diff --git a/providers/gptzzz/logo.svg b/providers/gptzzz/logo.svg
new file mode 100644
index 00000000000..f8be18d2183
--- /dev/null
+++ b/providers/gptzzz/logo.svg
@@ -0,0 +1 @@
+<svg fill="currentColor" viewBox="0 0 24 24" xmlns="http://www.w3.org/2000/svg"><title>GPTZZZ</title><path transform="rotate(-22 12 12)" d="M.7 12a11.3 4.05 0 1 1 22.6 0 11.3 4.05 0 1 1-22.6 0ZM2.1 12a9.9 2.55 0 1 0 19.8 0 9.9 2.55 0 1 0-19.8 0ZM6.05 12a5.95 5.95 0 1 1 11.9 0 5.95 5.95 0 1 1-11.9 0Z"/></svg>
diff --git a/providers/gptzzz/models/gpt-5.6-sol.toml b/providers/gptzzz/models/gpt-5.6-sol.toml
new file mode 100644
index 00000000000..a803e00af59
--- /dev/null
+++ b/providers/gptzzz/models/gpt-5.6-sol.toml
@@ -0,0 +1,24 @@
+# GPTZZZ is an OpenAI-compatible relay. Effort values verified per model against the
+# live /v1/chat/completions endpoint on 2026-09-16 and re-probed on 2026-09-29: none, low,
+# medium, high, xhigh and max all return HTTP 200; minimal returns HTTP 400.
+# Pricing source: https://gptzzz.ai/model-plaza (read from /api/v1/model-plaza on 2026-09-29).
+# GPTZZZ bills this model against its own published list price of 5.00 input / 30.00 output /
+# 0.50 cache read / 6.25 cache write per MTok (>272K context: 10.00 / 45.00 / 1.00 / 12.50),
+# which is higher than OpenAI's list for gpt-5.6-sol, multiplied by its public 0.05x group rate.
+# The host's /v1/usage "cost" field matches this 5.00 / 30.00 list exactly, so the values
+# below are what the host actually charges.
+base_model = "openai/gpt-5.6-sol"
+reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh", "max"] }]
+
+[cost]
+input = 0.25
+output = 1.5
+cache_read = 0.025
+cache_write = 0.3125
+
+[[cost.tiers]]
+tier = { size = 272_000 }
+input = 0.5
+output = 2.25
+cache_read = 0.05
+cache_write = 0.625
diff --git a/providers/gptzzz/models/gpt-5.6-terra.toml b/providers/gptzzz/models/gpt-5.6-terra.toml
new file mode 100644
index 00000000000..65e402b588a
--- /dev/null
+++ b/providers/gptzzz/models/gpt-5.6-terra.toml
@@ -0,0 +1,21 @@
+# GPTZZZ is an OpenAI-compatible relay. Effort values verified per model against the
+# live /v1/chat/completions endpoint on 2026-09-16 and re-probed on 2026-09-29: none, low,
+# medium, high, xhigh and max all return HTTP 200; minimal returns HTTP 400.
+# Pricing source: https://gptzzz.ai/model-plaza (read from /api/v1/model-plaza on 2026-09-29).
+# Published list price 2.00 / 12.00 / 0.20 / 2.50 per MTok (>272K: 4.00 / 18.00 / 0.40 / 5.00),
+# multiplied by the public 0.05x group rate. The host's /v1/usage "cost" field matches this list.
+base_model = "openai/gpt-5.6-terra"
+reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh", "max"] }]
+
+[cost]
+input = 0.1
+output = 0.6
+cache_read = 0.01
+cache_write = 0.125
+
+[[cost.tiers]]
+tier = { size = 272_000 }
+input = 0.2
+output = 0.9
+cache_read = 0.02
+cache_write = 0.25
diff --git a/providers/gptzzz/models/gpt-5.6.toml b/providers/gptzzz/models/gpt-5.6.toml
new file mode 100644
index 00000000000..02ec739c8e5
--- /dev/null
+++ b/providers/gptzzz/models/gpt-5.6.toml
@@ -0,0 +1,25 @@
+# GPTZZZ is an OpenAI-compatible relay. Effort values verified per model against the
+# live /v1/chat/completions endpoint on 2026-09-16 and re-probed on 2026-09-29: none, low,
+# medium, high, xhigh and max all return HTTP 200; minimal returns HTTP 400.
+# Pricing source: https://gptzzz.ai/model-plaza (read from /api/v1/model-plaza on 2026-09-29).
+# GPTZZZ bills this model against its own published list price of 5.00 input / 30.00 output /
+# 0.50 cache read / 6.25 cache write per MTok (>272K context: 10.00 / 45.00 / 1.00 / 12.50),
+# which is higher than OpenAI's list for gpt-5.6-sol, multiplied by its public 0.05x group rate.
+# The host's /v1/usage "cost" field matches this 5.00 / 30.00 list exactly, so the values
+# below are what the host actually charges.
+base_model = "openai/gpt-5.6-sol"
+name = "GPT-5.6"
+reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh", "max"] }]
+
+[cost]
+input = 0.25
+output = 1.5
+cache_read = 0.025
+cache_write = 0.3125
+
+[[cost.tiers]]
+tier = { size = 272_000 }
+input = 0.5
+output = 2.25
+cache_read = 0.05
+cache_write = 0.625
diff --git a/providers/gptzzz/models/gpt-6-astra.toml b/providers/gptzzz/models/gpt-6-astra.toml
new file mode 100644
index 00000000000..cbc0c8dee38
--- /dev/null
+++ b/providers/gptzzz/models/gpt-6-astra.toml
@@ -0,0 +1,24 @@
+# GPTZZZ is an OpenAI-compatible relay. Effort values probed one by one against the live
+# /v1/chat/completions endpoint on 2026-09-29: low, medium, high, xhigh and max return
+# HTTP 200 (reasoning_tokens 10 / 11 / 14 at high / xhigh / max); none and minimal return
+# HTTP 400 ("Supported values are: 'low', 'medium', 'high', 'xhigh', and 'max'").
+# This matches the OpenAI lab entry.
+# Pricing source: https://gptzzz.ai/model-plaza (read from /api/v1/model-plaza on 2026-09-29).
+# Published list price 10.00 / 50.00 / 1.00 / 12.50 per MTok (>272K: 20.00 / 75.00 / 2.00 / 25.00),
+# multiplied by the public 0.05x group rate. The host's /v1/usage "cost" field matches this
+# list price exactly.
+base_model = "openai/gpt-6-astra"
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
+
+[cost]
+input = 0.5
+output = 2.5
+cache_read = 0.05
+cache_write = 0.625
+
+[[cost.tiers]]
+tier = { size = 272_000 }
+input = 1.0
+output = 3.75
+cache_read = 0.1
+cache_write = 1.25
diff --git a/providers/gptzzz/provider.toml b/providers/gptzzz/provider.toml
new file mode 100644
index 00000000000..c8f9cd3abee
--- /dev/null
+++ b/providers/gptzzz/provider.toml
@@ -0,0 +1,5 @@
+name = "GPTZZZ"
+env = ["GPTZZZ_API_KEY"]
+npm = "@ai-sdk/openai-compatible"
+api = "https://gptzzz.ai/v1"
+doc = "https://gptzzz.ai/docs/"