MetaMeta

Llama 3.3 70B Retired

This model is no longer available to run. Its stats below reflect past sessions and are kept for reference.

Meta's 2024 open-weight model. Second-highest checkout rate (67%) in our benchmark but the least efficient — averaging 11.8 turns per checkout. Retired: superseded by Llama 4 Maverick.

Avg Tokens58,192Avg Duration40.4sTurns to Checkout11.8
CheckoutCartSearchTurnsTokens39.352.782.1213
Shopping Score37/100
Fair

Scores computed from real agent sessions against live UCP-enabled stores. Not estimated — every data point is from an actual tool call, checkout attempt, and store response.

Shopping Score Breakdown

Checkout Rate40%39.3%
Cart Rate20%52.7%
Search Rate10%82.1%
Turn Efficiency15%2%
Token Efficiency15%13%
17% of sessions failed with errors.

Top Stores

StoreCheckout %Cart %
•••••••••••••••••••100%100%
••••••••••••••••••••••••••••••••100%100%
••••••••••••••••••••••••••••100%100%
everlane.com75%75%
•••••••••••••••66.7%83.3%
houseofparfum.nl58.3%62.5%
ucp.travel50%50%
••••••••••••••••••••••••••••••••40%40%
kyliecosmetics.com33.3%33.3%
••••••••••••••••••••••••••25%75%

Known Issues

No known issues documented for this model yet.

Token Usage

Avg per Session58,192
Fleet Average66,626
vs Fleet−13%
Median (p50)39,765
p90138,622
Prompt / Completion split
Prompt 57,687 (99%) Completion 504 (1%)
No daily trend data yet. Run more sessions to see token usage over time.
Distribution range
Min3,037p2518,233p5039,765p7571,067p90138,622Max392,012

Test Llama 3.3 70B on Your Store

Run a live agent session to see how Llama 3.3 70B handles your store's checkout flow end-to-end.

Run Agent Session