Pydantic Recursive model with discriminated union has

The Fix

pip install pydantic==2.6.3

Based on closed pydantic/pydantic issue #8709 · PR/commit linked

Jump to Verify Open PR/Commit

@@ -54,6 +54,12 @@ def inner(s: core_schema.CoreSchema, recurse: _core_utils.Recurse) -> core_schem
             if definitions is None:
                 definitions = collect_definitions(schema)
+                # After we collect the definitions schemas, we must run through the discriminator
+                # application logic for each one. This step is crucial to prevent an exponential
+                # increase in complexity that occurs if schemas are left as 'union' schemas

repro.py

from __future__ import annotations
from typing import Literal, Annotated
from pydantic import Field, TypeAdapter, BaseModel

class NestedState(BaseModel):
    state_type: Literal["nested"]
    substate: AnyState

# If this type is left out, the model behaves normally again
class LoopState(BaseModel):
    state_type: Literal["loop"]
    substate: AnyState

class LeafState(BaseModel):
    state_type: Literal["leaf"]

AnyState = Annotated[NestedState | LoopState | LeafState, Field(..., discriminator="state_type")]

def build_nested_state(n):
    if n <= 0:
        return {"state_type": "leaf"}
    else:
        return {"state_type": "loop", "substate": {"state_type": "nested", "substate": build_nested_state(n-1)}}
        
adapter = TypeAdapter(AnyState)

# the next statement takes around 0.8s
adapter.validate_python(build_nested_state(9))

# the next statement takes around 3.5s
adapter.validate_python(build_nested_state(10))

# the next statement takes around 12.5s
adapter.validate_python(build_nested_state(11))

verify

Re-run the minimal reproduction on your broken version, then apply the fix and re-run.

fix.md

Option A — Upgrade to fixed release\npip install pydantic==2.6.3\nWhen NOT to use: This fix should not be applied if the model does not use discriminated unions.\n\nOption C — Workaround\nfor this performance issue.\nWhen NOT to use: This fix should not be applied if the model does not use discriminated unions.\n\n

Why This Fix Works in Production

Trigger: Recursive model with discriminated union has exponential time and space complexity for validation
Mechanism: The recursive model validation had exponential time and space complexity due to improper handling of discriminated unions
Why the fix works: Fixes exponential time and space complexity in recursive model validation by improving discriminated union schema generation. (first fixed release: 2.6.3).

Production impact:

If left unfixed, this can cause silent data inconsistencies that propagate (bad cache entries, incorrect downstream decisions).

Why This Breaks in Prod

Shows up under Python 3.11 in real deployments (not just unit tests).
The recursive model validation had exponential time and space complexity due to improper handling of discriminated unions
Production symptom (often without a traceback): Recursive model with discriminated union has exponential time and space complexity for validation

Proof / Evidence

GitHub issue: #8709
Fix PR: https://github.com/pydantic/pydantic/pull/8904
First fixed release: 2.6.3
Reproduced locally: No (not executed)
Last verified: 2026-02-09
Confidence: 0.75
Did this fix it?: Yes (upstream fix exists)
Own content ratio: 0.44

Discussion

High-signal excerpts from the issue thread (symptoms, repros, edge-cases).

Jump to Sources Open on GitHub

“Hmm, this looks similar to this issue: https://github.com/pydantic/pydantic/issues/8688, which we fixed here https://github.com/pydantic/pydantic/pull/8702 for the 2.6.1 release”

@sydney-runkle · 2024-02-08 · confirmation · source

“@sydney-runkle unfortunately still reproducible with 2.6.1, as David already noticed.”

@palle-k · 2024-02-08 · source

“@palle-k, Thanks for reporting this. @davidhewitt, could you please take a look at this performance issue when you have a chance? Thank you!”

@sydney-runkle · 2024-02-06 · source

“The schema simplifies to: The issue is that the tagged union is not carried along so it's loss after the first union and thus performance…”

@adriangb · 2024-02-08 · source

Failure Signature (Search String)

Recursive model with discriminated union has exponential time and space complexity for validation

Copy-friendly signature

signature.txt

Failure Signature
-----------------
Recursive model with discriminated union has exponential time and space complexity for validation
python version: 3.11.7 (main, Dec  4 2023, 18:10:11) [Clang 15.0.0 (clang-1500.1.0.2.5)]

Error Message

Signature-only (no traceback captured)

error.txt

Error Message
-------------
Recursive model with discriminated union has exponential time and space complexity for validation
python version: 3.11.7 (main, Dec  4 2023, 18:10:11) [Clang 15.0.0 (clang-1500.1.0.2.5)]

Minimal Reproduction

repro.py

from __future__ import annotations
from typing import Literal, Annotated
from pydantic import Field, TypeAdapter, BaseModel

class NestedState(BaseModel):
    state_type: Literal["nested"]
    substate: AnyState

# If this type is left out, the model behaves normally again
class LoopState(BaseModel):
    state_type: Literal["loop"]
    substate: AnyState

class LeafState(BaseModel):
    state_type: Literal["leaf"]

AnyState = Annotated[NestedState | LoopState | LeafState, Field(..., discriminator="state_type")]

def build_nested_state(n):
    if n <= 0:
        return {"state_type": "leaf"}
    else:
        return {"state_type": "loop", "substate": {"state_type": "nested", "substate": build_nested_state(n-1)}}
        
adapter = TypeAdapter(AnyState)

# the next statement takes around 0.8s
adapter.validate_python(build_nested_state(9))

# the next statement takes around 3.5s
adapter.validate_python(build_nested_state(10))

# the next statement takes around 12.5s
adapter.validate_python(build_nested_state(11))

Environment

Python: 3.11
Pydantic: 2

What Broke

Validation time and memory usage could grow indefinitely, leading to potential denial-of-service attacks.

Why It Broke

The recursive model validation had exponential time and space complexity due to improper handling of discriminated unions

Fix Options (Details)

Option A — Upgrade to fixed release Safe default (recommended)

pip install pydantic==2.6.3

When NOT to use: This fix should not be applied if the model does not use discriminated unions.

Use when you can deploy the upstream fix. It is usually lower-risk than long-lived workarounds.

Option C — Workaround Temporary workaround

for this performance issue.

When NOT to use: This fix should not be applied if the model does not use discriminated unions.

Use only if you cannot change versions today. Treat this as a stopgap and remove once upgraded.

Fix reference: https://github.com/pydantic/pydantic/pull/8904

First fixed release: 2.6.3

Last verified: 2026-02-09. Validate in your environment.

When NOT to Use This Fix

This fix should not be applied if the model does not use discriminated unions.

Verify Fix

verify

Re-run the minimal reproduction on your broken version, then apply the fix and re-run.

Did This Fix Work in Your Case?

Quick signal helps us prioritize which fixes to verify and improve.

Prevention

Add a CI check that diffs key outputs after upgrades (OpenAPI schema snapshots, JSON payload shapes, CLI output).
Upgrade behind a canary and run integration tests against the canary before 100% rollout.

Version Compatibility Table

Version	Status
2.6.3	Fixed

Related Issues

No related fixes found.

Cluster: pydantic:data-consistency Pydantic hub Pydantic best practices All hubs All clusters

Related clusters: Configuration error OpenAPI schema Serialization

Sources

We don’t republish the full GitHub discussion text. Use the links above for context.

Pydantic Recursive model with discriminated union has exponential time and space complexity for (Fix)