AlibabaReleased Aug 8, 2026
Open weights

Qwen3.8 2.4T A95B

Strongest in Academic Reasoning (#37 of 223) and Computer Use (#33 of 186); weakest in Reliability (#58 of 188). Abstract Reasoning is estimated, because no results are published yet.

Composite IQ ?
122
#46 of 186 ranked models · ahead of 76%
Alibaba's #4 model, behind Qwen4 Max (128)
Effective cost
$13.22/1M
$2 in · $6 out
Speed
41tok/s
65s typical response
Context
1Mtokens
1,010,000 tokens in
Evidence
6of 7 dims
14 of 39 benchmarks

IQ profile

Seven dimensions, averaged equally into Composite IQ
Compare shape with
100 avg human130160● Abstract100#80 · est.● Math126#47● Academic131#37● Coding131#47● Computer Use126#33● Reliability118#58● Knowledge Work126#34
MeasuredEstimated95% range
#1 in dimensionTop-10 median

IQ dimensions

Ranks are among models scored in each dimension. Ranges are model-based estimates, not calibrated confidence intervals.

DimensionIQPosition · 70–160RankEvidencevs Top 10
100estimated
#80 / 221
0/5
-32.5
126measured
#47 / 186
1/5
-23.5
131measured
#37 / 223
4/6
-4
131measured
#47 / 187
1/6
-20
126measured
#33 / 186
1/5
-20.5
118measured
#58 / 188
2/6
-17.5
126measured
#34 / 186
5/6
-6
Qwen3.8 2.4T A95B95% range (model-based, uncalibrated)Top-10 median#1 in dimension100 = average human

Benchmark results

14 of 39 Composite IQ benchmarks have a published result for Qwen3.8 2.4T A95B. Benchmarks without a result stay listed so gaps are visible.

Abstract Reasoning IQ 100 · estimatedNo published results0/5
No published results yet for Abstract Reasoning. The 100 shown is estimated from Qwen3.8 2.4T A95B's other dimensions and will be replaced as results come in. How estimates work
ARC-AGI-3ARC-AGI-2ARC-AGI-1Chess PuzzlesMystery Game Puzzles
Mathematical Reasoning IQ 126 · measuredBest: #29 of 70 on MathArena1/5
BenchmarkScorePercentileRankLeader · source
MathArena
Full source leaderboard reconciliation · Nov 18
30.8%
#29 / 70
Leader gpt-6.2-astra · 92.2%
Full source leaderboard reconciliation · Nov 18
Not yet reported
FrontierMath Tier 4FrontierMath Tier 1-3ProofBench 1.1AIME
Academic Reasoning IQ 131 · measuredBest: #15 of 147 on GPQA Diamond4/6
BenchmarkScorePercentileRankLeader · source
GPQA Diamond
Artificial Analysis v5.0.1 · Nov 13
93.5%
#15 / 147
Leader gpt-6-astra · 96.1%
Artificial Analysis v5.0.1 · Nov 13
CritPt
Artificial Analysis v4.3.2 · Oct 27
20.0%
#26 / 159
Leader Muse Wonder Full · 32.9%
Artificial Analysis v4.3.2 · Oct 27
Humanity's Last Exam
Artificial Analysis v4.3.2 · Oct 27
42.4%
#38 / 167
Leader opus-5.5 · 61.4%
Artificial Analysis v4.3.2 · Oct 27
SciCode
Artificial Analysis v4.3.2 · Oct 27
54.1%
#42 / 167
Leader opus-5.5 · 66.9%
Artificial Analysis v4.3.2 · Oct 27
Not yet reported
MMLU-ProMMMU-Pro
Programmatic Reasoning IQ 131 · measuredBest: #46 of 109 on Terminal-Bench 4.01/6
BenchmarkScorePercentileRankLeader · source
Terminal-Bench 4.0
Artificial Analysis v5.0.1 · Nov 13
11.1%
#46 / 109
Leader Kimi K3.1 Pro · 67.2%
Artificial Analysis v5.0.1 · Nov 13
Not yet reported
LiveCodeBenchIOI V2ProgramBenchSWE-rebenchMirrorCode
Computer Use IQ 126 · measuredBest: #7 of 35 on MCP Atlas1/5
BenchmarkScorePercentileRankLeader · source
MCP Atlas
Full source leaderboard reconciliation · Oct 23
84.5%
#7 / 35
Leader Gemini 4 Argon · 90.7%
Full source leaderboard reconciliation · Oct 23
Not yet reported
BrowseCompOSWorld-VerifiedArena.ai Agent ArenaAgents' Last Exam
Reliability IQ 118 · measuredBest: #43 of 154 on AA Long Context Reasoning v1.12/6
BenchmarkScorePercentileRankLeader · source
AA Long Context Reasoning v1.1
Artificial Analysis v4.3.2 · Oct 27
80.3%
#43 / 154
Leader kimi-k3 · 88.7%
Artificial Analysis v4.3.2 · Oct 27
AA Omniscience
Artificial Analysis v4.3.2 · Oct 27
4.3
#57 / 158
Leader opus-5.5 · 46.4
Artificial Analysis v4.3.2 · Oct 27
Not yet reported
SimpleQA VerifiedBullshitBench v2IFBenchFACTS Grounding
Knowledge Work IQ 126 · measuredBest: #22 of 124 on GDPval-AA v2.1 normalized score5/6
BenchmarkScorePercentileRankLeader · source
GDPval-AA v2.1 normalized score
Artificial Analysis v4.3.2 · Oct 23
55.3%
#22 / 124
Leader opus-5.5 · 68.3%
Artificial Analysis v4.3.2 · Oct 23
EnterpriseOps-Gym-AA
Full source leaderboard reconciliation v1117-task-strict-2026-09-29 · Oct 27
47.4%
#8 / 31
Leader fable-5 · 51.1%
Full source leaderboard reconciliation v1117-task-strict-2026-09-29 · Oct 27
AA-Briefcase v1.1 Elo
Artificial Analysis v5.0.1 · Nov 13
1,430
#31 / 116
Leader sonnet-5.5 · 1,816
Artificial Analysis v5.0.1 · Nov 13
AutomationBench-AA
Artificial Analysis v5.0.1 · Nov 13
57.2%
#32 / 112
Leader Gemini 4 Carbon · 77.8%
Artificial Analysis v5.0.1 · Nov 13
GDP.pdf
Artificial Analysis v5.0.1 · Nov 13
15.6%
#43 / 111
Leader Muse Wonder Full · 32.6%
Artificial Analysis v5.0.1 · Nov 13
Not yet reported
AA-AnalystAgent

