From 77f63b79800d16080bca167ff8c0c1a7cd4c7c03 Mon Sep 17 00:00:00 2001
From: moonmath <zro@moonmath.ai>
Date: Tue, 11 Aug 2026 08:31:52 +0000
Subject: [PATCH 1/8] feat: add Zro provider (OpenAI-compatible gateway)

Adds the Zro provider (zro.moonmath.ai), an OpenAI-compatible gateway
fronting LiteLLM->SGLang, serving three open-weight models:

- GLM-5.2             none|high|max effort, interleaved, 524K ctx / 64K out
- DeepSeek V4 Flash   none|high effort, interleaved, 1M ctx / 384K out
- Kimi K3             low|high|max effort, interleaved, 1M ctx / 1M out

Pricing/model set and reasoning effort tiers sourced from the product
bundled catalog and https://zro.moonmath.ai/pricing (2026-08-11).
Interleaved reasoning via choices[...].message.reasoning_content.
---
 providers/zro/logo.svg                        | 20 +++++++++++++++++++
 .../zro/models/deepseek-v4-flash-0731.toml    | 15 ++++++++++++++
 providers/zro/models/glm-5.2.toml             | 16 +++++++++++++++
 providers/zro/models/kimi-k3.toml             | 16 +++++++++++++++
 providers/zro/provider.toml                   |  5 +++++
 5 files changed, 72 insertions(+)
 create mode 100644 providers/zro/logo.svg
 create mode 100644 providers/zro/models/deepseek-v4-flash-0731.toml
 create mode 100644 providers/zro/models/glm-5.2.toml
 create mode 100644 providers/zro/models/kimi-k3.toml
 create mode 100644 providers/zro/provider.toml

