From c71d0c80658e9f557baec3f05e0df0403ecf4b2a Mon Sep 17 00:00:00 2001
From: Aiden Cline <aidenpcline@gmail.com>
Date: Wed, 10 Jun 2026 16:05:01 -0500
Subject: [PATCH 1/4] fix(deepinfra): reconcile reasoning controls

---
 providers/deepinfra/models/MiniMaxAI/MiniMax-M2.5.toml          | 2 +-
 .../models/Qwen/Qwen3-Coder-480B-A35B-Instruct-Turbo.toml       | 1 -
 providers/deepinfra/models/Qwen/Qwen3.5-35B-A3B.toml            | 2 +-
 providers/deepinfra/models/Qwen/Qwen3.5-397B-A17B.toml          | 2 +-
 providers/deepinfra/models/Qwen/Qwen3.6-35B-A3B.toml            | 2 +-
 providers/deepinfra/models/XiaomiMiMo/MiMo-V2.5-Pro.toml        | 2 +-
 providers/deepinfra/models/XiaomiMiMo/MiMo-V2.5.toml            | 2 +-
 providers/deepinfra/models/deepseek-ai/DeepSeek-R1-0528.toml    | 2 +-
 providers/deepinfra/models/deepseek-ai/DeepSeek-V3.2.toml       | 2 +-
 providers/deepinfra/models/deepseek-ai/DeepSeek-V4-Flash.toml   | 2 +-
 providers/deepinfra/models/deepseek-ai/DeepSeek-V4-Pro.toml     | 2 +-
 providers/deepinfra/models/google/gemma-4-26B-A4B-it.toml       | 2 +-
 providers/deepinfra/models/google/gemma-4-31B-it.toml           | 2 +-
 .../models/meta-llama/Llama-3.3-70B-Instruct-Turbo.toml         | 1 -
 .../meta-llama/Llama-4-Maverick-17B-128E-Instruct-FP8.toml      | 1 -
 .../models/meta-llama/Llama-4-Scout-17B-16E-Instruct.toml       | 1 -
 providers/deepinfra/models/moonshotai/Kimi-K2.5.toml            | 2 +-
 providers/deepinfra/models/moonshotai/Kimi-K2.6.toml            | 2 +-
 providers/deepinfra/models/openai/gpt-oss-120b.toml             | 2 +-
 providers/deepinfra/models/openai/gpt-oss-20b.toml              | 2 +-
 providers/deepinfra/models/zai-org/GLM-4.6.toml                 | 2 +-
 providers/deepinfra/models/zai-org/GLM-4.7-Flash.toml           | 2 +-
 providers/deepinfra/models/zai-org/GLM-4.7.toml                 | 2 +-
 providers/deepinfra/models/zai-org/GLM-5.1.toml                 | 2 +-
 providers/deepinfra/models/zai-org/GLM-5.toml                   | 2 +-
 25 files changed, 21 insertions(+), 25 deletions(-)

diff --git a/providers/deepinfra/models/MiniMaxAI/MiniMax-M2.5.toml b/providers/deepinfra/models/MiniMaxAI/MiniMax-M2.5.toml
index 0a46f9656..d156d1ef0 100644
--- a/providers/deepinfra/models/MiniMaxAI/MiniMax-M2.5.toml
+++ b/providers/deepinfra/models/MiniMaxAI/MiniMax-M2.5.toml
@@ -5,7 +5,7 @@ release_date = "2026-02-12"
 last_updated = "2026-02-12"
 attachment = false
 reasoning = true
-reasoning_options = []
+reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high"] }]
 temperature = true
 tool_call = true
 knowledge = "2025-06"
diff --git a/providers/deepinfra/models/Qwen/Qwen3-Coder-480B-A35B-Instruct-Turbo.toml b/providers/deepinfra/models/Qwen/Qwen3-Coder-480B-A35B-Instruct-Turbo.toml
index d8cedbb09..0593a47db 100644
--- a/providers/deepinfra/models/Qwen/Qwen3-Coder-480B-A35B-Instruct-Turbo.toml
+++ b/providers/deepinfra/models/Qwen/Qwen3-Coder-480B-A35B-Instruct-Turbo.toml
@@ -4,7 +4,6 @@ release_date = "2025-07-23"
 last_updated = "2025-07-23"
 attachment = false
 reasoning = false
-reasoning_options = []
 temperature = true
 knowledge = "2025-04"
 tool_call = true
diff --git a/providers/deepinfra/models/Qwen/Qwen3.5-35B-A3B.toml b/providers/deepinfra/models/Qwen/Qwen3.5-35B-A3B.toml
index 772dd6d1e..4c2b2b2cf 100644
--- a/providers/deepinfra/models/Qwen/Qwen3.5-35B-A3B.toml
+++ b/providers/deepinfra/models/Qwen/Qwen3.5-35B-A3B.toml
@@ -4,7 +4,7 @@ release_date = "2026-02-01"
 last_updated = "2026-04-20"
 attachment = true
 reasoning = true
-reasoning_options = [{ type = "toggle" }]
+reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high"] }]
 temperature = true
 knowledge = "2025-01"
 tool_call = true
diff --git a/providers/deepinfra/models/Qwen/Qwen3.5-397B-A17B.toml b/providers/deepinfra/models/Qwen/Qwen3.5-397B-A17B.toml
index 8a7c912a5..641c474c1 100644
--- a/providers/deepinfra/models/Qwen/Qwen3.5-397B-A17B.toml
+++ b/providers/deepinfra/models/Qwen/Qwen3.5-397B-A17B.toml
@@ -4,7 +4,7 @@ release_date = "2026-02-01"
 last_updated = "2026-04-20"
 attachment = true
 reasoning = true
-reasoning_options = [{ type = "toggle" }]
+reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high"] }]
 temperature = true
 knowledge = "2025-01"
 tool_call = true
diff --git a/providers/deepinfra/models/Qwen/Qwen3.6-35B-A3B.toml b/providers/deepinfra/models/Qwen/Qwen3.6-35B-A3B.toml
index 6208ed020..7c966c087 100644
--- a/providers/deepinfra/models/Qwen/Qwen3.6-35B-A3B.toml
+++ b/providers/deepinfra/models/Qwen/Qwen3.6-35B-A3B.toml
@@ -4,7 +4,7 @@ release_date = "2026-04-01"
 last_updated = "2026-04-01"
 attachment = true
 reasoning = true
-reasoning_options = [{ type = "toggle" }]
+reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high"] }]
 temperature = true
 tool_call = true
 open_weights = true
diff --git a/providers/deepinfra/models/XiaomiMiMo/MiMo-V2.5-Pro.toml b/providers/deepinfra/models/XiaomiMiMo/MiMo-V2.5-Pro.toml
index a6dc5161b..444e6173a 100644
--- a/providers/deepinfra/models/XiaomiMiMo/MiMo-V2.5-Pro.toml
+++ b/providers/deepinfra/models/XiaomiMiMo/MiMo-V2.5-Pro.toml
@@ -1,5 +1,5 @@
 base_model = "xiaomi/mimo-v2.5-pro"
-reasoning_options = [{ type = "toggle" }]
+reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high"] }]
 
