TOPIC HUB

#benchmarks

Browse SynapseWire coverage, updates, and related reads about benchmarks.

CURATION NOTE

Topic hubs connect recurring themes across breaking news, hands-on reviews, tutorials, and strategic analysis.

#01 2026.10.21

Inception Mercury 3: 1,068 Tokens per Second and How Its Benchmarks Compare

Inception's Oct. 19 Mercury 3 is 1,068 tokens a second in three writeups and 'more than 1,000' in a fourth. One scorecard is blank. Another has Haiku ahead.

A steel desk by a tall window with a sealed kraft envelope, a closed notebook, an unplugged metal brick, and a coiled cable.

AI Industry Analysis

#02 2026.10.12

Haiku 5.5 Lists at $0.10. OSWorld Hit 72.4%.

Anthropic's short-prompt Haiku 5.5 rate is $0.10 in and $0.50 out. The 75% average saving is not the same cut as the 90% list-price drop.

A warehouse desk with printed order slips, a laptop closed, and a stamped ledger, cool overhead light, no logos and no readable text.

AI Industry Analysis

#03 2026.10.03

Opus 5.5 Cleared 34.8% of Argo-Bench

Argo-Bench, submitted Oct. 1, grades 210 warehouse tasks by consequences. The strongest of 14 models averages 59.5. A writeup names that model Opus 5.5.

Two office workers comparing notes at a long table with coffee cups and large windows.

AI News

#04 2026.08.17

Claude Opus 5 vs GPT-5.6: Which AI Model Should You Actually Use?

Two frontier AI models launched within two weeks of each other in July 2026. Here's how they actually compare on benchmarks, pricing, and real-world use cases.