Reasoning models
Reasoning models can allocate computation to internal reasoning before and between answers or tool calls. More effort can help difficult tasks, with possible latency and token costs. The benefit is task-dependent, so compare quality, time and cost on your own examples.
What it is
OpenAI says its reasoning models use internal reasoning tokens before producing a response. This helps them plan, use tools, consider alternatives and solve harder multi-step tasks. Anthropic’s current models support adaptive thinking, but the allowed controls vary: Fable 5.1 and Opus 5.5 keep thinking on, while other models permit different settings.
How it works
- You send a task as usual.
- The model works through the problem in reasoning tokens before it answers.
- You steer the effort:
- OpenAI offers a reasoning-effort setting with model-dependent levels from none to max.
- Anthropic offers an effort parameter as soft guidance.
- Billing (checked 2026-10-18). Anthropic bills thinking tokens as output tokens, whether or not the thinking is shown. OpenAI counts reasoning tokens toward the output token limit you set with max_output_tokens.
Benefits
- Potentially better results on complex, multi-step tasks; validate the gain for your workload.
- Fewer prompt tricks needed to get careful work.
- Good for planning steps an agent will carry out.
Limits and risks
- More effort can increase latency and billed tokens. Lower it where supported and measure the trade-off.
- Routine rewriting or tagging may work well at lower effort (see model tiers). Reasoning-capable models can still be economical in suitable configurations.
- More thinking isn’t verification. Reasoning models still make mistakes.
Agent compatibility
Not tool-specific. See Agent-ready for connector support by tool.
Levels come from each tool's recorded, sourced compatibility. How the levels work.
Related entries
Sources
- Reasoning models · OpenAIofficial accessed Oct 18, 2026
- Models overview · Anthropicofficial accessed Oct 18, 2026
- Steering thinking (adaptive thinking) · Anthropicofficial accessed Oct 18, 2026
Practitioner views are restated in our words and labeled as views, not measured results. Product facts come from official sources dated above.
Suggest an edit
IonBin's wiki isn't user-edited. Corrections go to our Directory Curator, who checks them against primary sources before anything changes. Spot something wrong or out of date on “Reasoning models”? Email us with a link to a primary source.
Suggest an edit by email or write to contact@ionbin.com