diff --git a/providers/zro/logo.svg b/providers/zro/logo.svg
new file mode 100644
index 00000000000..3aa2236ba67
--- /dev/null
+++ b/providers/zro/logo.svg
@@ -0,0 +1,20 @@
+<?xml version="1.0" encoding="UTF-8"?>
+<svg version="1.1" xmlns="http://www.w3.org/2000/svg" width="24" height="24" viewBox="0 0 24 24">
+<g fill="currentColor" transform="scale(0.09375)">
+<path d="M0 0 C26.4 0 52.8 0 80 0 C80 9.24 80 18.48 80 28 C53.6 28 27.2 28 0 28 C0 18.76 0 9.52 0 0 Z " transform="translate(88,200)"/>
+<path d="M0 0 C26.4 0 52.8 0 80 0 C80 9.24 80 18.48 80 28 C53.6 28 27.2 28 0 28 C0 18.76 0 9.52 0 0 Z " transform="translate(88,28)"/>
+<path d="M0 0 C17.82 0 35.64 0 54 0 C54 9.24 54 18.48 54 28 C36.18 28 18.36 28 0 28 C0 18.76 0 9.52 0 0 Z " transform="translate(165,114)"/>
+<path d="M0 0 C17.82 0 35.64 0 54 0 C54 9.24 54 18.48 54 28 C36.18 28 18.36 28 0 28 C0 18.76 0 9.52 0 0 Z " transform="translate(37,114)"/>
+<path d="M0 0 C1.03777588 -0.00315216 2.07555176 -0.00630432 3.14477539 -0.009552 C4.83848267 -0.00651566 4.83848267 -0.00651566 6.56640625 -0.00341797 C7.71681396 -0.00437469 8.86722168 -0.00533142 10.05249023 -0.00631714 C12.49039769 -0.00699916 14.92830695 -0.00514337 17.36621094 -0.00097656 C21.11267143 0.00437136 24.85901202 -0.00092086 28.60546875 -0.00732422 C30.96744818 -0.00666344 33.32942754 -0.00538229 35.69140625 -0.00341797 C36.82054443 -0.0054422 37.94968262 -0.00746643 39.11303711 -0.009552 C40.15081299 -0.00639984 41.18858887 -0.00324768 42.2578125 0 C43.1763501 0.00079559 44.0948877 0.00159119 45.04125977 0.00241089 C47.12890625 0.12939453 47.12890625 0.12939453 48.12890625 1.12939453 C48.12890625 10.03939453 48.12890625 18.94939453 48.12890625 28.12939453 C30.30890625 28.12939453 12.48890625 28.12939453 -5.87109375 28.12939453 C-5.87109375 0.00508526 -5.87109375 0.00508526 0 0 Z " transform="translate(68.87109375,170.87060546875)"/>
+<path d="M0 0 C17.82 0 35.64 0 54 0 C54 8.91 54 17.82 54 27 C52.34591544 28.65408456 50.40446108 28.12742355 48.12890625 28.12939453 C46.57224243 28.13412277 46.57224243 28.13412277 44.98413086 28.13894653 C43.29042358 28.13591019 43.29042358 28.13591019 41.5625 28.1328125 C39.83688843 28.13424759 39.83688843 28.13424759 38.07641602 28.13571167 C35.63850856 28.13639369 33.2005993 28.1345379 30.76269531 28.13037109 C27.01623482 28.12502317 23.26989423 28.13031539 19.5234375 28.13671875 C17.16145807 28.13605797 14.79947871 28.13477682 12.4375 28.1328125 C11.30836182 28.13483673 10.17922363 28.13686096 9.01586914 28.13894653 C7.45920532 28.13421829 7.45920532 28.13421829 5.87109375 28.12939453 C4.95255615 28.12859894 4.03401855 28.12780334 3.08764648 28.12698364 C1 28 1 28 0 27 C0 18.09 0 9.18 0 0 Z " transform="translate(139,57)"/>
+<path d="M0 0 C17.82 0 35.64 0 54 0 C54 8.91 54 17.82 54 27 C36.18 27 18.36 27 0 27 C0 18.09 0 9.18 0 0 Z " transform="translate(165,143)"/>
+<path d="M0 0 C17.82 0 35.64 0 54 0 C54 8.91 54 17.82 54 27 C36.18 27 18.36 27 0 27 C0 18.09 0 9.18 0 0 Z " transform="translate(165,86)"/>
+<path d="M0 0 C17.82 0 35.64 0 54 0 C54 8.91 54 17.82 54 27 C36.18 27 18.36 27 0 27 C0 18.09 0 9.18 0 0 Z " transform="translate(37,86)"/>
+<path d="M0 0 C7.52974139 -0.02290222 15.05936291 -0.04283138 22.58911133 -0.05493164 C25.15291757 -0.0599737 27.71672091 -0.06680714 30.28051758 -0.07543945 C33.95655021 -0.08751064 37.63254498 -0.09322942 41.30859375 -0.09765625 C42.46274277 -0.10281754 43.61689178 -0.10797882 44.80601501 -0.11329651 C46.3975515 -0.11340981 46.3975515 -0.11340981 48.02124023 -0.11352539 C48.96055893 -0.115746 49.89987762 -0.11796661 50.86766052 -0.12025452 C53 0 53 0 54 1 C54 9.58 54 18.16 54 27 C36.18 27 18.36 27 0 27 C0 18.09 0 9.18 0 0 Z " transform="translate(37,143)"/>
+<path d="M0 0 C8.91 0 17.82 0 27 0 C27 9.24 27 18.48 27 28 C18.09 28 9.18 28 0 28 C0 18.76 0 9.52 0 0 Z " transform="translate(166,171)"/>
+<path d="M0 0 C8.91 0 17.82 0 27 0 C27 9.24 27 18.48 27 28 C18.09 28 9.18 28 0 28 C0 18.76 0 9.52 0 0 Z " transform="translate(63,57)"/>
+<path d="M0 0 C4.95 0 9.9 0 15 0 C15 8.91 15 17.82 15 27 C10.05 27 5.1 27 0 27 C0 18.09 0 9.18 0 0 Z " transform="translate(107,143)"/>
+<path d="M0 0 C4.95 0 9.9 0 15 0 C15 8.91 15 17.82 15 27 C10.05 27 5.1 27 0 27 C0 18.09 0 9.18 0 0 Z " transform="translate(134,86)"/>
+<path d="M0 0 C4.62 0 9.24 0 14 0 C14 9.24 14 18.48 14 28 C9.38 28 4.76 28 0 28 C0 18.76 0 9.52 0 0 Z " transform="translate(121,114)"/>
+</g>
+</svg>
diff --git a/providers/zro/models/deepseek-v4-flash-0731.toml b/providers/zro/models/deepseek-v4-flash-0731.toml
new file mode 100644
index 00000000000..76cdaae46d0
--- /dev/null
+++ b/providers/zro/models/deepseek-v4-flash-0731.toml
@@ -0,0 +1,15 @@
+# DeepSeek V4 Flash; exposes reasoning_effort = none|high (none disables; high is the only tier).
+# Limits 1M context / 384K output. Pricing: https://zro.moonmath.ai/pricing
+base_model = "deepseek/deepseek-v4-flash-0731"
+reasoning_options = [{ type = "effort", values = ["none", "high"] }]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.14
+output = 0.28
+cache_read = 0.0028
+
+[limit]
+context = 1_048_576
\ No newline at end of file
diff --git a/providers/zro/models/glm-5.2.toml b/providers/zro/models/glm-5.2.toml
new file mode 100644
index 00000000000..846253a8168
--- /dev/null
+++ b/providers/zro/models/glm-5.2.toml
@@ -0,0 +1,16 @@
+# GLM-5.2; exposes reasoning_effort = none|high|max (none disables; high/max tiers).
+# Pricing: https://zro.moonmath.ai/pricing
+base_model = "zhipuai/glm-5.2"
+reasoning_options = [{ type = "effort", values = ["none", "high", "max"] }]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 1.10
+output = 4.00
+cache_read = 0.20
+
+[limit]
+context = 524_288
+output = 64_000
\ No newline at end of file
diff --git a/providers/zro/models/kimi-k3.toml b/providers/zro/models/kimi-k3.toml
new file mode 100644
index 00000000000..01090d7b6cc
--- /dev/null
+++ b/providers/zro/models/kimi-k3.toml
@@ -0,0 +1,16 @@
+# Kimi K3; exposes reasoning_effort = low|high|max (no off level).
+# Limits 1M context / 1M output. Pricing: https://zro.moonmath.ai/pricing
+base_model = "moonshotai/kimi-k3"
+reasoning_options = [{ type = "effort", values = ["low", "high", "max"] }]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 2.50
+output = 12.00
+cache_read = 0.45
+
+[limit]
+context = 1_048_576
+output = 1_048_576
\ No newline at end of file
diff --git a/providers/zro/provider.toml b/providers/zro/provider.toml
new file mode 100644
index 00000000000..944889964f2
--- /dev/null
+++ b/providers/zro/provider.toml
@@ -0,0 +1,5 @@
+name = "Zro"
+env = ["ZRO_API_KEY"]
+npm = "@ai-sdk/openai-compatible"
+api = "https://zro.moonmath.ai/v1"
+doc = "https://zro.moonmath.ai/pricing"
\ No newline at end of file

