Ordinal

Which model is worth using

1 Oct 2026

Alibaba

Qwen3.8 2.4T A95B

The open Qwen MoE on the leaderboard at index 40. One published score. No list price in this snapshot.

Open weights
40Index · Published
$2.16Per measured task
18.5 pts/$Points / $
40Tokens / sec

No High score is published. Showing Published.

What this workload costs

List price for one call. Reasoning tokens count as output. A cache hit uses the lab’s cache rate when one is published.

Qwen3.8 2.4T A95B
No list price on this card.
—

Specs

API id
Not published here
Released
Sep 2026
Context
262k
Max output
—
Input / 1M
—
Output / 1M
—
Cache read / 1M
—
Knowledge
—
Inputs
Text
Weights
Open weights. Confirm whether this checkpoint uses the Qwen license.

No token price is attached here. The leaderboard does publish a measured task cost. 2.4T total, 95B active, 262k window.

Use it when

An open Qwen MoE when Max's API price or Max's measured task cost is the problem, and a 262k window is enough.

Leave it when

You needed a token price to approve the spend. It is not on this card. Use the measured task cost as a comparison, then read the host's rate card.

Sources