 [interleaved]
 field = "reasoning_content"
diff --git a/providers/deepinfra/models/XiaomiMiMo/MiMo-V2.5.toml b/providers/deepinfra/models/XiaomiMiMo/MiMo-V2.5.toml
index 766f6c20f..cce93c855 100644
--- a/providers/deepinfra/models/XiaomiMiMo/MiMo-V2.5.toml
+++ b/providers/deepinfra/models/XiaomiMiMo/MiMo-V2.5.toml
@@ -1,5 +1,5 @@
 base_model = "xiaomi/mimo-v2.5"
-reasoning_options = [{ type = "toggle" }]
+reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high"] }]
 
 [interleaved]
 field = "reasoning_content"
diff --git a/providers/deepinfra/models/deepseek-ai/DeepSeek-R1-0528.toml b/providers/deepinfra/models/deepseek-ai/DeepSeek-R1-0528.toml
index bd080cd58..678bfe7bc 100644
--- a/providers/deepinfra/models/deepseek-ai/DeepSeek-R1-0528.toml
+++ b/providers/deepinfra/models/deepseek-ai/DeepSeek-R1-0528.toml
@@ -3,7 +3,7 @@ release_date = "2025-05-28"
 last_updated = "2025-05-28"
 attachment = false
 reasoning = true
-reasoning_options = []
+reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high"] }]
 temperature = true
 knowledge = "2024-07"
 tool_call = false
diff --git a/providers/deepinfra/models/deepseek-ai/DeepSeek-V3.2.toml b/providers/deepinfra/models/deepseek-ai/DeepSeek-V3.2.toml
index 07c3cdb0d..d121cb1d9 100644
--- a/providers/deepinfra/models/deepseek-ai/DeepSeek-V3.2.toml
+++ b/providers/deepinfra/models/deepseek-ai/DeepSeek-V3.2.toml
@@ -3,7 +3,7 @@ release_date = "2025-12-02"
 last_updated = "2025-12-02"
 attachment = false
 reasoning = true
-reasoning_options = [{ type = "toggle" }]
+reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high"] }]
 temperature = true
 knowledge = "2024-12"
 tool_call = true
diff --git a/providers/deepinfra/models/deepseek-ai/DeepSeek-V4-Flash.toml b/providers/deepinfra/models/deepseek-ai/DeepSeek-V4-Flash.toml
index 6c708c932..7fddc8db7 100644
--- a/providers/deepinfra/models/deepseek-ai/DeepSeek-V4-Flash.toml
+++ b/providers/deepinfra/models/deepseek-ai/DeepSeek-V4-Flash.toml
@@ -1,5 +1,5 @@
 base_model = "deepseek/deepseek-v4-flash"
-reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high", "xhigh"] }]
+reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high"] }]
 
 [interleaved]
 field = "reasoning_content"
diff --git a/providers/deepinfra/models/deepseek-ai/DeepSeek-V4-Pro.toml b/providers/deepinfra/models/deepseek-ai/DeepSeek-V4-Pro.toml
index fb50a1e0f..ad77c847b 100644
--- a/providers/deepinfra/models/deepseek-ai/DeepSeek-V4-Pro.toml
+++ b/providers/deepinfra/models/deepseek-ai/DeepSeek-V4-Pro.toml
@@ -1,5 +1,5 @@
 base_model = "deepseek/deepseek-v4-pro"
-reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high", "xhigh"] }]
+reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high"] }]
 
 [interleaved]
 field = "reasoning_content"
diff --git a/providers/deepinfra/models/google/gemma-4-26B-A4B-it.toml b/providers/deepinfra/models/google/gemma-4-26B-A4B-it.toml
index ee3319c85..811637712 100644
--- a/providers/deepinfra/models/google/gemma-4-26B-A4B-it.toml
+++ b/providers/deepinfra/models/google/gemma-4-26B-A4B-it.toml
@@ -1,5 +1,5 @@
 base_model = "google/gemma-4-26b-a4b-it"
-reasoning_options = [{ type = "toggle" }]
+reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high"] }]
 
 [cost]
 input = 0.07
diff --git a/providers/deepinfra/models/google/gemma-4-31B-it.toml b/providers/deepinfra/models/google/gemma-4-31B-it.toml
index f9dc05e17..53541f4cc 100644
--- a/providers/deepinfra/models/google/gemma-4-31B-it.toml
+++ b/providers/deepinfra/models/google/gemma-4-31B-it.toml
@@ -1,5 +1,5 @@
 base_model = "google/gemma-4-31b-it"
-reasoning_options = [{ type = "toggle" }]
+reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high"] }]
 
 [cost]
 input = 0.13
diff --git a/providers/deepinfra/models/meta-llama/Llama-3.3-70B-Instruct-Turbo.toml b/providers/deepinfra/models/meta-llama/Llama-3.3-70B-Instruct-Turbo.toml
index e2ca7401f..df433558b 100644
--- a/providers/deepinfra/models/meta-llama/Llama-3.3-70B-Instruct-Turbo.toml
+++ b/providers/deepinfra/models/meta-llama/Llama-3.3-70B-Instruct-Turbo.toml
@@ -4,7 +4,6 @@ release_date = "2024-12-06"
 last_updated = "2024-12-06"
 attachment = false
 reasoning = false
-reasoning_options = []
 tool_call = true
 open_weights = true
 
diff --git a/providers/deepinfra/models/meta-llama/Llama-4-Maverick-17B-128E-Instruct-FP8.toml b/providers/deepinfra/models/meta-llama/Llama-4-Maverick-17B-128E-Instruct-FP8.toml
index 98ca558fc..1a2b88bbc 100644
--- a/providers/deepinfra/models/meta-llama/Llama-4-Maverick-17B-128E-Instruct-FP8.toml
+++ b/providers/deepinfra/models/meta-llama/Llama-4-Maverick-17B-128E-Instruct-FP8.toml
@@ -4,7 +4,6 @@ release_date = "2025-04-05"
 last_updated = "2025-04-05"
 attachment = false
 reasoning = false
-reasoning_options = []
 tool_call = true
 open_weights = true
 
diff --git a/providers/deepinfra/models/meta-llama/Llama-4-Scout-17B-16E-Instruct.toml b/providers/deepinfra/models/meta-llama/Llama-4-Scout-17B-16E-Instruct.toml
index 0bcfdc0c0..8f7bab10f 100644
--- a/providers/deepinfra/models/meta-llama/Llama-4-Scout-17B-16E-Instruct.toml
+++ b/providers/deepinfra/models/meta-llama/Llama-4-Scout-17B-16E-Instruct.toml
@@ -4,7 +4,6 @@ release_date = "2025-04-05"
 last_updated = "2025-04-05"
 attachment = false
 reasoning = false
-reasoning_options = []
 tool_call = true
 open_weights = true
 