From f99fcc1f4c98ae5821c62c454b69cf9bed9c85ba Mon Sep 17 00:00:00 2001
From: Jeremy Felder <jeremy.felder1@gmail.com>
Date: Tue, 1 Sep 2026 14:50:27 +0300
Subject: [PATCH 2/8] fix(provider): rework Zro logo to clean viewBox without
 fixed dimensions

---
 providers/zro/logo.svg | 5 +----
 1 file changed, 1 insertion(+), 4 deletions(-)

diff --git a/providers/zro/logo.svg b/providers/zro/logo.svg
index 3aa2236ba67..440252733ac 100644
--- a/providers/zro/logo.svg
+++ b/providers/zro/logo.svg
@@ -1,6 +1,4 @@
-<?xml version="1.0" encoding="UTF-8"?>
-<svg version="1.1" xmlns="http://www.w3.org/2000/svg" width="24" height="24" viewBox="0 0 24 24">
-<g fill="currentColor" transform="scale(0.09375)">
+<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 256 256" fill="currentColor">
 <path d="M0 0 C26.4 0 52.8 0 80 0 C80 9.24 80 18.48 80 28 C53.6 28 27.2 28 0 28 C0 18.76 0 9.52 0 0 Z " transform="translate(88,200)"/>
 <path d="M0 0 C26.4 0 52.8 0 80 0 C80 9.24 80 18.48 80 28 C53.6 28 27.2 28 0 28 C0 18.76 0 9.52 0 0 Z " transform="translate(88,28)"/>
 <path d="M0 0 C17.82 0 35.64 0 54 0 C54 9.24 54 18.48 54 28 C36.18 28 18.36 28 0 28 C0 18.76 0 9.52 0 0 Z " transform="translate(165,114)"/>
@@ -16,5 +14,4 @@
 <path d="M0 0 C4.95 0 9.9 0 15 0 C15 8.91 15 17.82 15 27 C10.05 27 5.1 27 0 27 C0 18.09 0 9.18 0 0 Z " transform="translate(107,143)"/>
 <path d="M0 0 C4.95 0 9.9 0 15 0 C15 8.91 15 17.82 15 27 C10.05 27 5.1 27 0 27 C0 18.09 0 9.18 0 0 Z " transform="translate(134,86)"/>
 <path d="M0 0 C4.62 0 9.24 0 14 0 C14 9.24 14 18.48 14 28 C9.38 28 4.76 28 0 28 C0 18.76 0 9.52 0 0 Z " transform="translate(121,114)"/>
-</g>
 </svg>

