fix: [Privatemode] sync supported models (#4449)

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
Co-authored-by: Aiden Cline <aidenpcline@gmail.com>

providers/privatemode-ai/models/deepseek-ocr-2.toml

@@ -1,14 +1,14 @@
 # Sources (accessed 2026-08-10):
-# Model overview (preview, ID deepseek-ocr-2, 8192 context, text+image):
-#   https://docs.privatemode.ai/models/overview/#deepseek-ocr-2
+# Model overview (preview): https://docs.privatemode.ai/models/overview/
+# Pricing (EUR): https://docs.privatemode.ai/pricing/
 # Added in v1.51.0: https://docs.privatemode.ai/release/#v1510
-# Pricing (EUR per 1M tokens): https://docs.privatemode.ai/pricing/
-# Cost converted EUR→USD at 1 EUR = 1.1554 USD (mid-market 2026-08-10).
-# EUR list: input €0.77 / output €1.27 / cached input €0.08 per MTok.
+# Upstream: https://huggingface.co/deepseek-ai/DeepSeek-OCR-2
+# Cost: EUR list prices converted at 1 EUR = 1.1555 USD (Frankfurter 2026-08-10).
+# EUR/MTok: input 0.77 / output 1.27 / cached input 0.08
 base_model = "deepseek/deepseek-ocr-2"
 status = "beta"
 
 [cost]
-input = 0.89
-output = 1.47
-cache_read = 0.09
+input = 0.8897
+output = 1.4675
+cache_read = 0.0924

providers/privatemode-ai/models/gpt-oss-120b.toml

@@ -1,14 +1,18 @@
+# Sources (accessed 2026-08-10):
+# Model overview: https://docs.privatemode.ai/models/overview/
+# Pricing (EUR): https://docs.privatemode.ai/pricing/
+# Upstream: https://huggingface.co/openai/gpt-oss-120b
+# Cost: EUR list prices converted at 1 EUR = 1.1555 USD (Frankfurter 2026-08-10).
+# EUR/MTok: input 0.43 / output 1.70 / cached input 0.04
+# Effort: OpenAI-style reasoning_effort on this host (low|medium|high).
 base_model = "openai/gpt-oss-120b"
 name = "gpt-oss-120b"
-release_date = "2025-08-04"
-last_updated = "2025-08-14"
-knowledge = "2025-08"
 reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }]
 
 [cost]
-input = 0.00
-output = 0.00
+input = 0.4969
+output = 1.9644
+cache_read = 0.0462
 
 [limit]
 context = 128_000
-output = 128_000

providers/privatemode-ai/models/kimi-k2.6.toml

@@ -1,7 +1,11 @@
-# Sources (accessed 2026-07-15):
+# Sources (accessed 2026-08-10):
 # Model overview & reasoning toggle: https://docs.privatemode.ai/models/overview/#kimi-k26
+# Pricing (EUR): https://docs.privatemode.ai/pricing/
 # Added in v1.40.0 (default model): https://docs.privatemode.ai/release/#v1400
-# Reasoning control: $.chat_template_kwargs.thinking = true|false (default true).
+# Upstream: https://huggingface.co/moonshotai/Kimi-K2.6
+# Toggle: chat_template_kwargs.thinking = true|false (default true / reasoning on).
+# Cost: EUR list prices converted at 1 EUR = 1.1555 USD (Frankfurter 2026-08-10).
+# EUR/MTok: input 1.55 / output 7.74 / cached input 0.15
 base_model = "moonshotai/kimi-k2.6"
 reasoning_options = [{ type = "toggle" }]
 
@@ -9,5 +13,9 @@ reasoning_options = [{ type = "toggle" }]
 input = ["text", "image"]
 
