thinking_mismatch_allowed
Anthropic (Claude) input_transformations entry: type: "thinking_mismatch_allowed", reason: "prefix_binding_mismatch".
What it means
The thinking block failed the prefix check, but with the setting unset this model let it through: Claude read it unchanged and you were billed for it. It’s reasoning about content that has since changed.
thinking_mismatch_allowed,prefix_binding_mismatch: The same, but Claude read the block unchanged, and it’s billed. This entry appears only where the setting is unset, on some models. Yours to fix.
Fix a thinking block bound to a different conversation (checked 2026-11-20)
When no error is returned, the failing blocks reach the model. Claude reads them like any other input, you’re billed for them, and Claude works from reasoning about content that has since changed.
Prefix checks on thinking blocks (checked 2026-11-20)
Is it yours to fix?
Yes. Your code changed something, so it’s yours to fix.
You were billed for this block, and Claude read reasoning about content that has since changed. Send prefix_mismatch_behavior: "error" in tests so every failure shows up as a 400.
How to fix it
Send every earlier message exactly as you sent it before, and put each update in a new message at the end. Store any context your code adds to a message with that message.
An example that produces it (Date in system prompt):
{
"max_tokens": 4096,
"messages": [
{
"content": "Fix the failing test.",
"role": "user"
}
],
"model": "claude-fable-5-2",
"system": "You are a coding agent. Today is 2026-11-18."
}{
"max_tokens": 4096,
"messages": [
{
"content": "Fix the failing test.",
"role": "user"
},
{
"content": [
{
"signature": "sigA1",
"thinking": "",
"type": "thinking"
},
{
"text": "Done.",
"type": "text"
}
],
"role": "assistant"
},
{
"content": "Now run the linter.",
"role": "user"
}
],
"model": "claude-fable-5-2",
"system": "You are a coding agent. Today is 2026-11-19."
}Related reason codes
prefix_binding_mismatch: prefix_binding_mismatch: thinking dropped because the conversation before it changedmodel_binding_mismatch: model_binding_mismatch: this Claude model can’t read thinking from the model that produced itorganization_binding_mismatch: organization_binding_mismatch: thinking from another account was dropped
Last checked 2026-11-20 against the official Anthropic docs.
FAQ
Why didn't I get a 400 error?
On Claude Fable 5.1, Opus 5.5, Sonnet 5.5 and Haiku 5.5, a request that fails the prefix check with the setting unset might get a 400 or no error. With no error, the failing blocks reach Claude and are billed. Send prefix_mismatch_behavior: "error" in tests so every failure is a 400.
Does the beta header change what Claude reads?
No. Sending thinking-binding-controls-2026-08-01 with the setting unset only adds the input_transformations array so you can log these entries.