diff --git a/.github/workflows/python-ci.yml b/.github/workflows/python-ci.yml
index 537fe540e7..df97444a17 100644
--- a/.github/workflows/python-ci.yml
+++ b/.github/workflows/python-ci.yml
@@ -58,6 +58,7 @@ jobs:
     - uses: actions/checkout@df4cb1c069e1874edd31b4311f1884172cec0e10 # v6.0.3
       with:
         persist-credentials: false
+        submodules: recursive
     - uses: actions/setup-python@a309ff8b426b58ec0e2a45f0f869d46889d02405 # v6.2.0
       with:
         python-version: ${{ matrix.python }}
diff --git a/.gitmodules b/.gitmodules
new file mode 100644
index 0000000000..8b7b6e8b71
--- /dev/null
+++ b/.gitmodules
@@ -0,0 +1,3 @@
+[submodule "iceberg-testing"]
+	path = iceberg-testing
+	url = https://github.com/sungwy/iceberg-testing.git
diff --git a/dev/.rat-excludes b/dev/.rat-excludes
index 485579eff4..633f20e102 100644
--- a/dev/.rat-excludes
+++ b/dev/.rat-excludes
@@ -1,5 +1,7 @@
 .github/**
 dev/.rat-excludes
+.gitmodules
+iceberg-testing/**
 uv.lock
 .ruff_cache/**
 .pytest_cache/**
diff --git a/iceberg-testing b/iceberg-testing
new file mode 160000
index 0000000000..1bd33d3b34
--- /dev/null
+++ b/iceberg-testing
@@ -0,0 +1 @@
+Subproject commit 1bd33d3b3410d214ab942b704b35ed0cac44d951
diff --git a/tests/conformance/_shared.py b/tests/conformance/_shared.py
new file mode 100644
index 0000000000..0c28ae545d
--- /dev/null
+++ b/tests/conformance/_shared.py
@@ -0,0 +1,45 @@
+# Licensed to the Apache Software Foundation (ASF) under one
+# or more contributor license agreements.  See the NOTICE file
+# distributed with this work for additional information
+# regarding copyright ownership.  The ASF licenses this file
+# to you under the Apache License, Version 2.0 (the
+# "License"); you may not use this file except in compliance
+# with the License.  You may obtain a copy of the License at
+#
+#   http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing,
+# software distributed under the License is distributed on an
+# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+# KIND, either express or implied.  See the License for the
+# specific language governing permissions and limitations
+# under the License.
+"""Shared helpers for the iceberg-testing conformance suite.
+
+The cross-language fixtures are language-neutral: a static input plus the expected
+result the spec fixes for it, pinned in the ``iceberg-testing/`` submodule. The
+per-surface test modules walk them and run them against PyIceberg. Cases PyIceberg
+does not satisfy yet are marked xfail in each module with a tracking issue; that
+staged-adoption list is a consumer-side concern. The submodule must be checked out
+(``git submodule update --init``); otherwise these tests skip.
+"""
+
+import json
+import os
+from typing import Any
+
+import pytest
+
+ROOT = os.path.abspath(os.path.join(os.path.dirname(__file__), "..", ".."))
+FIXTURES = os.path.join(ROOT, "iceberg-testing", "table-spec")
+FIXTURES_PRESENT = os.path.isdir(FIXTURES)
+
+skip_if_absent = pytest.mark.skipif(
+    not FIXTURES_PRESENT,
+    reason="iceberg-testing submodule not initialized (git submodule update --init --recursive)",
+)
+
+
+def load_jsonl(path: str) -> list[dict[str, Any]]:
+    with open(path) as handle:
+        return [json.loads(line) for line in handle if line.strip()]
diff --git a/tests/conformance/test_bucket_hash.py b/tests/conformance/test_bucket_hash.py
new file mode 100644
index 0000000000..b26824de11
--- /dev/null
+++ b/tests/conformance/test_bucket_hash.py
@@ -0,0 +1,80 @@
+# Licensed to the Apache Software Foundation (ASF) under one
+# or more contributor license agreements.  See the NOTICE file
+# distributed with this work for additional information
+# regarding copyright ownership.  The ASF licenses this file
+# to you under the Apache License, Version 2.0 (the
+# "License"); you may not use this file except in compliance
+# with the License.  You may obtain a copy of the License at
+#
+#   http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing,
+# software distributed under the License is distributed on an
+# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+# KIND, either express or implied.  See the License for the
+# specific language governing permissions and limitations
+# under the License.
