Expired50% confidenceBenchmarkExact time
In the Atomic Chat comparative evaluation, HY3 consumed approximately 30,000 tokens to build three HTML5 Canvas physics demos at a cost of only fractions of a cent, while Gemini 3.5 Flash cost approximately $0.21 and GLM 5.2 cost $0.07.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
7/13/2026
First Seen
Valid until: 10/11/2026(expired)
Sources
Related Claims
UnverifiedWhen building three canvas physics demos, the model used approximately 62,000 tokens at a cost of about $3, while GLM 5.2 used 36,000 tokens at a cost of only 8 cents.79% similarUnverifiedIn the 3D art museum test, Opus 4.8 output 437,000 tokens and Fable 5 output 280,000 tokens, with Fable 5 costing approximately $64, which is about 37% more expensive than Opus 4.8's approximately $46.75% similarUnverifiedThe inference cost of GPT-4-level models is approximately $10–$30 per million tokens, while lightweight models such as Gemini Nano can compress costs to below one-tenth of that.66% similarUnverifiedFor Google AI Studio's Gemini model, text/image/video input costs approximately $0.25 and output approximately $1.5; audio input costs approximately $0.566% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/495161API
curl https://kongchang.com/api/v1/knowledge/claims/495161MCP
get_claim(id=495161)