diff --git a/providers/deepinfra/models/moonshotai/Kimi-K2.5.toml b/providers/deepinfra/models/moonshotai/Kimi-K2.5.toml
index 17a816502..bc030925f 100644
--- a/providers/deepinfra/models/moonshotai/Kimi-K2.5.toml
+++ b/providers/deepinfra/models/moonshotai/Kimi-K2.5.toml
@@ -4,7 +4,7 @@ release_date = "2026-01-27"
 last_updated = "2026-01-27"
 attachment = true
 reasoning = true
-reasoning_options = [{ type = "toggle" }]
+reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high"] }]
 structured_output = true
 temperature = true
 tool_call = true
diff --git a/providers/deepinfra/models/moonshotai/Kimi-K2.6.toml b/providers/deepinfra/models/moonshotai/Kimi-K2.6.toml
index 60a745ac7..a5b127070 100644
--- a/providers/deepinfra/models/moonshotai/Kimi-K2.6.toml
+++ b/providers/deepinfra/models/moonshotai/Kimi-K2.6.toml
@@ -4,7 +4,7 @@ release_date = "2026-04-21"
 last_updated = "2026-04-21"
 attachment = true
 reasoning = true
-reasoning_options = [{ type = "toggle" }]
+reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high"] }]
 structured_output = true
 temperature = true
 tool_call = true
diff --git a/providers/deepinfra/models/openai/gpt-oss-120b.toml b/providers/deepinfra/models/openai/gpt-oss-120b.toml
index 6da7ba780..7a713c352 100644
--- a/providers/deepinfra/models/openai/gpt-oss-120b.toml
+++ b/providers/deepinfra/models/openai/gpt-oss-120b.toml
@@ -6,7 +6,7 @@ release_date = "2025-08-05"
 last_updated = "2025-08-05"
 attachment = false
 reasoning = true
-reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }]
+reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high"] }]
 temperature = true
 tool_call = true
 open_weights = true
diff --git a/providers/deepinfra/models/openai/gpt-oss-20b.toml b/providers/deepinfra/models/openai/gpt-oss-20b.toml
index eebce8f2e..d7166c7c5 100644
--- a/providers/deepinfra/models/openai/gpt-oss-20b.toml
+++ b/providers/deepinfra/models/openai/gpt-oss-20b.toml
@@ -6,7 +6,7 @@ release_date = "2025-08-05"
 last_updated = "2025-08-05"
 attachment = false
 reasoning = true
-reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }]
+reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high"] }]
 temperature = true
 tool_call = true
 open_weights = true
diff --git a/providers/deepinfra/models/zai-org/GLM-4.6.toml b/providers/deepinfra/models/zai-org/GLM-4.6.toml
index 614f0d683..2352f5451 100644
--- a/providers/deepinfra/models/zai-org/GLM-4.6.toml
+++ b/providers/deepinfra/models/zai-org/GLM-4.6.toml
@@ -6,7 +6,7 @@ last_updated = "2025-09-30"
 knowledge = "2025-04"
 attachment = false
 reasoning = true
-reasoning_options = [{ type = "toggle" }]
+reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high"] }]
 temperature = true
 tool_call = true
 open_weights = true
diff --git a/providers/deepinfra/models/zai-org/GLM-4.7-Flash.toml b/providers/deepinfra/models/zai-org/GLM-4.7-Flash.toml
index b64fc596e..d3dba78bd 100644
--- a/providers/deepinfra/models/zai-org/GLM-4.7-Flash.toml
+++ b/providers/deepinfra/models/zai-org/GLM-4.7-Flash.toml
@@ -6,7 +6,7 @@ last_updated = "2026-01-19"
 knowledge = "2025-04"
 attachment = false
 reasoning = true
-reasoning_options = [{ type = "toggle" }]
+reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high"] }]
 temperature = true
 tool_call = true
 open_weights = true
diff --git a/providers/deepinfra/models/zai-org/GLM-4.7.toml b/providers/deepinfra/models/zai-org/GLM-4.7.toml
index 0291e9572..b45850427 100644
--- a/providers/deepinfra/models/zai-org/GLM-4.7.toml
+++ b/providers/deepinfra/models/zai-org/GLM-4.7.toml
@@ -6,7 +6,7 @@ last_updated = "2025-12-22"
 knowledge = "2025-04"
 attachment = false
 reasoning = true
-reasoning_options = [{ type = "toggle" }]
+reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high"] }]
 temperature = true
 tool_call = true
 open_weights = true
diff --git a/providers/deepinfra/models/zai-org/GLM-5.1.toml b/providers/deepinfra/models/zai-org/GLM-5.1.toml
index b2624d58a..2994ed6b8 100644
--- a/providers/deepinfra/models/zai-org/GLM-5.1.toml
+++ b/providers/deepinfra/models/zai-org/GLM-5.1.toml
@@ -6,7 +6,7 @@ last_updated = "2026-04-07"
 knowledge = "2025-04"
 attachment = false
 reasoning = true
-reasoning_options = [{ type = "toggle" }]
+reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high"] }]
 temperature = true
 tool_call = true
 open_weights = true
diff --git a/providers/deepinfra/models/zai-org/GLM-5.toml b/providers/deepinfra/models/zai-org/GLM-5.toml
index 595b286fc..00d9de73b 100644
--- a/providers/deepinfra/models/zai-org/GLM-5.toml
+++ b/providers/deepinfra/models/zai-org/GLM-5.toml
@@ -6,7 +6,7 @@ last_updated = "2026-02-12"
 knowledge = "2025-12"
 attachment = false
 reasoning = true
-reasoning_options = [{ type = "toggle" }]
+reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high"] }]
 temperature = true
 tool_call = true
 open_weights = true

From 79e47be9522fec55b04c78fc7857ed1ec8c7237d Mon Sep 17 00:00:00 2001
From: Aiden Cline <aidenpcline@gmail.com>
Date: Wed, 10 Jun 2026 19:50:58 -0500
Subject: [PATCH 2/4] fix(deepinfra): use model-specific reasoning controls

---
 packages/core/test/deepinfra.test.ts          | 82 +++++++++++++++++++
 .../models/MiniMaxAI/MiniMax-M2.5.toml        |  8 +-
 .../Qwen3-Coder-480B-A35B-Instruct-Turbo.toml |  2 +-
 .../models/Qwen/Qwen3.5-35B-A3B.toml          |  8 +-
 .../models/Qwen/Qwen3.5-397B-A17B.toml        |  8 +-
 .../models/Qwen/Qwen3.6-35B-A3B.toml          |  6 +-
 .../models/XiaomiMiMo/MiMo-V2.5-Pro.toml      |  2 +-
 .../models/XiaomiMiMo/MiMo-V2.5.toml          |  2 +-
 .../models/deepseek-ai/DeepSeek-R1-0528.toml  |  3 +-
 .../models/deepseek-ai/DeepSeek-V3.2.toml     |  2 +-
 .../models/deepseek-ai/DeepSeek-V4-Flash.toml |  2 +-
 .../models/deepseek-ai/DeepSeek-V4-Pro.toml   |  2 +-
 .../models/google/gemma-4-26B-A4B-it.toml     |  2 +-
 .../models/google/gemma-4-31B-it.toml         |  2 +-
 ...lama-4-Maverick-17B-128E-Instruct-FP8.toml |  4 +-
 .../Llama-4-Scout-17B-16E-Instruct.toml       |  4 +-
 .../models/moonshotai/Kimi-K2.5.toml          |  8 +-
 .../models/moonshotai/Kimi-K2.6.toml          |  2 +-
 .../deepinfra/models/openai/gpt-oss-120b.toml |  6 +-
 .../deepinfra/models/openai/gpt-oss-20b.toml  |  2 +-
 .../deepinfra/models/zai-org/GLM-4.6.toml     |  4 +-
 .../models/zai-org/GLM-4.7-Flash.toml         |  2 +-
 .../deepinfra/models/zai-org/GLM-4.7.toml     |  4 +-
 .../deepinfra/models/zai-org/GLM-5.1.toml     |  8 +-
 providers/deepinfra/models/zai-org/GLM-5.toml |  8 +-
 25 files changed, 132 insertions(+), 51 deletions(-)
 create mode 100644 packages/core/test/deepinfra.test.ts

