AI Models
Are Smaller AI Models Good Enough?
Don't just read what happened. See what could happen next.
Prediction
Small and mid-size models handle a majority of enterprise tasks within 12 months; frontier models stay reserved for hard reasoning and agents.
- Confidence
- 69%
- Horizon
- 6–18 months
- Impact
- High
- Direction
- Increasing
Prediction changed +8 points in 21 days
Short answer
Distillation and open mid-size releases made “good enough” real for summarization, classification, and many copilots. Past-year cost pressure accelerates the shift; hard agentic work still leans large.
Why this question matters
If small models suffice, inference moves on-prem and on-device — reshaping chip mix, privacy, and API revenue.
What's happening now?
Strong open mid-size releases
Sub-frontier weights punch above parameter count on practical tasks.
Signal · strong
On-device and PC AI SKUs
OEMs ship NPUs marketed for local assistants.
Signal · moderate
Router architectures
Gateways send easy traffic to small models, hard traffic to frontier.
Signal · strong
Fine-tune economics
Domain small models beat general giants on narrow KPIs.
Signal · moderate
What could happen next?
Scenario A
Small-default world
Most tokens served by efficient models; frontier is premium.
Scenario B
Frontier still default
Agent complexity keeps buyers on the largest APIs.
Scenario C
Edge-first consumer AI
Phones and PCs absorb casual AI without cloud round-trips.
Key companies / entities
- Meta
- Mistral
- Apple
- Qualcomm
- Microsoft
Evidence
- research
Small-model benchmark comparisons
- company
On-device AI product launches
- news
Enterprise model-routing case studies
Prediction history
- Sep 22, 202669%
- Sep 15, 202666%
- Sep 8, 202664%
- Sep 1, 202661%