Tell an agent it has 1M token budget and it reasons 3-5x longer: a metacognition experiment
paraschopra · x · 2026-09-10
Paras Chopra tested agents that are told their token budget for the run inside the prompt. Two agents ran the exact same model, settings and prompt — one told it had a 10k token budget, the other 1 million.
For the same task, the agent told it had the higher budget produced 3-5x longer reasoning. A clear sign of meta-cognitive abilities beyond mere task execution. He notes this isn't a separate thinking-effort setting — only the stated budget differed, yet token expenditure changed dramatically.
Related event: Telling an Agent It Has 1M Tokens Multiplies Its Reasoning 3-5x(2 posts)→
More from Models
- DeepSeek V4.1 Flash is actually 748B params, safetensors analysis shows — DistanceSolar1449 · 2026-09-10
- Kimi K3 lands on RunPod: 2.8T params, 1M context, $3/$15 per 1M tokens — Kimi_Moonshot · 2026-09-10
- Dev burns 300M tokens on GLM 5.3 in a week and still has quota left — saibharadwaj · 2026-09-10
- Follow-up: a 3T-parameter model may already exist, scaling issues remain the wildcard — teortaxesTex · 2026-09-10
- Speculation: DeepSeek V4.1 Pro could be a 3.1T-param MoE with 2.6TB disk footprint — teortaxesTex · 2026-09-10
- New Book Teaches Beginners to Build and Fine-Tune Their Own GPT-Style SLMs, With Colab Notebooks — Roger_M_Taylor · 2026-09-10