Unverified50% confidenceFactExact time
In the backend of the evaluation organization Cognition, a hidden model codenamed Obsidian-3 was found: under high compute on specific tasks it scored 46.8%, with an average cost of 8.62 dollars per task; Opus 5.5 in its full-power state scored 45.5%, with a task cost of 12.82 dollars.
1
Sources
50%
Confidence
Short-term (~14 days)
Relevance
10/29/2026
First Seen
Valid until: 11/12/2026
Sources
Gemini 4 Carbon Tops the Intelligence Leaderboard, But Coding Agents Still Trail Claude Opus 5.5
bilibili进化中的阿陈10/29/2026
Related Entities
Related Claims
UnverifiedObsidian 3's single inference cost is about $47, 3.6 times that of Opus 5.5 at the same intensity.74% similarUnverifiedIn Terminal Bench testing, GPT-6.1 Soul's max/high performance is nearly on par with Opus 5.5, but the cost per task is approximately $0.96 cents, while Opus costs $0.11 to $0.13 per task, roughly one-thirteenth of the price.72% similarUnverifiedClaude Opus 5.5 ranked first in the Code2Video Benchmark with a total score of 1536, at an average task cost of $3.08.72% similarUnverifiedRunning the same benchmark suite, Grok 4.6 costs $1.86, Astra costs $3.26, and Grok 4.7 costs as much as $3.74 per task.71% similarUnverifiedAmong 3,571 request logs, 2,321 intermediate Agent queries routed to the lightweight model cost a total of $1.73, and 1,250 complex reasoning queries routed to the flagship model cost a total of $70.55, for an actual total spend of $72.28, versus $156.78 required to send everything to the flagship model, saving $84.50, a cost reduction of about 54%.71% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1111474API
curl https://kongchang.com/api/v1/knowledge/claims/1111474MCP
get_claim(id=1111474)