diff --git a/packages/core/test/deepinfra.test.ts b/packages/core/test/deepinfra.test.ts
new file mode 100644
index 000000000..fda06b97f
--- /dev/null
+++ b/packages/core/test/deepinfra.test.ts
@@ -0,0 +1,82 @@
+import { expect, test } from "bun:test";
+import path from "node:path";
+
+import { generate } from "../src/index.js";
+
+test("DeepInfra models expose only verified reasoning controls", async () => {
+  const root = path.join(import.meta.dirname, "..", "..", "..");
+  const deepinfra = (await generate(path.join(root, "providers"))).deepinfra;
+
+  expect(deepinfra).toBeDefined();
+
+  const fixed = ["MiniMaxAI/MiniMax-M2.5"];
+  const toggle = [
+    "Qwen/Qwen3.5-35B-A3B",
+    "Qwen/Qwen3.5-397B-A17B",
+    "Qwen/Qwen3.6-35B-A3B",
+    "XiaomiMiMo/MiMo-V2.5-Pro",
+    "XiaomiMiMo/MiMo-V2.5",
+    "deepseek-ai/DeepSeek-V3.2",
+    "google/gemma-4-26B-A4B-it",
+    "google/gemma-4-31B-it",
+    "moonshotai/Kimi-K2.5",
+    "moonshotai/Kimi-K2.6",
+    "zai-org/GLM-4.6",
+    "zai-org/GLM-4.7-Flash",
+    "zai-org/GLM-4.7",
+    "zai-org/GLM-5.1",
+    "zai-org/GLM-5",
+  ];
+  const standardEffort = ["openai/gpt-oss-120b", "openai/gpt-oss-20b"];
+  const v4Effort = [
+    "deepseek-ai/DeepSeek-V4-Flash",
+    "deepseek-ai/DeepSeek-V4-Pro",
+  ];
+  const nonReasoning = [
+    "Qwen/Qwen3-Coder-480B-A35B-Instruct-Turbo",
+    "meta-llama/Llama-3.3-70B-Instruct-Turbo",
+    "meta-llama/Llama-4-Maverick-17B-128E-Instruct-FP8",
+    "meta-llama/Llama-4-Scout-17B-16E-Instruct",
+  ];
+  const unresolved = ["deepseek-ai/DeepSeek-R1-0528"];
+
+  expect(Object.keys(deepinfra?.models ?? {}).sort()).toEqual(
+    [
+      ...fixed,
+      ...toggle,
+      ...standardEffort,
+      ...v4Effort,
+      ...nonReasoning,
+      ...unresolved,
+    ].sort(),
+  );
+
+  for (const id of fixed) {
+    expect(deepinfra?.models[id]?.reasoning).toBe(true);
+    expect(deepinfra?.models[id]?.reasoning_options).toEqual([]);
+  }
+  for (const id of toggle) {
+    expect(deepinfra?.models[id]?.reasoning).toBe(true);
+    expect(deepinfra?.models[id]?.reasoning_options).toEqual([{ type: "toggle" }]);
+  }
+  for (const id of standardEffort) {
+    expect(deepinfra?.models[id]?.reasoning).toBe(true);
+    expect(deepinfra?.models[id]?.reasoning_options).toEqual([
+      { type: "effort", values: ["low", "medium", "high"] },
+    ]);
+  }
+  for (const id of v4Effort) {
+    expect(deepinfra?.models[id]?.reasoning).toBe(true);
+    expect(deepinfra?.models[id]?.reasoning_options).toEqual([
+      { type: "effort", values: ["none", "high", "max"] },
+    ]);
+  }
+  for (const id of nonReasoning) {
+    expect(deepinfra?.models[id]?.reasoning).toBe(false);
+    expect(deepinfra?.models[id]?.reasoning_options).toBeUndefined();
+  }
+  for (const id of unresolved) {
+    expect(deepinfra?.models[id]?.reasoning).toBe(true);
+    expect(deepinfra?.models[id]?.reasoning_options).toBeUndefined();
+  }
+});
diff --git a/providers/deepinfra/models/MiniMaxAI/MiniMax-M2.5.toml b/providers/deepinfra/models/MiniMaxAI/MiniMax-M2.5.toml
index d156d1ef0..57bc81952 100644
--- a/providers/deepinfra/models/MiniMaxAI/MiniMax-M2.5.toml
+++ b/providers/deepinfra/models/MiniMaxAI/MiniMax-M2.5.toml
@@ -5,20 +5,20 @@ release_date = "2026-02-12"
 last_updated = "2026-02-12"
 attachment = false
 reasoning = true
-reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high"] }]
+reasoning_options = []
 temperature = true
 tool_call = true
 knowledge = "2025-06"
 open_weights = true
 
 [cost]
-input = 0.27
-output = 0.95
+input = 0.15
+output = 1.15
 cache_read = 0.03
 cache_write = 0.375
 
 [limit]
-context = 204_800
+context = 196_608
 output = 131_072
 
 [modalities]
diff --git a/providers/deepinfra/models/Qwen/Qwen3-Coder-480B-A35B-Instruct-Turbo.toml b/providers/deepinfra/models/Qwen/Qwen3-Coder-480B-A35B-Instruct-Turbo.toml
index 0593a47db..44dd1e436 100644
--- a/providers/deepinfra/models/Qwen/Qwen3-Coder-480B-A35B-Instruct-Turbo.toml
+++ b/providers/deepinfra/models/Qwen/Qwen3-Coder-480B-A35B-Instruct-Turbo.toml
@@ -11,7 +11,7 @@ open_weights = true
 
 [cost]
 input = 0.3
-output = 1.2
+output = 1.0
 
 [limit]
 context = 262_144
diff --git a/providers/deepinfra/models/Qwen/Qwen3.5-35B-A3B.toml b/providers/deepinfra/models/Qwen/Qwen3.5-35B-A3B.toml
index 4c2b2b2cf..f89e336f2 100644
--- a/providers/deepinfra/models/Qwen/Qwen3.5-35B-A3B.toml
+++ b/providers/deepinfra/models/Qwen/Qwen3.5-35B-A3B.toml
@@ -4,16 +4,16 @@ release_date = "2026-02-01"
 last_updated = "2026-04-20"
 attachment = true
 reasoning = true
-reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high"] }]
+reasoning_options = [{ type = "toggle" }]
 temperature = true
 knowledge = "2025-01"
 tool_call = true
 open_weights = true
 
 [cost]
-input = 0.2
-output = 0.95
-cached = 0.1
+input = 0.14
+output = 1.00
+cached = 0.05
 