 [cost]
-input = 0.00
-output = 0.00
+input = 1.7910
+output = 8.9436
+cache_read = 0.1733
+
+[limit]
+context = 256_000

providers/privatemode-ai/models/kimi-latest.toml

@@ -0,0 +1,21 @@
+# Sources (accessed 2026-08-10):
+# Alias routes to the latest available Kimi model (currently Kimi K2.6):
+# https://docs.privatemode.ai/models/overview/#kimi-k26
+# Pricing (EUR): https://docs.privatemode.ai/pricing/
+# Toggle: chat_template_kwargs.thinking = true|false (default true / reasoning on).
+# Cost: EUR list prices converted at 1 EUR = 1.1555 USD (Frankfurter 2026-08-10).
+# EUR/MTok: input 1.55 / output 7.74 / cached input 0.15
+base_model = "moonshotai/kimi-k2.6"
+name = "Kimi (latest)"
+reasoning_options = [{ type = "toggle" }]
+
+[modalities]
+input = ["text", "image"]
+
+[cost]
+input = 1.7910
+output = 8.9436
+cache_read = 0.1733
+
+[limit]
+context = 256_000

providers/privatemode-ai/models/qwen3-embedding-4b.toml

@@ -1,3 +1,10 @@
+# Sources (accessed 2026-08-10):
+# Model overview: https://docs.privatemode.ai/models/overview/#qwen3-embedding-4b
+# Pricing (EUR): https://docs.privatemode.ai/pricing/
+# Upstream quant checkpoint: https://huggingface.co/boboliu/Qwen3-Embedding-4B-W4A16-G128
+# Cost: EUR list prices converted at 1 EUR = 1.1555 USD (Frankfurter 2026-08-10).
+# EUR/MTok: input 0.13 (embeddings; output not billed).
+# Output dimensions: 1024 or 2560 via request `dimensions` (limit.output = max dims).
 name = "Qwen3-Embedding 4B"
 description = "Embedding model for semantic search, retrieval, clustering, and ranking pipelines"
 family = "qwen"
@@ -12,7 +19,7 @@ knowledge = "2025-06"
 open_weights = true
 
 [cost]
-input = 0.00
+input = 0.1502
 output = 0.00
 
 [limit]
@@ -21,4 +28,4 @@ output = 2560
 
 [modalities]
 input = ["text"]
-output = ["text"]
+output = ["text"]

providers/privatemode-ai/models/voxtral-mini-3b.toml

@@ -1,8 +1,10 @@
-# Sources (accessed 2026-07-15):
+# Sources (accessed 2026-08-10):
 # Model overview (GA since v1.39.0): https://docs.privatemode.ai/models/overview/#voxtral-mini-3b
+# Pricing (EUR per audio minute): https://docs.privatemode.ai/pricing/
 # Release notes: https://docs.privatemode.ai/release/#v1390
-# Underlying model (32k token context, audio transcription):
-# https://huggingface.co/mistralai/Voxtral-Mini-3B-2507
+# Upstream: https://huggingface.co/mistralai/Voxtral-Mini-3B-2507
+# Cost: EUR 0.004/minute converted at 1 EUR = 1.1555 USD (Frankfurter 2026-08-10).
+# cost.input is USD per audio minute (STT; not token-based).
 name = "Voxtral Mini 3B"
 description = "Speech-to-text model for audio transcription, translation, and audio understanding"
 family = "voxtral"
@@ -16,7 +18,7 @@ last_updated = "2025-07"
 open_weights = true
 
 [cost]
-input = 0.00
+input = 0.00462
 output = 0.00
 
 [limit]

providers/privatemode-ai/models/whisper-large-v3.toml

@@ -1,24 +1,13 @@
+# Sources (accessed 2026-08-10):
+# Model overview: https://docs.privatemode.ai/models/overview/
+# Pricing (EUR per audio minute): https://docs.privatemode.ai/pricing/
+# Upstream: https://huggingface.co/openai/whisper-large-v3
+# Cost: EUR 0.014/minute converted at 1 EUR = 1.1555 USD (Frankfurter 2026-08-10).
+# cost.input is USD per audio minute (STT; not token-based).
+base_model = "openai/whisper-large-v3"
 name = "Whisper large-v3"
-description = "Speech transcription model for accurate audio-to-text and captioning workflows"
-family = "whisper"
 attachment = true
-reasoning = false
-structured_output = false
-temperature = true
-tool_call = false
-release_date = "2023-09-01"
-last_updated = "2023-09-01"
-knowledge = "2023-09"
-open_weights = true
 
 [cost]
-input = 0.00
+input = 0.01618
 output = 0.00
-
-[limit]
-context = 0
-output = 4096
-
-[modalities]
-input = ["audio"]
-output = ["text"]