HolyGrAIL

All tasks / Models

Qwen3.8-2.4T-A95B

Alibaba's largest open-weights Qwen model (2.4T total, 95B active), the first Qwen-Max-class open release.

  • Alibaba (Qwen)
  • Open weights
  • Released Aug 2026

Official page

Where it ranks

  • B
    Chat & assistants
    Models · Local · 40 on Artificial Analysis Intelligence Index (secondary benchmark)
  • C
    Writing
    Models · Local · 1,843 on EQ-Bench Creative Writing v3 (secondary benchmark)
  • B
    Coding
    Models · Local · 40 on Artificial Analysis Intelligence Index (secondary benchmark)

Facts

Vendor
Alibaba (Qwen)
Family
Qwen
Released
Aug 2026
Weights
Open: download and run it yourself
License
Qwen3.8-Max license (custom)
Parameters
2400B (95B active)
GPU memory to run locally
~1370 GB (4-bit ≈ 2.4 T × 0.57 GB ≈ 1.37 TB, estimate (multi-node datacenter).)
API price
$2 in / $6 out per 1M tokens
Access
Free (open weights)
Last verified
18 Oct 2026

Price from openrouter.ai, 18 Oct 2026.

Benchmark results

BenchmarkResultEntry on the leaderboardAs ofSource
Artificial Analysis Intelligence Index40Qwen3.8 2.4T A95B
Raw value from the page's embedded model data; the site displays it rounded to an integer.
18 Oct 2026artificialanalysis.ai
EQ-Bench Creative Writing v31,843Qwen/Qwen3.8-2.4T-A95B
Elo score from the leaderboard data embedded in https://eqbench.com/creative_writing.js (file last modified 2026-09-24); rank 19 of 140; rubric score 16.72.
24 Sep 2026eqbench.com

Where you can use it

Earlier versions

  • Qwen3.5-397B-A17B · —

Sources