 [limit]
 context = 262_144
diff --git a/providers/deepinfra/models/Qwen/Qwen3.5-397B-A17B.toml b/providers/deepinfra/models/Qwen/Qwen3.5-397B-A17B.toml
index 641c474c1..427846ef5 100644
--- a/providers/deepinfra/models/Qwen/Qwen3.5-397B-A17B.toml
+++ b/providers/deepinfra/models/Qwen/Qwen3.5-397B-A17B.toml
@@ -4,16 +4,16 @@ release_date = "2026-02-01"
 last_updated = "2026-04-20"
 attachment = true
 reasoning = true
-reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high"] }]
+reasoning_options = [{ type = "toggle" }]
 temperature = true
 knowledge = "2025-01"
 tool_call = true
 open_weights = true
 
 [cost]
-input = 0.54
-output = 3.4
-cached = 0.27
+input = 0.45
+output = 3.00
+cached = 0.22
 
 [limit]
 context = 262_144
diff --git a/providers/deepinfra/models/Qwen/Qwen3.6-35B-A3B.toml b/providers/deepinfra/models/Qwen/Qwen3.6-35B-A3B.toml
index 7c966c087..9feaa1218 100644
--- a/providers/deepinfra/models/Qwen/Qwen3.6-35B-A3B.toml
+++ b/providers/deepinfra/models/Qwen/Qwen3.6-35B-A3B.toml
@@ -4,14 +4,14 @@ release_date = "2026-04-01"
 last_updated = "2026-04-01"
 attachment = true
 reasoning = true
-reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high"] }]
+reasoning_options = [{ type = "toggle" }]
 temperature = true
 tool_call = true
 open_weights = true
 
 [cost]
-input = 0.20
-output = 1.00
+input = 0.15
+output = 0.95
 
 [limit]
 context = 262_144
diff --git a/providers/deepinfra/models/XiaomiMiMo/MiMo-V2.5-Pro.toml b/providers/deepinfra/models/XiaomiMiMo/MiMo-V2.5-Pro.toml
index 444e6173a..a6dc5161b 100644
--- a/providers/deepinfra/models/XiaomiMiMo/MiMo-V2.5-Pro.toml
+++ b/providers/deepinfra/models/XiaomiMiMo/MiMo-V2.5-Pro.toml
@@ -1,5 +1,5 @@
 base_model = "xiaomi/mimo-v2.5-pro"
-reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high"] }]
+reasoning_options = [{ type = "toggle" }]
 
 [interleaved]
 field = "reasoning_content"
diff --git a/providers/deepinfra/models/XiaomiMiMo/MiMo-V2.5.toml b/providers/deepinfra/models/XiaomiMiMo/MiMo-V2.5.toml
index cce93c855..766f6c20f 100644
--- a/providers/deepinfra/models/XiaomiMiMo/MiMo-V2.5.toml
+++ b/providers/deepinfra/models/XiaomiMiMo/MiMo-V2.5.toml
@@ -1,5 +1,5 @@
 base_model = "xiaomi/mimo-v2.5"
-reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high"] }]
+reasoning_options = [{ type = "toggle" }]
 
 [interleaved]
 field = "reasoning_content"
diff --git a/providers/deepinfra/models/deepseek-ai/DeepSeek-R1-0528.toml b/providers/deepinfra/models/deepseek-ai/DeepSeek-R1-0528.toml
index 678bfe7bc..2b4bc6be5 100644
--- a/providers/deepinfra/models/deepseek-ai/DeepSeek-R1-0528.toml
+++ b/providers/deepinfra/models/deepseek-ai/DeepSeek-R1-0528.toml
@@ -3,10 +3,9 @@ release_date = "2025-05-28"
 last_updated = "2025-05-28"
 attachment = false
 reasoning = true
-reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high"] }]
 temperature = true
 knowledge = "2024-07"
-tool_call = false
+tool_call = true
 open_weights = false
 
 [interleaved]
diff --git a/providers/deepinfra/models/deepseek-ai/DeepSeek-V3.2.toml b/providers/deepinfra/models/deepseek-ai/DeepSeek-V3.2.toml
index d121cb1d9..07c3cdb0d 100644
--- a/providers/deepinfra/models/deepseek-ai/DeepSeek-V3.2.toml
+++ b/providers/deepinfra/models/deepseek-ai/DeepSeek-V3.2.toml
@@ -3,7 +3,7 @@ release_date = "2025-12-02"
 last_updated = "2025-12-02"
 attachment = false
 reasoning = true
-reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high"] }]
+reasoning_options = [{ type = "toggle" }]
 temperature = true
 knowledge = "2024-12"
 tool_call = true
diff --git a/providers/deepinfra/models/deepseek-ai/DeepSeek-V4-Flash.toml b/providers/deepinfra/models/deepseek-ai/DeepSeek-V4-Flash.toml
index 7fddc8db7..5a1b7b870 100644
--- a/providers/deepinfra/models/deepseek-ai/DeepSeek-V4-Flash.toml
+++ b/providers/deepinfra/models/deepseek-ai/DeepSeek-V4-Flash.toml
@@ -1,5 +1,5 @@
 base_model = "deepseek/deepseek-v4-flash"
-reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high"] }]
+reasoning_options = [{ type = "effort", values = ["none", "high", "max"] }]
 
 [interleaved]
 field = "reasoning_content"
diff --git a/providers/deepinfra/models/deepseek-ai/DeepSeek-V4-Pro.toml b/providers/deepinfra/models/deepseek-ai/DeepSeek-V4-Pro.toml
index ad77c847b..b371b586b 100644
--- a/providers/deepinfra/models/deepseek-ai/DeepSeek-V4-Pro.toml
+++ b/providers/deepinfra/models/deepseek-ai/DeepSeek-V4-Pro.toml
@@ -1,5 +1,5 @@
 base_model = "deepseek/deepseek-v4-pro"
-reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high"] }]
+reasoning_options = [{ type = "effort", values = ["none", "high", "max"] }]
 
 [interleaved]
 field = "reasoning_content"
diff --git a/providers/deepinfra/models/google/gemma-4-26B-A4B-it.toml b/providers/deepinfra/models/google/gemma-4-26B-A4B-it.toml
index 811637712..ee3319c85 100644
--- a/providers/deepinfra/models/google/gemma-4-26B-A4B-it.toml
+++ b/providers/deepinfra/models/google/gemma-4-26B-A4B-it.toml
@@ -1,5 +1,5 @@
 base_model = "google/gemma-4-26b-a4b-it"
-reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high"] }]
+reasoning_options = [{ type = "toggle" }]
 
 [cost]
 input = 0.07
diff --git a/providers/deepinfra/models/google/gemma-4-31B-it.toml b/providers/deepinfra/models/google/gemma-4-31B-it.toml
index 53541f4cc..f9dc05e17 100644
--- a/providers/deepinfra/models/google/gemma-4-31B-it.toml
+++ b/providers/deepinfra/models/google/gemma-4-31B-it.toml
@@ -1,5 +1,5 @@
 base_model = "google/gemma-4-31b-it"
-reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high"] }]
+reasoning_options = [{ type = "toggle" }]
 
 [cost]
 input = 0.13
diff --git a/providers/deepinfra/models/meta-llama/Llama-4-Maverick-17B-128E-Instruct-FP8.toml b/providers/deepinfra/models/meta-llama/Llama-4-Maverick-17B-128E-Instruct-FP8.toml
index 1a2b88bbc..e56d05fac 100644
--- a/providers/deepinfra/models/meta-llama/Llama-4-Maverick-17B-128E-Instruct-FP8.toml
+++ b/providers/deepinfra/models/meta-llama/Llama-4-Maverick-17B-128E-Instruct-FP8.toml
@@ -4,7 +4,7 @@ release_date = "2025-04-05"
 last_updated = "2025-04-05"
 attachment = false
 reasoning = false
-tool_call = true
+tool_call = false
 open_weights = true
 
