thenaivekid/small-models-cannot-budget-their-thinking

Blind spot: small reasoning models cannot budget their own thinking Models evaluated (all 0.6B–4B, open weights, Hugging Face): Model Mode Role in this study Qwen/Qwen3-4B-Thinking-2507 (FP8) thinking primary model: the closest 0.6–6B analogue of a frontier reasoning model (same recipe family as Qwen3-235B; <think> channel, RL-tuned long CoT) Qwen/Qwen3-0.6B non-thinking smallest model of the frontier family; full 16-budget sweep Qwen/Qwen2.5-Math-1.5B-Instruct… See the full description on the dataset page: https://huggingface.co/datasets/thenaivekid/small-models-cannot-budget-their-thinking.

<iframe src="https://silytics.si/embed/item/ed5aa60f-1f51-4398-ba66-c1aa6efdd745" width="420" height="190" style="border:0" loading="lazy" title="Silytics entry"></iframe>

Silytics shows public metadata from Hugging Face and links to the original. Information may be incomplete or out of date. Last refreshed 22 October 2026. Licenses and terms of the original apply to the work itself. Not financial, legal or professional advice.