From 03ba8e59980fbbbdcb37cdb1c37f6b8cb5d7eacd Mon Sep 17 00:00:00 2001
From: Jeremy Felder <jeremy.felder1@gmail.com>
Date: Tue, 1 Sep 2026 14:50:27 +0300
Subject: [PATCH 3/8] fix(provider): align Zro model reasoning/limits with lab
 baselines

- deepseek-v4-flash-0731: thinking toggle + effort low|high|max (matches first-party deepseek); add output limit
- kimi-k3: add thinking toggle; drop redundant context limit (base already 1M), keep output delta
- glm-5.2: effective effort high|max per zhipuai baseline
---
 providers/zro/models/deepseek-v4-flash-0731.toml | 14 +++++++++++---
 providers/zro/models/glm-5.2.toml                |  6 ++++--
 providers/zro/models/kimi-k3.toml                | 15 +++++++++++----
 3 files changed, 26 insertions(+), 9 deletions(-)

diff --git a/providers/zro/models/deepseek-v4-flash-0731.toml b/providers/zro/models/deepseek-v4-flash-0731.toml
index 76cdaae46d0..ea2a039d623 100644
--- a/providers/zro/models/deepseek-v4-flash-0731.toml
+++ b/providers/zro/models/deepseek-v4-flash-0731.toml
@@ -1,7 +1,14 @@
-# DeepSeek V4 Flash; exposes reasoning_effort = none|high (none disables; high is the only tier).
+# DeepSeek V4 Flash; OpenAI-compatible reasoning controls: thinking toggle + reasoning_effort low|high|max
+# (matches first-party providers/deepseek/models/deepseek-v4-flash.toml).
 # Limits 1M context / 384K output. Pricing: https://zro.moonmath.ai/pricing
 base_model = "deepseek/deepseek-v4-flash-0731"
-reasoning_options = [{ type = "effort", values = ["none", "high"] }]
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
 
 [interleaved]
 field = "reasoning_content"
@@ -12,4 +19,5 @@ output = 0.28
 cache_read = 0.0028
 
 [limit]
-context = 1_048_576
\ No newline at end of file
+context = 1_048_576
+output = 384_000
\ No newline at end of file
diff --git a/providers/zro/models/glm-5.2.toml b/providers/zro/models/glm-5.2.toml
index 846253a8168..58327173dff 100644
--- a/providers/zro/models/glm-5.2.toml
+++ b/providers/zro/models/glm-5.2.toml
@@ -1,7 +1,9 @@
-# GLM-5.2; exposes reasoning_effort = none|high|max (none disables; high/max tiers).
+# GLM-5.2; reasoning_effort effective levels high|max (default max).
+# low|medium map to high, xhigh maps to max; see providers/zhipuai/models/glm-5.2.toml.
+# `none` (off) is accepted via the effort field to disable thinking.
 # Pricing: https://zro.moonmath.ai/pricing
 base_model = "zhipuai/glm-5.2"
-reasoning_options = [{ type = "effort", values = ["none", "high", "max"] }]
+reasoning_options = [{ type = "effort", values = ["high", "max"] }]
 
 [interleaved]
 field = "reasoning_content"
diff --git a/providers/zro/models/kimi-k3.toml b/providers/zro/models/kimi-k3.toml
index 01090d7b6cc..84345ec4038 100644
--- a/providers/zro/models/kimi-k3.toml
+++ b/providers/zro/models/kimi-k3.toml
@@ -1,7 +1,15 @@
-# Kimi K3; exposes reasoning_effort = low|high|max (no off level).
-# Limits 1M context / 1M output. Pricing: https://zro.moonmath.ai/pricing
+# Kimi K3; OpenAI-compatible reasoning controls: thinking toggle + reasoning effort low|high|max
+# (matches providers/moonshotai/models/kimi-k3.toml and providers/neuralwatt/models/kimi-k3.toml).
+# Limits 1M context / 1M output (context matches base model; output is the delta).
+# Pricing: https://zro.moonmath.ai/pricing
 base_model = "moonshotai/kimi-k3"
-reasoning_options = [{ type = "effort", values = ["low", "high", "max"] }]
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
 
 [interleaved]
 field = "reasoning_content"
@@ -12,5 +20,4 @@ output = 12.00
 cache_read = 0.45
 
