Unverified50% confidenceFactExact time
The experiments use three mid-tier judges: two local models (Qwen 9B/4B, run via Ollama) and one gpt-4o-mini, with total API spending of $0.29.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/19/2026
First Seen
Valid until: 1/17/2027
Sources
Related Entities
Related Claims
UnverifiedThe per-simulation user-side cost of the seven agents ranges from 0.0063 to 0.1841; Qwen3.6-35B, Gemma-31B, and GPT-5.4-mini are marked as strictly dominated on the cost-UFS plane.74% similarUnverifiedCost data from this experiment: all three applications combined at low cost less than 4 dollars; medium about 14 dollars; high about 27 dollars; extra about 50 dollars; max about 82 dollars; UltraCode over 300 dollars.72% similarUnverifiedArtificial Analysis previously measured the V4-Flash series and found its per-task cost far lower than competitors (3 cents in its testing framework, versus $1.86 for GPT-5.6 Sol).71% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1069902API
curl https://kongchang.com/api/v1/knowledge/claims/1069902MCP
get_claim(id=1069902)