
 .../stackit/models/Qwen/Qwen3.6-27B.toml      | 39 +++++++++++++++++++
 1 file changed, 39 insertions(+)
 create mode 100644 providers/stackit/models/Qwen/Qwen3.6-27B.toml

diff --git a/providers/stackit/models/Qwen/Qwen3.6-27B.toml b/providers/stackit/models/Qwen/Qwen3.6-27B.toml
new file mode 100644
index 000000000..510073a25
--- /dev/null
+++ b/providers/stackit/models/Qwen/Qwen3.6-27B.toml
@@ -0,0 +1,39 @@
+name = "Qwen3.6-27B"
+family = "qwen"
+release_date = "2026-04-22"
+last_updated = "2026-04-22"
+attachment = true
+reasoning = false
+temperature = true
+tool_call = true
+structured_output = true
+open_weights = true
+
+[cost]
+input = 0.53
+output = 0.76
+
+[limit]
+context = 262_144
+output = 16_384
+
+[modalities]
+input = ["text"]
+output = ["text"]
+
+[[benchmarks]]
+name = "SWE-bench Verified"
+score = 77.2
+metric = "accuracy"
+source = "https://qwen.ai/blog?id=qwen3.6-27b"
+
+[[benchmarks]]
+name = "GPQA Diamond"
+score = 87.8
+metric = "accuracy"
+source = "https://qwen.ai/blog?id=qwen3.6-27b"
+
+[[weights]]
+label = "Model weights"
+url = "https://huggingface.co/Qwen/Qwen3.6-27B"
+format = "safetensors"

From 43a3ec3aceba608a15a640229f53be2fb36472e1 Mon Sep 17 00:00:00 2001
From: Florian Heuer <florian.heuer@digits.schwarz>
Date: Mon, 8 Jun 2026 10:47:02 +0200
Subject: [PATCH 6/6] Update Qwen3.6-27B.toml

---
 providers/stackit/models/Qwen/Qwen3.6-27B.toml | 17 -----------------
 1 file changed, 17 deletions(-)

diff --git a/providers/stackit/models/Qwen/Qwen3.6-27B.toml b/providers/stackit/models/Qwen/Qwen3.6-27B.toml
index 510073a25..0be7b70d2 100644
--- a/providers/stackit/models/Qwen/Qwen3.6-27B.toml
+++ b/providers/stackit/models/Qwen/Qwen3.6-27B.toml
@@ -20,20 +20,3 @@ output = 16_384
 [modalities]
 input = ["text"]
 output = ["text"]
-
-[[benchmarks]]
-name = "SWE-bench Verified"
-score = 77.2
-metric = "accuracy"
-source = "https://qwen.ai/blog?id=qwen3.6-27b"
-
-[[benchmarks]]
-name = "GPQA Diamond"
-score = 87.8
-metric = "accuracy"
-source = "https://qwen.ai/blog?id=qwen3.6-27b"
-
-[[weights]]
-label = "Model weights"
-url = "https://huggingface.co/Qwen/Qwen3.6-27B"
-format = "safetensors"