 [limit]
-context = 1_048_576
 output = 1_048_576
\ No newline at end of file

From e05f19cb79002d534c91401fa7da2d35d089316f Mon Sep 17 00:00:00 2001
From: Jeremy Felder <jeremy.felder1@gmail.com>
Date: Wed, 2 Sep 2026 09:30:31 +0300
Subject: [PATCH 4/8] fix(provider): catalog Zro host reasoning and add GLM-5.3

Use Zro's live reasoning_effort surface instead of lab toggles, add GLM-5.3 and Flash as override-only entries, and drop restated identical limits.
---
 .../zro/models/deepseek-v4-flash-0731.toml    | 14 +++----------
 providers/zro/models/glm-5.2.toml             | 12 +++++------
 providers/zro/models/glm-5.3-flash.toml       | 21 +++++++++++++++++++
 providers/zro/models/glm-5.3.toml             | 16 ++++++++++++++
 providers/zro/models/kimi-k3.toml             | 20 +++++++-----------
 providers/zro/provider.toml                   |  2 +-
 6 files changed, 55 insertions(+), 30 deletions(-)
 create mode 100644 providers/zro/models/glm-5.3-flash.toml
 create mode 100644 providers/zro/models/glm-5.3.toml

diff --git a/providers/zro/models/deepseek-v4-flash-0731.toml b/providers/zro/models/deepseek-v4-flash-0731.toml
index ea2a039d623..515cd1dadf2 100644
--- a/providers/zro/models/deepseek-v4-flash-0731.toml
+++ b/providers/zro/models/deepseek-v4-flash-0731.toml
@@ -1,14 +1,7 @@
-# DeepSeek V4 Flash; OpenAI-compatible reasoning controls: thinking toggle + reasoning_effort low|high|max
-# (matches first-party providers/deepseek/models/deepseek-v4-flash.toml).
-# Limits 1M context / 384K output. Pricing: https://zro.moonmath.ai/pricing
+# Effort: reasoning_effort = low|high|max (default high)
+# Pricing: https://zro.moonmath.ai/pricing (USD / 1M tokens)
 base_model = "deepseek/deepseek-v4-flash-0731"
-
-[[reasoning_options]]
-type = "toggle"
-
-[[reasoning_options]]
-type = "effort"
-values = ["low", "high", "max"]
+reasoning_options = [{ type = "effort", values = ["low", "high", "max"] }]
 
 [interleaved]
 field = "reasoning_content"
@@ -20,4 +13,3 @@ cache_read = 0.0028
 
 [limit]
 context = 1_048_576
-output = 384_000
\ No newline at end of file
diff --git a/providers/zro/models/glm-5.2.toml b/providers/zro/models/glm-5.2.toml
index 58327173dff..dafa74579aa 100644
--- a/providers/zro/models/glm-5.2.toml
+++ b/providers/zro/models/glm-5.2.toml
@@ -1,9 +1,9 @@
-# GLM-5.2; reasoning_effort effective levels high|max (default max).
-# low|medium map to high, xhigh maps to max; see providers/zhipuai/models/glm-5.2.toml.
-# `none` (off) is accepted via the effort field to disable thinking.
-# Pricing: https://zro.moonmath.ai/pricing
+# Effort: reasoning_effort = none|high|max (default max)
+# Off is effort=none. thinking.type=disabled is accepted as an alias for none.
+# low|medium map to high; xhigh maps to max.
+# Pricing: https://zro.moonmath.ai/pricing (USD / 1M tokens)
 base_model = "zhipuai/glm-5.2"
-reasoning_options = [{ type = "effort", values = ["high", "max"] }]
+reasoning_options = [{ type = "effort", values = ["none", "high", "max"] }]
 
