From bc6aa481e3256be84b4a18e9fe232cc4dcbf2a30 Mon Sep 17 00:00:00 2001
From: Claude <noreply@anthropic.com>
Date: Wed, 6 May 2026 18:17:49 +0000
Subject: [PATCH 1/5] Add Infomaniak provider with 5 open-source models

Infomaniak is a Swiss cloud provider offering an OpenAI-compatible AI API
with European data sovereignty. Adds provider config and models:

- mistralai/Ministral-3-14B-Instruct-2512 (vision, tool calling)
- google/Gemma-4-31B-it (extends canonical, vision, reasoning)
- moonshotai/Kimi-K2.6 (extends canonical, reasoning, vision, video)
- Qwen/Qwen3.5-122B-A10B-FP8 (extends canonical, reasoning, multimodal)
- swiss-ai/Apertus-70B-Instruct-2509 (open multilingual model, beta)

https://claude.ai/code/session_01MQforMKhaKerRcDUMZjnVg
---
 .../models/Qwen/Qwen3.5-122B-A10B-FP8.toml    |  4 ++++
 .../models/google/Gemma-4-31B-it.toml         |  8 +++++++
 .../Ministral-3-14B-Instruct-2512.toml        | 22 +++++++++++++++++++
 .../models/moonshotai/Kimi-K2.6.toml          |  9 ++++++++
 .../swiss-ai/Apertus-70B-Instruct-2509.toml   | 21 ++++++++++++++++++
 providers/infomaniak/provider.toml            |  5 +++++
 6 files changed, 69 insertions(+)
 create mode 100644 providers/infomaniak/models/Qwen/Qwen3.5-122B-A10B-FP8.toml
 create mode 100644 providers/infomaniak/models/google/Gemma-4-31B-it.toml
 create mode 100644 providers/infomaniak/models/mistralai/Ministral-3-14B-Instruct-2512.toml
 create mode 100644 providers/infomaniak/models/moonshotai/Kimi-K2.6.toml
 create mode 100644 providers/infomaniak/models/swiss-ai/Apertus-70B-Instruct-2509.toml
 create mode 100644 providers/infomaniak/provider.toml

diff --git a/providers/infomaniak/models/Qwen/Qwen3.5-122B-A10B-FP8.toml b/providers/infomaniak/models/Qwen/Qwen3.5-122B-A10B-FP8.toml
new file mode 100644
index 000000000..252f04071
--- /dev/null
+++ b/providers/infomaniak/models/Qwen/Qwen3.5-122B-A10B-FP8.toml
@@ -0,0 +1,4 @@
+name = "Qwen3.5 122B-A10B FP8"
+
+[extends]
+from = "alibaba/qwen3.5-122b-a10b"
diff --git a/providers/infomaniak/models/google/Gemma-4-31B-it.toml b/providers/infomaniak/models/google/Gemma-4-31B-it.toml
new file mode 100644
index 000000000..d28b1e3b7
--- /dev/null
+++ b/providers/infomaniak/models/google/Gemma-4-31B-it.toml
@@ -0,0 +1,8 @@
+name = "Gemma 4 31B"
+
+[extends]
+from = "google/gemma-4-31b-it"
+
+[cost]
+input = 0.20
+output = 0.40
diff --git a/providers/infomaniak/models/mistralai/Ministral-3-14B-Instruct-2512.toml b/providers/infomaniak/models/mistralai/Ministral-3-14B-Instruct-2512.toml
new file mode 100644
index 000000000..64d1bbe21
--- /dev/null
+++ b/providers/infomaniak/models/mistralai/Ministral-3-14B-Instruct-2512.toml
@@ -0,0 +1,22 @@
+name = "Ministral 3 14B Instruct 2512"
+family = "ministral"
+release_date = "2025-12"
+last_updated = "2025-12"
+attachment = false
+reasoning = false
+temperature = true
+tool_call = true
+structured_output = true
+open_weights = true
+
+[cost]
+input = 0.30
+output = 0.40
+
+[limit]
+context = 262_144
+output = 16_384
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
diff --git a/providers/infomaniak/models/moonshotai/Kimi-K2.6.toml b/providers/infomaniak/models/moonshotai/Kimi-K2.6.toml
new file mode 100644
index 000000000..f571775dd
--- /dev/null
+++ b/providers/infomaniak/models/moonshotai/Kimi-K2.6.toml
@@ -0,0 +1,9 @@
+name = "Kimi K2.6"
+
+[extends]
+from = "moonshotai/kimi-k2.6"
+omit = ["cost.cache_read"]
+
+[cost]
+input = 0.60
+output = 3.00
diff --git a/providers/infomaniak/models/swiss-ai/Apertus-70B-Instruct-2509.toml b/providers/infomaniak/models/swiss-ai/Apertus-70B-Instruct-2509.toml
new file mode 100644
index 000000000..e6016bd73
--- /dev/null
+++ b/providers/infomaniak/models/swiss-ai/Apertus-70B-Instruct-2509.toml
@@ -0,0 +1,21 @@
+name = "Apertus 70B Instruct 2509"
+release_date = "2025-09"
+last_updated = "2025-09"
+attachment = false
+reasoning = false
+temperature = true
+tool_call = false
+open_weights = true
+status = "beta"
+
+[cost]
+input = 0.70
+output = 2.50
+
+[limit]
+context = 65_536
+output = 8_192
+
+[modalities]
+input = ["text"]
+output = ["text"]
diff --git a/providers/infomaniak/provider.toml b/providers/infomaniak/provider.toml
new file mode 100644
index 000000000..f18f5a268
--- /dev/null
+++ b/providers/infomaniak/provider.toml
@@ -0,0 +1,5 @@
+name = "Infomaniak"
+npm = "@ai-sdk/openai-compatible"
+api = "https://api.infomaniak.com/2/ai/${INFOMANIAK_PROJECT_ID}/openai/v1"
+env = ["INFOMANIAK_API_KEY", "INFOMANIAK_PROJECT_ID"]
+doc = "https://developer.infomaniak.com/docs/api/get/1/ai/models"