+"""Bucket-transform conformance: the Appendix B 32-bit hash that bucket[N] builds on."""
+
+import os
+import uuid as uuidmod
+from decimal import Decimal
+from typing import Any
+
+import pytest
+from _shared import FIXTURES, FIXTURES_PRESENT, load_jsonl, skip_if_absent
+
+from pyiceberg.transforms import BucketTransform
+from pyiceberg.types import PrimitiveType
+from pyiceberg.utils.datetime import date_str_to_days, time_str_to_micros, timestamp_to_micros
+
+pytestmark = skip_if_absent
+
+# PyIceberg under-sizes the minimal two's-complement of byte-boundary decimals, so it
+# hashes (and buckets) these to a different value than the spec fixes. Tracked in the
+# open issue below; remove these entries when it lands and the cases XPASS.
+BUCKET_XFAIL = {
+    "hash-decimal-neg-1.28": "minimal-byte decimal serialization differs from spec; apache/iceberg-python#3522",
+    "hash-decimal-neg-327.68": "minimal-byte decimal serialization differs from spec; apache/iceberg-python#3522",
+    "hash-decimal-neg-83886.08": "minimal-byte decimal serialization differs from spec; apache/iceberg-python#3522",
+}
+_BUCKET_N = 2_000_003  # large prime; comparing bucket[N] avoids needing the raw hash API
+
+
+def _physical_value(type_str: str, value: str) -> Any:
+    """Convert a fixture's string value to the physical form PyIceberg buckets."""
+    if type_str in ("int", "long"):
+        return int(value)
+    if type_str.startswith("decimal"):
+        return Decimal(value)
+    if type_str == "date":
+        return date_str_to_days(value)
+    if type_str == "time":
+        return time_str_to_micros(value)
+    if type_str == "timestamp":
+        return timestamp_to_micros(value)
+    if type_str == "string":
+        return value
+    if type_str == "uuid":
+        return uuidmod.UUID(value)
+    if type_str.startswith("fixed") or type_str == "binary":
+        return bytes.fromhex(value)
+    raise ValueError(type_str)
+
+
+def _bucket_cases() -> list[Any]:
+    if not FIXTURES_PRESENT:
+        return []
+    out: list[Any] = []
+    for case in load_jsonl(os.path.join(FIXTURES, "transforms", "bucket", "cases.jsonl")):
+        marks = [pytest.mark.xfail(reason=BUCKET_XFAIL[case["id"]], strict=False)] if case["id"] in BUCKET_XFAIL else []
+        out.append(pytest.param(case, id=case["id"], marks=marks))
+    return out
+
+
+@pytest.mark.parametrize("case", _bucket_cases())
+def test_bucket_hash(case: dict[str, Any]) -> None:
+    source_type = PrimitiveType.model_validate(case["type"])
+    bucket = BucketTransform(num_buckets=_BUCKET_N).transform(source_type)
+    got = bucket(_physical_value(case["type"], case["value"]))
+    assert got == (case["hash"] & 0x7FFFFFFF) % _BUCKET_N
diff --git a/tests/conformance/test_delete_formats.py b/tests/conformance/test_delete_formats.py
new file mode 100644
index 0000000000..1fb0500418
--- /dev/null
+++ b/tests/conformance/test_delete_formats.py
@@ -0,0 +1,57 @@
+# Licensed to the Apache Software Foundation (ASF) under one
+# or more contributor license agreements.  See the NOTICE file
+# distributed with this work for additional information
+# regarding copyright ownership.  The ASF licenses this file
+# to you under the Apache License, Version 2.0 (the
+# "License"); you may not use this file except in compliance
+# with the License.  You may obtain a copy of the License at
+#
+#   http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing,
+# software distributed under the License is distributed on an
+# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+# KIND, either express or implied.  See the License for the
+# specific language governing permissions and limitations
+# under the License.
+"""Delete-file conformance: decode positional and equality delete files by field-id."""