 [interleaved]
 field = "reasoning_content"
@@ -15,4 +15,4 @@ cache_read = 0.20
 
 [limit]
 context = 524_288
-output = 64_000
\ No newline at end of file
+output = 64_000
diff --git a/providers/zro/models/glm-5.3-flash.toml b/providers/zro/models/glm-5.3-flash.toml
new file mode 100644
index 00000000000..f2835eddd2b
--- /dev/null
+++ b/providers/zro/models/glm-5.3-flash.toml
@@ -0,0 +1,21 @@
+# Effort: reasoning_effort = none|high|max (default max)
+# Off is effort=none. thinking.type=disabled is accepted as an alias for none.
+# Pricing: https://zro.moonmath.ai/pricing (USD / 1M tokens)
+base_model = "zhipuai/glm-5.3-flash"
+attachment = false
+reasoning_options = [{ type = "effort", values = ["none", "high", "max"] }]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.15
+output = 0.50
+cache_read = 0.03
+
+[limit]
+context = 1_048_576
+output = 64_000
+
+[modalities]
+input = ["text"]
diff --git a/providers/zro/models/glm-5.3.toml b/providers/zro/models/glm-5.3.toml
new file mode 100644
index 00000000000..7c59e143727
--- /dev/null
+++ b/providers/zro/models/glm-5.3.toml
@@ -0,0 +1,16 @@
+# Effort: reasoning_effort = none|high|max (default max)
+# Off is effort=none. thinking.type=disabled is accepted as an alias for none.
+# Pricing: https://zro.moonmath.ai/pricing (USD / 1M tokens)
+base_model = "zhipuai/glm-5.3"
+reasoning_options = [{ type = "effort", values = ["none", "high", "max"] }]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 1.40
+output = 4.40
+cache_read = 0.26
+
+[limit]
+context = 1_048_576
diff --git a/providers/zro/models/kimi-k3.toml b/providers/zro/models/kimi-k3.toml
index 84345ec4038..b47152f8c38 100644
--- a/providers/zro/models/kimi-k3.toml
+++ b/providers/zro/models/kimi-k3.toml
@@ -1,15 +1,8 @@
-# Kimi K3; OpenAI-compatible reasoning controls: thinking toggle + reasoning effort low|high|max
-# (matches providers/moonshotai/models/kimi-k3.toml and providers/neuralwatt/models/kimi-k3.toml).
-# Limits 1M context / 1M output (context matches base model; output is the delta).
-# Pricing: https://zro.moonmath.ai/pricing
+# Effort: reasoning_effort = low|high|max (default high)
+# Pricing: https://zro.moonmath.ai/pricing (USD / 1M tokens)
 base_model = "moonshotai/kimi-k3"
-
-[[reasoning_options]]
-type = "toggle"
-
-[[reasoning_options]]
-type = "effort"
-values = ["low", "high", "max"]
+temperature = true
+reasoning_options = [{ type = "effort", values = ["low", "high", "max"] }]
 
 [interleaved]
 field = "reasoning_content"
@@ -20,4 +13,7 @@ output = 12.00
 cache_read = 0.45
 
 [limit]
-output = 1_048_576
\ No newline at end of file
+output = 1_048_576
+
+[modalities]
+input = ["text", "image"]
diff --git a/providers/zro/provider.toml b/providers/zro/provider.toml
index 944889964f2..3b6d9ee77d0 100644
--- a/providers/zro/provider.toml
+++ b/providers/zro/provider.toml
@@ -2,4 +2,4 @@ name = "Zro"
 env = ["ZRO_API_KEY"]
 npm = "@ai-sdk/openai-compatible"
 api = "https://zro.moonmath.ai/v1"
-doc = "https://zro.moonmath.ai/pricing"
\ No newline at end of file
+doc = "https://zro.moonmath.ai/"

From 11129333a2a97d018727ef7bed00eae9dc855b82 Mon Sep 17 00:00:00 2001
From: Jeremy Felder <jeremy.felder1@gmail.com>
Date: Wed, 2 Sep 2026 11:42:44 +0300
Subject: [PATCH 5/8] chore: retrigger CI checks


From 75fa7a756e848615e301c5ca6c381f749588c5b6 Mon Sep 17 00:00:00 2001
From: Jeremy Felder <jeremy.felder1@gmail.com>
Date: Thu, 3 Sep 2026 10:09:50 +0300
Subject: [PATCH 6/8] fix(provider): align Zro reasoning_options with lab peers

Match first-party effort sets, drop host-specific none/off claims, and restore GLM-5.3 Flash vision from the lab.
---
 providers/zro/models/deepseek-v4-flash-0731.toml | 2 --
 providers/zro/models/glm-5.2.toml                | 6 +-----
 providers/zro/models/glm-5.3-flash.toml          | 9 +--------
 providers/zro/models/glm-5.3.toml                | 5 +----
 providers/zro/models/kimi-k3.toml                | 2 --
 5 files changed, 3 insertions(+), 21 deletions(-)

diff --git a/providers/zro/models/deepseek-v4-flash-0731.toml b/providers/zro/models/deepseek-v4-flash-0731.toml
index 515cd1dadf2..2866c706cdb 100644
--- a/providers/zro/models/deepseek-v4-flash-0731.toml
+++ b/providers/zro/models/deepseek-v4-flash-0731.toml
@@ -1,5 +1,3 @@
-# Effort: reasoning_effort = low|high|max (default high)
-# Pricing: https://zro.moonmath.ai/pricing (USD / 1M tokens)
 base_model = "deepseek/deepseek-v4-flash-0731"
 reasoning_options = [{ type = "effort", values = ["low", "high", "max"] }]
 