 [cost]
@@ -12,7 +12,7 @@ input = 0.15
 output = 0.60
 
 [limit]
-context = 1_000_000
+context = 1_048_576
 output = 16_384
 
 [modalities]
diff --git a/providers/deepinfra/models/meta-llama/Llama-4-Scout-17B-16E-Instruct.toml b/providers/deepinfra/models/meta-llama/Llama-4-Scout-17B-16E-Instruct.toml
index 8f7bab10f..12d7d50db 100644
--- a/providers/deepinfra/models/meta-llama/Llama-4-Scout-17B-16E-Instruct.toml
+++ b/providers/deepinfra/models/meta-llama/Llama-4-Scout-17B-16E-Instruct.toml
@@ -8,11 +8,11 @@ tool_call = true
 open_weights = true
 
 [cost]
-input = 0.08
+input = 0.10
 output = 0.30
 
 [limit]
-context = 10_000_000
+context = 327_680
 output = 16_384
 
 [modalities]
diff --git a/providers/deepinfra/models/moonshotai/Kimi-K2.5.toml b/providers/deepinfra/models/moonshotai/Kimi-K2.5.toml
index bc030925f..5cb0e8a79 100644
--- a/providers/deepinfra/models/moonshotai/Kimi-K2.5.toml
+++ b/providers/deepinfra/models/moonshotai/Kimi-K2.5.toml
@@ -4,7 +4,7 @@ release_date = "2026-01-27"
 last_updated = "2026-01-27"
 attachment = true
 reasoning = true
-reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high"] }]
+reasoning_options = [{ type = "toggle" }]
 structured_output = true
 temperature = true
 tool_call = true
@@ -12,9 +12,9 @@ knowledge = "2025-01"
 open_weights = true
 
 [cost]
-input = 0.50
-output = 2.80
-cached_input = 0.09
+input = 0.45
+output = 2.25
+cached_input = 0.07
 
 [limit]
 context = 262_144
diff --git a/providers/deepinfra/models/moonshotai/Kimi-K2.6.toml b/providers/deepinfra/models/moonshotai/Kimi-K2.6.toml
index a5b127070..60a745ac7 100644
--- a/providers/deepinfra/models/moonshotai/Kimi-K2.6.toml
+++ b/providers/deepinfra/models/moonshotai/Kimi-K2.6.toml
@@ -4,7 +4,7 @@ release_date = "2026-04-21"
 last_updated = "2026-04-21"
 attachment = true
 reasoning = true
-reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high"] }]
+reasoning_options = [{ type = "toggle" }]
 structured_output = true
 temperature = true
 tool_call = true
diff --git a/providers/deepinfra/models/openai/gpt-oss-120b.toml b/providers/deepinfra/models/openai/gpt-oss-120b.toml
index 7a713c352..734ae2d0f 100644
--- a/providers/deepinfra/models/openai/gpt-oss-120b.toml
+++ b/providers/deepinfra/models/openai/gpt-oss-120b.toml
@@ -6,14 +6,14 @@ release_date = "2025-08-05"
 last_updated = "2025-08-05"
 attachment = false
 reasoning = true
-reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high"] }]
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }]
 temperature = true
 tool_call = true
 open_weights = true
 
 [cost]
-input = 0.05
-output = 0.24
+input = 0.039
+output = 0.19
 
 [limit]
 context = 131_072
diff --git a/providers/deepinfra/models/openai/gpt-oss-20b.toml b/providers/deepinfra/models/openai/gpt-oss-20b.toml
index d7166c7c5..eebce8f2e 100644
--- a/providers/deepinfra/models/openai/gpt-oss-20b.toml
+++ b/providers/deepinfra/models/openai/gpt-oss-20b.toml
@@ -6,7 +6,7 @@ release_date = "2025-08-05"
 last_updated = "2025-08-05"
 attachment = false
 reasoning = true
-reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high"] }]
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }]
 temperature = true
 tool_call = true
 open_weights = true
diff --git a/providers/deepinfra/models/zai-org/GLM-4.6.toml b/providers/deepinfra/models/zai-org/GLM-4.6.toml
index 2352f5451..1485de47a 100644
--- a/providers/deepinfra/models/zai-org/GLM-4.6.toml
+++ b/providers/deepinfra/models/zai-org/GLM-4.6.toml
@@ -6,7 +6,7 @@ last_updated = "2025-09-30"
 knowledge = "2025-04"
 attachment = false
 reasoning = true
-reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high"] }]
+reasoning_options = [{ type = "toggle" }]
 temperature = true
 tool_call = true
 open_weights = true
@@ -20,7 +20,7 @@ output = 1.74
 cache_read = 0.08
 
 [limit]
-context = 204_800
+context = 202_752
 output = 131_072
 
 [modalities]
diff --git a/providers/deepinfra/models/zai-org/GLM-4.7-Flash.toml b/providers/deepinfra/models/zai-org/GLM-4.7-Flash.toml
index d3dba78bd..b64fc596e 100644
--- a/providers/deepinfra/models/zai-org/GLM-4.7-Flash.toml
+++ b/providers/deepinfra/models/zai-org/GLM-4.7-Flash.toml
@@ -6,7 +6,7 @@ last_updated = "2026-01-19"
 knowledge = "2025-04"
 attachment = false
 reasoning = true
-reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high"] }]
+reasoning_options = [{ type = "toggle" }]
 temperature = true
 tool_call = true
 open_weights = true
diff --git a/providers/deepinfra/models/zai-org/GLM-4.7.toml b/providers/deepinfra/models/zai-org/GLM-4.7.toml
index b45850427..f90cf3741 100644
--- a/providers/deepinfra/models/zai-org/GLM-4.7.toml
+++ b/providers/deepinfra/models/zai-org/GLM-4.7.toml
@@ -6,7 +6,7 @@ last_updated = "2025-12-22"
 knowledge = "2025-04"
 attachment = false
 reasoning = true
-reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high"] }]
+reasoning_options = [{ type = "toggle" }]
 temperature = true
 tool_call = true
 open_weights = true
@@ -15,7 +15,7 @@ open_weights = true
 field = "reasoning_content"
 
 [cost]
-input = 0.43
+input = 0.40
 output = 1.75
 cache_read = 0.08
 
diff --git a/providers/deepinfra/models/zai-org/GLM-5.1.toml b/providers/deepinfra/models/zai-org/GLM-5.1.toml
index 2994ed6b8..8ec261580 100644
--- a/providers/deepinfra/models/zai-org/GLM-5.1.toml
+++ b/providers/deepinfra/models/zai-org/GLM-5.1.toml
@@ -6,7 +6,7 @@ last_updated = "2026-04-07"
 knowledge = "2025-04"
 attachment = false
 reasoning = true
-reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high"] }]
+reasoning_options = [{ type = "toggle" }]
 temperature = true
 tool_call = true
 open_weights = true
@@ -16,9 +16,9 @@ structured_output = true
 field = "reasoning_content"
 
 [cost]
-input = 1.4
-output = 4.4
-cache_read = 0.26
+input = 1.05
+output = 3.50
+cache_read = 0.205
 