+
+import glob
+import json
+import os
+from typing import Any
+
+import pyarrow.parquet as pq
+import pytest
+from _shared import FIXTURES, skip_if_absent
+
+pytestmark = skip_if_absent
+
+
+def _encode(value: Any) -> Any:
+    if value is None:
+        return None
+    if isinstance(value, int):
+        return str(value)
+    if isinstance(value, bytes):
+        return value.hex().upper()
+    return value
+
+
+def _decode(path: str) -> list[dict[str, Any]]:
+    table = pq.read_table(path)
+    ids = {field.name: int((field.metadata or {})[b"PARQUET:field_id"]) for field in table.schema}
+    cols = {name: table.column(name).to_pylist() for name in table.column_names}
+    return [{str(ids[name]): _encode(cols[name][row]) for name in table.column_names} for row in range(table.num_rows)]
+
+
+@pytest.mark.parametrize(
+    "sidecar",
+    sorted(glob.glob(os.path.join(FIXTURES, "delete-formats", "*", "*.parquet.expected.json"))),
+    ids=lambda p: os.path.basename(os.path.dirname(p)),
+)
+def test_delete_file_decode(sidecar: str) -> None:
+    with open(sidecar) as handle:
+        expected = json.load(handle)
+    data_file = os.path.join(os.path.dirname(sidecar), expected["delete-file"])
+    assert _decode(data_file) == expected["decoded-rows"]
diff --git a/tests/conformance/test_type_strings.py b/tests/conformance/test_type_strings.py
new file mode 100644
index 0000000000..874a73f89d
--- /dev/null
+++ b/tests/conformance/test_type_strings.py
@@ -0,0 +1,76 @@
+# Licensed to the Apache Software Foundation (ASF) under one
+# or more contributor license agreements.  See the NOTICE file
+# distributed with this work for additional information
+# regarding copyright ownership.  The ASF licenses this file
+# to you under the Apache License, Version 2.0 (the
+# "License"); you may not use this file except in compliance
+# with the License.  You may obtain a copy of the License at
+#
+#   http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing,
+# software distributed under the License is distributed on an
+# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+# KIND, either express or implied.  See the License for the
+# specific language governing permissions and limitations
+# under the License.
+"""Type-string conformance: parse the input and re-serialize to the canonical form."""
+
+import glob
+import os
+from typing import Any
+
+import pytest
+from _shared import FIXTURES, FIXTURES_PRESENT, load_jsonl, skip_if_absent
+
+from pyiceberg.exceptions import ValidationError
+from pyiceberg.types import DecimalType, FixedType, PrimitiveType
+
+pytestmark = skip_if_absent
+
+# Cases PyIceberg does not satisfy yet (consumer-side staged-adoption list). Each
+# reason links the open issue that tracks the divergence; when it lands the case
+# passes (reported XPASS, since these xfails are non-strict) and the entry here can
+# be removed. fixed[...] whitespace has no dedicated issue; it is the same
+# type-string-whitespace class as the decimal case.
+TYPE_XFAIL = {
+    "decimal-space-around-params": "rejects optional whitespace around decimal params; apache/iceberg#16798",
+    "fixed-space-around": "rejects whitespace inside fixed[...] brackets; type-string whitespace, cf. apache/iceberg#16798",
+    "geometry-default": "geometry type not supported yet; apache/iceberg-python#3004",
+    "geometry-unquoted-crs": "geometry type not supported yet; apache/iceberg-python#3004",
+    "decimal-precision-over-max": "does not reject decimal precision > 38 (spec: precision must be 38 or less; Java rejects it)",
+}
+
+
+def _type_fields(parsed: PrimitiveType) -> dict[str, int]:
+    if isinstance(parsed, DecimalType):
+        return {"precision": parsed.precision, "scale": parsed.scale}
+    if isinstance(parsed, FixedType):
+        return {"length": parsed.root}
+    return {}
+
+
+def _type_cases(*, accept: bool) -> list[Any]:
+    if not FIXTURES_PRESENT:
+        return []
+    out: list[Any] = []
+    for path in sorted(glob.glob(os.path.join(FIXTURES, "types", "*", "cases.jsonl"))):
+        for case in load_jsonl(path):
+            if case["accept"] is not accept:
+                continue
+            marks = [pytest.mark.xfail(reason=TYPE_XFAIL[case["id"]], strict=False)] if case["id"] in TYPE_XFAIL else []
+            out.append(pytest.param(case, id=case["id"], marks=marks))
+    return out
+
+
+@pytest.mark.parametrize("case", _type_cases(accept=True))
+def test_type_string_accepted(case: dict[str, Any]) -> None:
+    parsed = PrimitiveType.model_validate(case["input"])
+    assert _type_fields(parsed) == case["parsed"]
+    assert str(parsed) == case["canonical"]
+
+
+@pytest.mark.parametrize("case", _type_cases(accept=False))
+def test_type_string_rejected(case: dict[str, Any]) -> None:
+    with pytest.raises(ValidationError):
+        PrimitiveType.model_validate(case["input"])