diff --git a/providers/zro/models/glm-5.2.toml b/providers/zro/models/glm-5.2.toml
index dafa74579aa..b837ce22598 100644
--- a/providers/zro/models/glm-5.2.toml
+++ b/providers/zro/models/glm-5.2.toml
@@ -1,9 +1,5 @@
-# Effort: reasoning_effort = none|high|max (default max)
-# Off is effort=none. thinking.type=disabled is accepted as an alias for none.
-# low|medium map to high; xhigh maps to max.
-# Pricing: https://zro.moonmath.ai/pricing (USD / 1M tokens)
 base_model = "zhipuai/glm-5.2"
-reasoning_options = [{ type = "effort", values = ["none", "high", "max"] }]
+reasoning_options = [{ type = "effort", values = ["high", "max"] }]
 
 [interleaved]
 field = "reasoning_content"
diff --git a/providers/zro/models/glm-5.3-flash.toml b/providers/zro/models/glm-5.3-flash.toml
index f2835eddd2b..c38a92a6ba7 100644
--- a/providers/zro/models/glm-5.3-flash.toml
+++ b/providers/zro/models/glm-5.3-flash.toml
@@ -1,9 +1,5 @@
-# Effort: reasoning_effort = none|high|max (default max)
-# Off is effort=none. thinking.type=disabled is accepted as an alias for none.
-# Pricing: https://zro.moonmath.ai/pricing (USD / 1M tokens)
 base_model = "zhipuai/glm-5.3-flash"
-attachment = false
-reasoning_options = [{ type = "effort", values = ["none", "high", "max"] }]
+reasoning_options = [{ type = "effort", values = ["low", "high", "max"] }]
 
 [interleaved]
 field = "reasoning_content"
@@ -16,6 +12,3 @@ cache_read = 0.03
 [limit]
 context = 1_048_576
 output = 64_000
-
-[modalities]
-input = ["text"]
diff --git a/providers/zro/models/glm-5.3.toml b/providers/zro/models/glm-5.3.toml
index 7c59e143727..b05c08f1a9f 100644
--- a/providers/zro/models/glm-5.3.toml
+++ b/providers/zro/models/glm-5.3.toml
@@ -1,8 +1,5 @@
-# Effort: reasoning_effort = none|high|max (default max)
-# Off is effort=none. thinking.type=disabled is accepted as an alias for none.
-# Pricing: https://zro.moonmath.ai/pricing (USD / 1M tokens)
 base_model = "zhipuai/glm-5.3"
-reasoning_options = [{ type = "effort", values = ["none", "high", "max"] }]
+reasoning_options = [{ type = "effort", values = ["low", "high", "max"] }]
 
 [interleaved]
 field = "reasoning_content"
diff --git a/providers/zro/models/kimi-k3.toml b/providers/zro/models/kimi-k3.toml
index b47152f8c38..e1fb9b3f352 100644
--- a/providers/zro/models/kimi-k3.toml
+++ b/providers/zro/models/kimi-k3.toml
@@ -1,5 +1,3 @@
-# Effort: reasoning_effort = low|high|max (default high)
-# Pricing: https://zro.moonmath.ai/pricing (USD / 1M tokens)
 base_model = "moonshotai/kimi-k3"
 temperature = true
 reasoning_options = [{ type = "effort", values = ["low", "high", "max"] }]

From fe08f96f3e743684ca78594f305f26a45813b2ed Mon Sep 17 00:00:00 2001
From: Jeremy Felder <jeremy.felder1@gmail.com>
Date: Thu, 3 Sep 2026 11:09:32 +0300
Subject: [PATCH 7/8] fix(provider): document Zro wire path and pricing on
 model files