 [limit]
 context = 202_752 
diff --git a/providers/deepinfra/models/zai-org/GLM-5.toml b/providers/deepinfra/models/zai-org/GLM-5.toml
index 00d9de73b..366f5fe88 100644
--- a/providers/deepinfra/models/zai-org/GLM-5.toml
+++ b/providers/deepinfra/models/zai-org/GLM-5.toml
@@ -6,7 +6,7 @@ last_updated = "2026-02-12"
 knowledge = "2025-12"
 attachment = false
 reasoning = true
-reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high"] }]
+reasoning_options = [{ type = "toggle" }]
 temperature = true
 tool_call = true
 open_weights = true
@@ -15,9 +15,9 @@ open_weights = true
 field = "reasoning_content"
 
 [cost]
-input = 0.8
-output = 2.56
-cache_read = 0.16
+input = 0.60
+output = 2.08
+cache_read = 0.12
 
 [limit]
 context = 202_752 

From f8229eba2e5854e55ec47acbcf9ae729fbc128d6 Mon Sep 17 00:00:00 2001
From: Aiden Cline <aidenpcline@gmail.com>
Date: Wed, 10 Jun 2026 20:07:25 -0500
Subject: [PATCH 3/4] fix(deepinfra): narrow reasoning controls

---
 packages/core/test/deepinfra.test.ts          | 30 +++++++++----------
 .../models/MiniMaxAI/MiniMax-M2.5.toml        |  1 -
 .../models/Qwen/Qwen3.5-35B-A3B.toml          |  1 -
 .../models/Qwen/Qwen3.5-397B-A17B.toml        |  1 -
 .../models/Qwen/Qwen3.6-35B-A3B.toml          |  1 -
 .../models/deepseek-ai/DeepSeek-V4-Flash.toml |  2 +-
 .../models/deepseek-ai/DeepSeek-V4-Pro.toml   |  2 +-
 .../models/moonshotai/Kimi-K2.5.toml          |  1 -
 .../models/zai-org/GLM-4.7-Flash.toml         |  1 -
 9 files changed, 16 insertions(+), 24 deletions(-)

diff --git a/packages/core/test/deepinfra.test.ts b/packages/core/test/deepinfra.test.ts
index fda06b97f..d3b4be726 100644
--- a/packages/core/test/deepinfra.test.ts
+++ b/packages/core/test/deepinfra.test.ts
@@ -9,26 +9,20 @@ test("DeepInfra models expose only verified reasoning controls", async () => {
 
   expect(deepinfra).toBeDefined();
 
-  const fixed = ["MiniMaxAI/MiniMax-M2.5"];
   const toggle = [
-    "Qwen/Qwen3.5-35B-A3B",
-    "Qwen/Qwen3.5-397B-A17B",
-    "Qwen/Qwen3.6-35B-A3B",
     "XiaomiMiMo/MiMo-V2.5-Pro",
     "XiaomiMiMo/MiMo-V2.5",
     "deepseek-ai/DeepSeek-V3.2",
     "google/gemma-4-26B-A4B-it",
     "google/gemma-4-31B-it",
-    "moonshotai/Kimi-K2.5",
     "moonshotai/Kimi-K2.6",
     "zai-org/GLM-4.6",
-    "zai-org/GLM-4.7-Flash",
     "zai-org/GLM-4.7",
     "zai-org/GLM-5.1",
     "zai-org/GLM-5",
   ];
   const standardEffort = ["openai/gpt-oss-120b", "openai/gpt-oss-20b"];
-  const v4Effort = [
+  const v4 = [
     "deepseek-ai/DeepSeek-V4-Flash",
     "deepseek-ai/DeepSeek-V4-Pro",
   ];
@@ -38,23 +32,26 @@ test("DeepInfra models expose only verified reasoning controls", async () => {
     "meta-llama/Llama-4-Maverick-17B-128E-Instruct-FP8",
     "meta-llama/Llama-4-Scout-17B-16E-Instruct",
   ];
-  const unresolved = ["deepseek-ai/DeepSeek-R1-0528"];
+  const unresolved = [
+    "MiniMaxAI/MiniMax-M2.5",
+    "Qwen/Qwen3.5-35B-A3B",
+    "Qwen/Qwen3.5-397B-A17B",
+    "Qwen/Qwen3.6-35B-A3B",
+    "deepseek-ai/DeepSeek-R1-0528",
+    "moonshotai/Kimi-K2.5",
+    "zai-org/GLM-4.7-Flash",
+  ];
 
   expect(Object.keys(deepinfra?.models ?? {}).sort()).toEqual(
     [
-      ...fixed,
       ...toggle,
       ...standardEffort,
-      ...v4Effort,
+      ...v4,
       ...nonReasoning,
       ...unresolved,
     ].sort(),
   );
 
-  for (const id of fixed) {
-    expect(deepinfra?.models[id]?.reasoning).toBe(true);
-    expect(deepinfra?.models[id]?.reasoning_options).toEqual([]);
-  }
   for (const id of toggle) {
     expect(deepinfra?.models[id]?.reasoning).toBe(true);
     expect(deepinfra?.models[id]?.reasoning_options).toEqual([{ type: "toggle" }]);
@@ -65,10 +62,11 @@ test("DeepInfra models expose only verified reasoning controls", async () => {
       { type: "effort", values: ["low", "medium", "high"] },
     ]);
   }
-  for (const id of v4Effort) {
+  for (const id of v4) {
     expect(deepinfra?.models[id]?.reasoning).toBe(true);
     expect(deepinfra?.models[id]?.reasoning_options).toEqual([
-      { type: "effort", values: ["none", "high", "max"] },
+      { type: "toggle" },
+      { type: "effort", values: ["high"] },
     ]);
   }
   for (const id of nonReasoning) {
diff --git a/providers/deepinfra/models/MiniMaxAI/MiniMax-M2.5.toml b/providers/deepinfra/models/MiniMaxAI/MiniMax-M2.5.toml
index 57bc81952..0c49851d3 100644
--- a/providers/deepinfra/models/MiniMaxAI/MiniMax-M2.5.toml
+++ b/providers/deepinfra/models/MiniMaxAI/MiniMax-M2.5.toml
@@ -5,7 +5,6 @@ release_date = "2026-02-12"
 last_updated = "2026-02-12"
 attachment = false
 reasoning = true
-reasoning_options = []
 temperature = true
 tool_call = true
 knowledge = "2025-06"
diff --git a/providers/deepinfra/models/Qwen/Qwen3.5-35B-A3B.toml b/providers/deepinfra/models/Qwen/Qwen3.5-35B-A3B.toml
index f89e336f2..3475b9525 100644
--- a/providers/deepinfra/models/Qwen/Qwen3.5-35B-A3B.toml
+++ b/providers/deepinfra/models/Qwen/Qwen3.5-35B-A3B.toml
@@ -4,7 +4,6 @@ release_date = "2026-02-01"
 last_updated = "2026-04-20"
 attachment = true
 reasoning = true
-reasoning_options = [{ type = "toggle" }]
 temperature = true
 knowledge = "2025-01"
 tool_call = true
diff --git a/providers/deepinfra/models/Qwen/Qwen3.5-397B-A17B.toml b/providers/deepinfra/models/Qwen/Qwen3.5-397B-A17B.toml
index 427846ef5..10c8fa71c 100644
--- a/providers/deepinfra/models/Qwen/Qwen3.5-397B-A17B.toml
+++ b/providers/deepinfra/models/Qwen/Qwen3.5-397B-A17B.toml
@@ -4,7 +4,6 @@ release_date = "2026-02-01"
 last_updated = "2026-04-20"
 attachment = true
 reasoning = true
-reasoning_options = [{ type = "toggle" }]
 temperature = true
 knowledge = "2025-01"
 tool_call = true
diff --git a/providers/deepinfra/models/Qwen/Qwen3.6-35B-A3B.toml b/providers/deepinfra/models/Qwen/Qwen3.6-35B-A3B.toml
index 9feaa1218..5c1c94a60 100644
--- a/providers/deepinfra/models/Qwen/Qwen3.6-35B-A3B.toml
+++ b/providers/deepinfra/models/Qwen/Qwen3.6-35B-A3B.toml
@@ -4,7 +4,6 @@ release_date = "2026-04-01"
 last_updated = "2026-04-01"
 attachment = true
 reasoning = true
-reasoning_options = [{ type = "toggle" }]
 temperature = true
 tool_call = true
 open_weights = true
diff --git a/providers/deepinfra/models/deepseek-ai/DeepSeek-V4-Flash.toml b/providers/deepinfra/models/deepseek-ai/DeepSeek-V4-Flash.toml
index 5a1b7b870..462a42e56 100644
--- a/providers/deepinfra/models/deepseek-ai/DeepSeek-V4-Flash.toml
+++ b/providers/deepinfra/models/deepseek-ai/DeepSeek-V4-Flash.toml
@@ -1,5 +1,5 @@
 base_model = "deepseek/deepseek-v4-flash"
-reasoning_options = [{ type = "effort", values = ["none", "high", "max"] }]
+reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["high"] }]
 
