PulseAugur
EN
LIVE 05:18:47

Ollama bug bypasses JSON schema when models skip thinking

By PulseAugur Editorial · [1 sources] ·

Ollama versions 0.34.4 and later exhibit a bug where the JSON schema enforcement is bypassed when a model, such as Gemma 4, chooses not to engage in its "thinking" process. This issue causes responses that skip the thinking step to be returned as plain text, even when a schema is specified, leading to potential parsing errors on the client side. While enabling the `think: false` parameter or using a specific fix (ollama/ollama#18783) can mitigate the problem by forcing the model to think or by ensuring the response adheres to the schema, the core issue remains unaddressed in current releases. AI

IMPACT This bug could lead to unexpected data formatting issues for users relying on Ollama for structured output generation.

RANK_REASON This is a bug report about a specific software product's behavior, not a new release or significant industry development.

Read on dev.to — LLM tag →

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →

Ollama bug bypasses JSON schema when models skip thinking

How we ranked this

Signal score
16 / 100
Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket
Tool
This is a bug report about a specific software product's behavior, not a new release or significant industry development.
Source corroboration
Single-source cluster
Only one publisher covered this so far. Single-source stories can still rank when the publisher is high-authority, but they lack cross-source corroboration.
Topics
product, other
Editorial topic classification. Feeds into how the story surfaces on /topic/<slug> hub pages and into the per-entity coverage mix.
AI-industry relevance
High
Clearly on-topic for AI-industry coverage.
Story freshness
Breaking (< 6h)
Fresh story with cross-source coverage still developing. Ranking may shift as more sources report.

Full methodology in our editorial standards.

COVERAGE [1]

  1. dev.to — LLM tag TIER_1 English(EN) · The Homelab Postmortem ·

    Ollama skips your JSON schema when a thinking model answers without thinking

    <p><strong>TL;DR</strong>: Since 0.34.4, Ollama applies a <code>format</code> schema on a thinking model as a single grammar: anything until the thinking block closes, then the schema. That grammar also allows the response to end before the closing, and it treats everything up to…

RELATED ENTITIES

RELATED TOPICS