Alibaba

Qwen3.8 2.4T A95B

Intelligence
39.9
Rank 61 of 788 measured models
Coding
71.9
Rank 43 of 301 measured models
Output speed
39.1 t/s
Rank 417 of 454 measured models
Price/1M
$3
Rank 550 of 741 measured models
Context window
262K
Rank 582 of 1334 measured models
Cost per task
$2.16
Rank 176 of 235 measured models

Qwen3.8 2.4T A95B scores 39.9 on the Artificial Analysis Intelligence Index, ranking 61st of 788 measured models. The closest models on this index are Qwen3.8 Max (40.2), Gemini 3.7 Flash (39.6), and Muse Spark 1.2 (Xhigh) (39.6). At $3 per million tokens (3:1 input/output blended price), it is priced close to Qwen3.8 Max ($3), more expensive than Gemini 3.7 Flash ($1.50) and Muse Spark 1.2 (Xhigh) ($2). With an output speed of 39.1 t/s, it is slower than Qwen3.8 Max (57.2 t/s), Gemini 3.7 Flash (383 t/s), and Muse Spark 1.2 (Xhigh) (207 t/s).

Artificial Analysis Indices
Intelligence39.9
Rank 61 of 788 measured models
Coding71.9
Rank 43 of 301 measured models
Agentic50.1
Rank 21 of 191 measured models
Capability indexes
Performance in specific industries, measured by Artificial Analysis
Finance & Accounting41.3
Rank 86 of 260 measured models
Strategy & Ops44.4
Rank 90 of 260 measured models
Legal37.0
Rank 104 of 260 measured models
Healthcare & Medical39.0
Rank 72 of 190 measured models
Engineering40.3
Rank 73 of 259 measured models
Economics37.3
Rank 119 of 267 measured models
Cyber Index
Artificial Analysis · v1
Cyber Index12.6%Rank 26 of 26 measured models
Successes
12.6%
Safety blocks
62.5%
Failures
24.9%
CWE-Bench-AA28.3% (safety blocks)
DeepsecBench-AA9.5% (safety blocks)
CyberGym-E2E-AA0.0% (safety blocks)
Cost per task—

* The model or provider declined some tasks on safety grounds.

Standard Benchmarks
GPQA Diamond93.5%
Rank 14 of 621 measured models
HLE42.4%
Rank 75 of 682 measured models
SciCode54.1%
Rank 73 of 249 measured models
AA-LCR80.3%
Rank 79 of 593 measured models
Terminal-Bench 2.182.0%
Rank 39 of 246 measured models
Terminal-Bench 4.011.1%
Rank 107 of 246 measured models
τ-Bench Banking49.1%
Rank 5 of 211 measured models
Models at a similar level
The closest models on the Intelligence Index, one configuration per family.

Preferences and analytics

Rybbit: visits, performance, clicks, copies, forms and session replay (masked inputs), with your permission. Refusing does not limit the site.

Privacy

Allow navigation analytics?

With your permission, Rybbit measures visits, performance and interactions (clicks, copies, forms), including session replay, to improve BenchSift. Input values are masked in replays. Refusing does not limit the site.

Your choice is stored for 180 days. Change it at any time using “Privacy preferences” in the footer. Withdrawal reloads the page.

No choice stored: Rybbit is disabled.

Privacy