 [interleaved]
 field = "reasoning_content"
diff --git a/providers/deepinfra/models/deepseek-ai/DeepSeek-V4-Pro.toml b/providers/deepinfra/models/deepseek-ai/DeepSeek-V4-Pro.toml
index b371b586b..d576e311d 100644
--- a/providers/deepinfra/models/deepseek-ai/DeepSeek-V4-Pro.toml
+++ b/providers/deepinfra/models/deepseek-ai/DeepSeek-V4-Pro.toml
@@ -1,5 +1,5 @@
 base_model = "deepseek/deepseek-v4-pro"
-reasoning_options = [{ type = "effort", values = ["none", "high", "max"] }]
+reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["high"] }]
 
 [interleaved]
 field = "reasoning_content"
diff --git a/providers/deepinfra/models/moonshotai/Kimi-K2.5.toml b/providers/deepinfra/models/moonshotai/Kimi-K2.5.toml
index 5cb0e8a79..e264128e6 100644
--- a/providers/deepinfra/models/moonshotai/Kimi-K2.5.toml
+++ b/providers/deepinfra/models/moonshotai/Kimi-K2.5.toml
@@ -4,7 +4,6 @@ release_date = "2026-01-27"
 last_updated = "2026-01-27"
 attachment = true
 reasoning = true
-reasoning_options = [{ type = "toggle" }]
 structured_output = true
 temperature = true
 tool_call = true
diff --git a/providers/deepinfra/models/zai-org/GLM-4.7-Flash.toml b/providers/deepinfra/models/zai-org/GLM-4.7-Flash.toml
index b64fc596e..dbb42a2d8 100644
--- a/providers/deepinfra/models/zai-org/GLM-4.7-Flash.toml
+++ b/providers/deepinfra/models/zai-org/GLM-4.7-Flash.toml
@@ -6,7 +6,6 @@ last_updated = "2026-01-19"
 knowledge = "2025-04"
 attachment = false
 reasoning = true
-reasoning_options = [{ type = "toggle" }]
 temperature = true
 tool_call = true
 open_weights = true

From b618a32341457dd2b8202ae8bcae2598dc05d1df Mon Sep 17 00:00:00 2001
From: Aiden Cline <aidenpcline@gmail.com>
Date: Wed, 10 Jun 2026 20:13:26 -0500
Subject: [PATCH 4/4] fix(deepinfra): verify R1 reasoning controls

---
 packages/core/test/deepinfra.test.ts                   | 10 +++++++++-
 .../deepinfra/models/deepseek-ai/DeepSeek-R1-0528.toml |  1 +
 2 files changed, 10 insertions(+), 1 deletion(-)

diff --git a/packages/core/test/deepinfra.test.ts b/packages/core/test/deepinfra.test.ts
index d3b4be726..46833b8ac 100644
--- a/packages/core/test/deepinfra.test.ts
+++ b/packages/core/test/deepinfra.test.ts
@@ -22,6 +22,7 @@ test("DeepInfra models expose only verified reasoning controls", async () => {
     "zai-org/GLM-5",
   ];
   const standardEffort = ["openai/gpt-oss-120b", "openai/gpt-oss-20b"];
+  const r1 = ["deepseek-ai/DeepSeek-R1-0528"];
   const v4 = [
     "deepseek-ai/DeepSeek-V4-Flash",
     "deepseek-ai/DeepSeek-V4-Pro",
@@ -37,7 +38,6 @@ test("DeepInfra models expose only verified reasoning controls", async () => {
     "Qwen/Qwen3.5-35B-A3B",
     "Qwen/Qwen3.5-397B-A17B",
     "Qwen/Qwen3.6-35B-A3B",
-    "deepseek-ai/DeepSeek-R1-0528",
     "moonshotai/Kimi-K2.5",
     "zai-org/GLM-4.7-Flash",
   ];
@@ -46,6 +46,7 @@ test("DeepInfra models expose only verified reasoning controls", async () => {
     [
       ...toggle,
       ...standardEffort,
+      ...r1,
       ...v4,
       ...nonReasoning,
       ...unresolved,
@@ -62,6 +63,13 @@ test("DeepInfra models expose only verified reasoning controls", async () => {
       { type: "effort", values: ["low", "medium", "high"] },
     ]);
   }
+  for (const id of r1) {
+    expect(deepinfra?.models[id]?.reasoning).toBe(true);
+    expect(deepinfra?.models[id]?.reasoning_options).toEqual([
+      { type: "toggle" },
+      { type: "effort", values: ["low", "medium", "high"] },
+    ]);
+  }
   for (const id of v4) {
     expect(deepinfra?.models[id]?.reasoning).toBe(true);
     expect(deepinfra?.models[id]?.reasoning_options).toEqual([
diff --git a/providers/deepinfra/models/deepseek-ai/DeepSeek-R1-0528.toml b/providers/deepinfra/models/deepseek-ai/DeepSeek-R1-0528.toml
index 2b4bc6be5..7e9e7b96e 100644
--- a/providers/deepinfra/models/deepseek-ai/DeepSeek-R1-0528.toml
+++ b/providers/deepinfra/models/deepseek-ai/DeepSeek-R1-0528.toml
@@ -3,6 +3,7 @@ release_date = "2025-05-28"
 last_updated = "2025-05-28"
 attachment = false
 reasoning = true
+reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high"] }]
 temperature = true
 knowledge = "2024-07"
 tool_call = true