Z.ai: GLM 5.3
74.8
coding index
$1.400 / M input
$4.400 / M output
1.05M context
THE IDLE JOURNAL
The waiting room has a story.
Field notes on model speed, developer economics, respectful ads, and the strange new market for free AI tools.
Live data desk
What the model market costs today.
A small, refreshable snapshot from OpenRouter's public model catalog. Prices are per million tokens; benchmark scores are the Artificial Analysis indices embedded in that catalog.
Live catalog response · observed 2026-10-03T15:10:42.709Z
Read the scoreboard note →Qwen: Qwen3.8 27B
68.1
coding index
$0.420 / M input
$3.000 / M output
1.00M context
Google: Gemini 3.7 Flash
76.1
coding index
$0.750 / M input
$3.750 / M output
1.05M context
Featured row: Z.ai: GLM 5.3 · intelligence 44.8 · agentic 53.1
DATA NOTELive-source methodology
The leaderboard is not the latency.
OpenRouter prices, Artificial Analysis indices, and the number that actually changes a developer's day: how long the cursor sits there before the answer begins.
Read the data note →Comparison · 8 min
RamenSplit vs Freebuff
Two ad-funded ideas, pointed at different moments in a developer's workflow. What we share, where we diverge, and why that matters.
Read the comparison →From the notebook
3 live model rows aboveECONOMICS · 5 MIN
The math behind AI latency
A transparent unit-economics sketch for the dead air between prompt and answer.
Open note →ADVERTISING · 4 MINWhy developers block display ads
Relevance, restraint, and the case for one useful card instead of a noisy web.
Open note →FIELD COMPARISON · 8 MINTwo ways to fund free AI
A candid comparison with the free coding agent next door.
Open note →Sources: OpenRouter public model catalog and its linked Artificial Analysis benchmark fields. Model prices and rankings move; each article labels the observation date. RamenSplit product facts describe the current implementation, not a promise of earnings.