From 385819a1fa8ff2e24cf450736745a005a1fe27a4 Mon Sep 17 00:00:00 2001
From: Claude <noreply@anthropic.com>
Date: Wed, 6 May 2026 18:21:22 +0000
Subject: [PATCH 2/5] Add explicit cost block to Infomaniak Qwen3.5 122B-A10B
 FP8

https://claude.ai/code/session_01MQforMKhaKerRcDUMZjnVg
---
 providers/infomaniak/models/Qwen/Qwen3.5-122B-A10B-FP8.toml | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/providers/infomaniak/models/Qwen/Qwen3.5-122B-A10B-FP8.toml b/providers/infomaniak/models/Qwen/Qwen3.5-122B-A10B-FP8.toml
index 252f04071..54022308a 100644
--- a/providers/infomaniak/models/Qwen/Qwen3.5-122B-A10B-FP8.toml
+++ b/providers/infomaniak/models/Qwen/Qwen3.5-122B-A10B-FP8.toml
@@ -2,3 +2,7 @@ name = "Qwen3.5 122B-A10B FP8"
 
 [extends]
 from = "alibaba/qwen3.5-122b-a10b"
+
+[cost]
+input = 0.40
+output = 3.20

From d4445ce07353fa695de6f47384f2d03dea5523b5 Mon Sep 17 00:00:00 2001
From: Claude <noreply@anthropic.com>
Date: Wed, 6 May 2026 18:43:14 +0000
Subject: [PATCH 3/5] Override Gemma 4 31B limits for Infomaniak (100K input
 cap)

Infomaniak caps this model at 100,000 input tokens, lower than the
canonical Google model's 256K context window.

https://claude.ai/code/session_01MQforMKhaKerRcDUMZjnVg
---
 providers/infomaniak/models/google/Gemma-4-31B-it.toml | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/providers/infomaniak/models/google/Gemma-4-31B-it.toml b/providers/infomaniak/models/google/Gemma-4-31B-it.toml
index d28b1e3b7..af9e7150d 100644
--- a/providers/infomaniak/models/google/Gemma-4-31B-it.toml
+++ b/providers/infomaniak/models/google/Gemma-4-31B-it.toml
@@ -6,3 +6,8 @@ from = "google/gemma-4-31b-it"
 [cost]
 input = 0.20
 output = 0.40
+
+[limit]
+context = 100_000
+input = 100_000
+output = 8_192

From 695ef47938d049615760944ea0f7d814d2a0e12c Mon Sep 17 00:00:00 2001
From: Claude <noreply@anthropic.com>
Date: Wed, 6 May 2026 18:45:35 +0000
Subject: [PATCH 4/5] Override Gemma 4 31B modalities for Infomaniak (text-only
 input)

Infomaniak only exposes text-to-text for this model, not the vision
capabilities available in the canonical Google model.

