MINA LABS
Start creating free
The Ollama project released version 0.34.3 on September 19, 2026, introducing a change to its /api/show endpoint that now includes detailed information about each model’s thinking controls. This update, detailed in the GitHub release notes, allows developers to query not just a model’s capabilities but also its default reasoning settings, such as the thinking parameter with options for low, high, and max depth.
For developers building applications that rely on local or self-hosted models, this change simplifies integration. Previously, configuring reasoning depth required manual intervention or hardcoded defaults. Now, the API explicitly advertises available options, enabling dynamic adjustments based on task complexity. For example, a conversational agent could automatically switch to high reasoning for nuanced queries while using low for straightforward responses, optimizing performance and user experience without over-provisioning resources.
The glm-5.3-flash:cloud model, available on Mina Labs, exemplifies how these controls can be leveraged. This model supports three reasoning levels, with max as the default. Developers can now programmatically toggle between depths to balance accuracy and speed. For instance, code generation tasks might benefit from max to ensure precision, while real-time chatbots could use low to reduce latency. The API’s structured response—{"thinking": {"values": ["low", "high", "max"], "default": "max"}}—provides a clear interface for implementing such logic.
Using this model on Mina Labs costs 4 per image generation. The platform’s infrastructure ensures consistent performance across reasoning depths, making it a reliable choice for applications requiring both speed and accuracy. Mina Labs’ integration also supports seamless deployment, allowing teams to test and iterate without managing underlying hardware.
Practical applications for these controls extend beyond chatbots. Developers could build debugging tools that escalate reasoning depth when detecting ambiguous code queries, or create educational apps that explain complex topics at higher reasoning levels while summarizing simpler concepts at lower depths. The API’s transparency reduces guesswork, enabling more responsive and context-aware systems.
Critically, this update aligns with broader trends in AI development, where fine-grained control over model behavior is becoming essential. By exposing these parameters at the API level, Ollama empowers developers to build adaptive workflows without deep diving into model-specific configurations. The change is backward-compatible, so existing integrations need not be rewritten—only enhanced.
As AI tools evolve, features like dynamic reasoning controls will become standard expectations. This release positions Ollama users to meet that standard while leveraging Mina Labs’ cost-effective infrastructure for scalable deployments. For teams prioritizing efficiency and flexibility, the update removes friction between experimentation and production-grade implementation.