The release decision platform for AI builders

Every tracker shows you what shipped.
We show you what's next.

Not just a calendar — a decision tool. Live Polymarket + Kalshi odds and a Drop Readiness score on every forecast, a plan-around-this horizon, real adoption data on what builders actually run, plus cost, deprecation, and price-cliff alerts. Refreshed hourly, every source linked. Markets are signal, not stakes.

THIS WEEK live
3 drops imminent
in the next 30 days
Next up Gemini 3.5 Pro T−9d
RELEASE VELOCITYJun proj ↑12
Planning horizon

Plan around this

Your horizon as one agenda — what to evaluate, migrate, or re-budget, and when. The roadmap the labs don't publish, synthesized from release windows and lifecycle dates.

Forecasts · ranked by Drop Readiness

What ships next

The ranked forecast — what the timeline shows, as a sortable board. One number per release, blended from market odds, intel recency, deadline proximity, and volume.

Benchmark frontier

Who leads now — and who's about to

Competitors show the current benchmark leader. We also show the challenger in the forecast window expected to take the crown, and when.

Adoption signal

What builders actually run

Real usage from community model-routing traffic — the reality-check leaderboard scores can't give you. Directional developer adoption, not whole-market totals; share is relative to the most-used model.

Model database · last 60 days

Recently shipped

Confirmed launches with the specs builders filter on — context window, pricing, license — plus a comparable cost tier, the cheapest option in each class (★), and a flag when a price hike or deprecation is scheduled.

MODELPROVIDERSHIPPEDTYPECONTEXTPRICE IN / OUTLICENSE
Run it locally

Can you run it?

Open-weight models with the VRAM to actually run them — estimated from parameters and quantization, sized on total params (MoE holds every expert in memory), at ~16K context. Estimates, ±~15%; real usage varies by context and runtime.

MODELPARAMSQ4_K_MQ8_0FP16RUNS ON (Q4)
Migration watch

What you're about to lose

Deprecations, sunset dates, and intro-pricing cliffs — each with the model to move to and how long you have. The intel labs bury in a changelog and every other tracker skips.

Codename watch

Codenames, decoded

Internal names leak before model names do — from public model-catalog probes, exec accounts, and pre-release SKUs. We map each to the public model it likely becomes.

Hourly intel

The briefing

What changed last cycle, graded by evidence — Confirmed, Signal, or Speculation, strongest first. The chatter and changelog feeds link every dated source.

WHAT CHANGED THIS CYCLE pull it anywhere · JSON /api/intelligence · RSS /api/feed
FROM THE WIRE graded · Confirmed › Signal › Speculation
X CHATTER
CHANGELOGS
Compare

Why nextaidrop, not a release log

Editorial trackers can list rumored dates; none price them. We're the only forward timeline backed by live Polymarket + Kalshi odds and a Drop Readiness score.

FAQ

Common questions