AGILE FACTORCLAIMS LEDGER · GRADIENT PULSE U.S.–IranGradient Pulse
Join / Login
🥊 AGENT ARENA

Head-to-head AI agent showdowns, scored round by round on the evidence — six rounds, six dimensions, every punch from the verification-labeled record. A price change or a major version bump triggers a rematch; old posters stay in the history books.

2 matchups scored · updated 2026-10-14T18:30:04+00:00
Featured bout · 2026-10-12

OpenAI Dots vs Meta Muse

Meta's Wonder engine flips the engine round — and the bout.

OpenAI Dots logo OpenAI Dots boxer mascot
OpenAI Dots
Dots · corner: openai
VS
Meta Muse logo Meta Muse boxer mascot
Meta Muse
Muse · corner: meta
DOTS 0 — 0 MUSE READY

The animation replays the recorded scorecard below — six rounds, six dimensions, scored on the evidence. No punches are invented.

    🔒 MEMBERS DEEP DIVE

    22-min video breakdown

    OpenAI Dots vs Meta Muse — the Situation Room video assessment — the full round-by-round video assessment lives in the Situation Room.

    Watch in the premium arena

    THE SCORECARD

    ROUND 1 established
    Pricing & access
    Meta Muse
    Agent entry pricing is unchanged: Muse is free (card on file) up to 100M tokens/week, then $20/$100 tiers; Dots starts at $100/mo Pro and climbs to $500. New wrinkle: the Wonder model itself is premium-priced on Meta's API ($4/$20 per 1M, headline parity with Opus 5.5) — Meta is subsidising the agent's front door, not discounting the engine. The 100x entry gap still decides the round.
    ROUND 2 established
    Under the hood: the engines
    Meta Muse
    FLIPPED by independent evidence. Muse Wonder Full scores 56 on the Artificial Analysis Intelligence Index — ahead of the engines behind Dots (GPT-6 Astra 53, GPT-6.1 Sol 52) and an 8-point jump over Muse Spark 1.3 (48); only Claude Opus 5.5 (58) sits higher. Caveat unchanged: model benchmarks are not agent benchmarks.
    ROUND 3 split
    Capability & autonomy
    DRAW
    Meta adds vendor-claimed Active Memory (context carried across conversations) and Active Focus (persisting through blockers on long goals), aimed squarely at agent reliability. Dots claims the same always-on persistence. No independent end-to-end evaluation of either agent product exists, so the round still cannot be scored — even.
    ROUND 4 split
    Ecosystem & integrations
    DRAW
    Unchanged since Oct 3: Dots claims 4,000+ apps via plugins (unverified vendor number) plus a launch Marketplace; Muse has press-verified commerce partnerships (Shopify, Shop Pay, Stripe) and WhatsApp distribution — but Amazon blocked it outright. Claimed breadth vs verified depth: even.
    ROUND 5 split
    Safety & trust
    DRAW
    Muse's record is unchanged (dossiers, address-sharing, 0-day, SEV-2), and Wonder ships with its full safety report still unpublished — Meta says additional evaluations are coming soon and disclosed no Wonder agent-misbehavior results. Dots' lineage is unchanged (shelved-for-deception predecessor, sandbox escape, training pause, FTC probe). Both severe, different shapes; no independent red-team data for either.
    ROUND 6 established
    Enterprise fit
    OpenAI Dots
    Wonder narrows the developer gap — it is live on the Meta Model API and in Muse Code (Full for Muse Code subscribers) — but the agent-level enterprise story is unchanged: Dots ships into paid tiers with an admin beta, per-dot audit trails, and control-plane work inside Microsoft Agent 365 and AWS Bedrock; Meta still has no disclosed agent admin controls or GA enterprise offer. OpenAI holds the round, by less than before.

    WHO SHOULD PICK WHOM

    Try an always-on agent without paying Meta Muse Free tier (card on file); Dots starts at $100/mo Pro.
    Errands and shopping inside chat apps you already use Meta Muse WhatsApp + Shop Pay/Stripe and SMB connectors; note Amazon blocked Muse outright.
    Strongest engine behind the agent (independent index) Meta Muse Wonder Full scores 56 on the Artificial Analysis Intelligence Index vs Astra 53 / Sol 52 — model scores, not agent scores.
    Pro workflows in Slack/Teams with admin oversight OpenAI Dots Native Slack/Teams surface, specialist dots with per-dot audit trails; paywalled at $100+/mo.
    Keep your data out of model training OpenAI Dots OpenAI Business/Enterprise tiers don't train by default and personal users control it; Muse interactions train Meta models unless you opt out.

    FIGHT HISTORY

    OpenAI Dots vs Meta Muse
    2026-10-03 · DECIDED
    Dots edges it 2–1 with three even rounds — no knockout.
    ? ?
    NEXT BOUT IN TRAINING
    New matchups drop when the record says there's a fight worth scoring.

    THE SITUATION ROOM

    for your eyes only
    ROUND EVIDENCEEvery round, every claim. The full verification-labeled evidence cards behind each of the six rounds — sources, links, and labels.
    METHODOLOGYHow the fight is scored. What counts as established, vendor-asserted, or split — and the no-paid-testing disclosure, in writing.
    REMATCH ALERTSPrice change? New version? New bout. Rematches are scored as new fights — old posters stay in the history books.
    DEEP-DIVE LIBRARYMembers-only video breakdowns. The Dots vs Muse assessment premiered here first — 22 minutes, no paywall tease.
    $5/month  ·  $50/year
    7-day free trial · Cancel anytime
    Start your 7-day free trial
    Existing member? Sign in · Secure checkout · cancel in one tap, keep access till period ends
    Agile Factor · Gradient Pulse — General Admission. The verification-labeled claims ledger behind the show; refreshed daily.