---
 providers/zro/models/deepseek-v4-flash-0731.toml | 4 ++++
 providers/zro/models/glm-5.2.toml                | 1 +
 providers/zro/models/glm-5.3-flash.toml          | 1 +
 providers/zro/models/glm-5.3.toml                | 1 +
 providers/zro/models/kimi-k3.toml                | 4 ++++
 5 files changed, 11 insertions(+)

diff --git a/providers/zro/models/deepseek-v4-flash-0731.toml b/providers/zro/models/deepseek-v4-flash-0731.toml
index 2866c706cdb..819ef662bef 100644
--- a/providers/zro/models/deepseek-v4-flash-0731.toml
+++ b/providers/zro/models/deepseek-v4-flash-0731.toml
@@ -1,3 +1,7 @@
+# Wire: only OpenAI-style `reasoning_effort` is forwarded.
+# No on/off toggle field is forwarded on this surface, so there is no toggle option to expose.
+# Effort: reasoning_effort = low|high|max
+# Pricing: https://zro.moonmath.ai/pricing (USD / 1M tokens)
 base_model = "deepseek/deepseek-v4-flash-0731"
 reasoning_options = [{ type = "effort", values = ["low", "high", "max"] }]
 
diff --git a/providers/zro/models/glm-5.2.toml b/providers/zro/models/glm-5.2.toml
index b837ce22598..4c7303a2538 100644
--- a/providers/zro/models/glm-5.2.toml
+++ b/providers/zro/models/glm-5.2.toml
@@ -1,3 +1,4 @@
+# Pricing: https://zro.moonmath.ai/pricing (USD / 1M tokens)
 base_model = "zhipuai/glm-5.2"
 reasoning_options = [{ type = "effort", values = ["high", "max"] }]
 
diff --git a/providers/zro/models/glm-5.3-flash.toml b/providers/zro/models/glm-5.3-flash.toml
index c38a92a6ba7..0c9c3c5b757 100644
--- a/providers/zro/models/glm-5.3-flash.toml
+++ b/providers/zro/models/glm-5.3-flash.toml
@@ -1,3 +1,4 @@
+# Pricing: https://zro.moonmath.ai/pricing (USD / 1M tokens)
 base_model = "zhipuai/glm-5.3-flash"
 reasoning_options = [{ type = "effort", values = ["low", "high", "max"] }]
 
diff --git a/providers/zro/models/glm-5.3.toml b/providers/zro/models/glm-5.3.toml
index b05c08f1a9f..39a0322f16d 100644
--- a/providers/zro/models/glm-5.3.toml
+++ b/providers/zro/models/glm-5.3.toml
@@ -1,3 +1,4 @@
+# Pricing: https://zro.moonmath.ai/pricing (USD / 1M tokens)
 base_model = "zhipuai/glm-5.3"
 reasoning_options = [{ type = "effort", values = ["low", "high", "max"] }]
 
diff --git a/providers/zro/models/kimi-k3.toml b/providers/zro/models/kimi-k3.toml
index e1fb9b3f352..d2e53afcc65 100644
--- a/providers/zro/models/kimi-k3.toml
+++ b/providers/zro/models/kimi-k3.toml
@@ -1,3 +1,7 @@
+# Wire: only OpenAI-style `reasoning_effort` is forwarded.
+# No on/off toggle field is forwarded on this surface, so there is no toggle option to expose.
+# Effort: reasoning_effort = low|high|max
+# Pricing: https://zro.moonmath.ai/pricing (USD / 1M tokens)
 base_model = "moonshotai/kimi-k3"
 temperature = true
 reasoning_options = [{ type = "effort", values = ["low", "high", "max"] }]

From a5c2f027778108901f7add4f4de9933ef3af4e3c Mon Sep 17 00:00:00 2001
From: Jeremy Felder <jeremy.felder1@gmail.com>
Date: Sun, 6 Sep 2026 08:56:55 +0300
Subject: [PATCH 8/8] fix(provider): expose DeepSeek off via
 reasoning_effort=none

---
 providers/zro/models/deepseek-v4-flash-0731.toml | 5 ++---
 1 file changed, 2 insertions(+), 3 deletions(-)

diff --git a/providers/zro/models/deepseek-v4-flash-0731.toml b/providers/zro/models/deepseek-v4-flash-0731.toml
index 819ef662bef..1e929fddb08 100644
--- a/providers/zro/models/deepseek-v4-flash-0731.toml
+++ b/providers/zro/models/deepseek-v4-flash-0731.toml
@@ -1,9 +1,8 @@
 # Wire: only OpenAI-style `reasoning_effort` is forwarded.
-# No on/off toggle field is forwarded on this surface, so there is no toggle option to expose.
-# Effort: reasoning_effort = low|high|max
+# Off is effort=none (no separate toggle field). Graded levels: low|high|max.
 # Pricing: https://zro.moonmath.ai/pricing (USD / 1M tokens)
 base_model = "deepseek/deepseek-v4-flash-0731"
-reasoning_options = [{ type = "effort", values = ["low", "high", "max"] }]
+reasoning_options = [{ type = "effort", values = ["none", "low", "high", "max"] }]
 
 [interleaved]
 field = "reasoning_content"