https://claude.ai/code/session_01MQforMKhaKerRcDUMZjnVg
---
 providers/infomaniak/models/google/Gemma-4-31B-it.toml | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/providers/infomaniak/models/google/Gemma-4-31B-it.toml b/providers/infomaniak/models/google/Gemma-4-31B-it.toml
index af9e7150d..287d76263 100644
--- a/providers/infomaniak/models/google/Gemma-4-31B-it.toml
+++ b/providers/infomaniak/models/google/Gemma-4-31B-it.toml
@@ -11,3 +11,7 @@ output = 0.40
 context = 100_000
 input = 100_000
 output = 8_192
+
+[modalities]
+input = ["text"]
+output = ["text"]

From 52fa57049da49da87d517b2c965c3701e505135f Mon Sep 17 00:00:00 2001
From: Claude <noreply@anthropic.com>
Date: Wed, 6 May 2026 18:49:09 +0000
Subject: [PATCH 5/5] Fix limits, modalities and status for Infomaniak models

Based on Infomaniak's model page screenshots:
- Qwen3.5 122B FP8: input cap 200K, image-text modality, beta
- Kimi K2.6: input cap 256K, image-text modality (no video), beta
- Ministral 3 14B: input cap 100K, beta
- Gemma 4 31B: beta
All non-Apertus models now correctly reflect Infomaniak's published specs.

https://claude.ai/code/session_01MQforMKhaKerRcDUMZjnVg
---
 .../infomaniak/models/Qwen/Qwen3.5-122B-A10B-FP8.toml  | 10 ++++++++++
 providers/infomaniak/models/google/Gemma-4-31B-it.toml |  1 +
 .../mistralai/Ministral-3-14B-Instruct-2512.toml       |  4 +++-
 providers/infomaniak/models/moonshotai/Kimi-K2.6.toml  | 10 ++++++++++
 4 files changed, 24 insertions(+), 1 deletion(-)

diff --git a/providers/infomaniak/models/Qwen/Qwen3.5-122B-A10B-FP8.toml b/providers/infomaniak/models/Qwen/Qwen3.5-122B-A10B-FP8.toml
index 54022308a..441ef9c77 100644
--- a/providers/infomaniak/models/Qwen/Qwen3.5-122B-A10B-FP8.toml
+++ b/providers/infomaniak/models/Qwen/Qwen3.5-122B-A10B-FP8.toml
@@ -1,4 +1,5 @@
 name = "Qwen3.5 122B-A10B FP8"
+status = "beta"
 
 [extends]
 from = "alibaba/qwen3.5-122b-a10b"
@@ -6,3 +7,12 @@ from = "alibaba/qwen3.5-122b-a10b"
 [cost]
 input = 0.40
 output = 3.20
+
+[limit]
+context = 200_000
+input = 200_000
+output = 65_536
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]
diff --git a/providers/infomaniak/models/google/Gemma-4-31B-it.toml b/providers/infomaniak/models/google/Gemma-4-31B-it.toml
index 287d76263..9e4ef25b8 100644
--- a/providers/infomaniak/models/google/Gemma-4-31B-it.toml
+++ b/providers/infomaniak/models/google/Gemma-4-31B-it.toml
@@ -1,4 +1,5 @@
 name = "Gemma 4 31B"
+status = "beta"
 
 [extends]
 from = "google/gemma-4-31b-it"
diff --git a/providers/infomaniak/models/mistralai/Ministral-3-14B-Instruct-2512.toml b/providers/infomaniak/models/mistralai/Ministral-3-14B-Instruct-2512.toml
index 64d1bbe21..efb7f27f8 100644
--- a/providers/infomaniak/models/mistralai/Ministral-3-14B-Instruct-2512.toml
+++ b/providers/infomaniak/models/mistralai/Ministral-3-14B-Instruct-2512.toml
@@ -8,13 +8,15 @@ temperature = true
 tool_call = true
 structured_output = true
 open_weights = true
+status = "beta"
 
 [cost]
 input = 0.30
 output = 0.40
 
 [limit]
-context = 262_144
+context = 100_000
+input = 100_000
 output = 16_384
 
 [modalities]
diff --git a/providers/infomaniak/models/moonshotai/Kimi-K2.6.toml b/providers/infomaniak/models/moonshotai/Kimi-K2.6.toml
index f571775dd..db66c5cc0 100644
--- a/providers/infomaniak/models/moonshotai/Kimi-K2.6.toml
+++ b/providers/infomaniak/models/moonshotai/Kimi-K2.6.toml
@@ -1,4 +1,5 @@
 name = "Kimi K2.6"
+status = "beta"
 
 [extends]
 from = "moonshotai/kimi-k2.6"
@@ -7,3 +8,12 @@ omit = ["cost.cache_read"]
 [cost]
 input = 0.60
 output = 3.00
+
+[limit]
+context = 256_000
+input = 256_000
+output = 32_768
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]