Release & availability

Facts from the publisher's announcement and documentation.

Released
Developer
Alibaba
Access
Open weights for local deployment and hosted access through Qwen Cloud
Weights
Open weightsDownloadable and self-hostable
Inputs
text
Context
1,010,000 tokens
API price
$2 in · $6 outper 1M tokens · full pricing

What Alibaba says

  • Qwen3.8-2.4T-A95B is the open-weight checkpoint underlying the Qwen3.8 Max family, with 2.4T total and 95B active parameters.
  • It has a native 262k-token context window that can be extended to about 1.01M tokens.

Limitations

  • Unlike the hosted Max configuration, the open checkpoint does not include vision input or non-thinking mode.

Publisher statements, not AI IQ measurements.

About this score

How much of Qwen3.8 2.4T A95B's IQ is measured, and how the rest is estimated.

14/ 39benchmark results
6/ 7dimensions measured
1estimated
0/5Abstract
1/5Math
4/6Academic
1/6Coding
1/5Computer Use
2/6Reliability
5/6Knowledge Work
Bar height = share of benchmarks with resultsEstimatedNo score
Measured results are shown as published

Every score in the benchmark tables comes from its source unchanged. Composite IQ averages all seven dimensions equally.

Estimates fill the gaps

Abstract Reasoning has no published results yet, so its IQ is estimated from Qwen3.8 2.4T A95B's other dimensions. Missing benchmarks can also draw on its predecessor, Qwen3.8 Max; a predecessor is an estimation input, not evidence that this model has the predecessor's measured results.

A missing result is not a zero

Benchmark versions and evaluation harnesses can differ between sources, so small gaps between models may not be meaningful.

Read the scoring methodology →

Compare Qwen3.8 2.4T A95B

Side-by-side pages for the models it is most often weighed against.

Qwen3.8 2.4T A95BComparison model
Browse all comparisons →

Frequently asked questions

Quick answers about Qwen3.8 2.4T A95B, drawn from the same data as the rest of this page.

How AI IQ scores models →Get this data from the API →
What is the IQ of Qwen3.8 2.4T A95B?
Qwen3.8 2.4T A95B has an estimated IQ of 122, which ranks #46 out of 186 public models on the AI IQ leaderboard. The score combines seven equally weighted reasoning dimensions using measured benchmark results and conservative estimates for missing coverage.
What is Qwen3.8 2.4T A95B best at?
Among the dimensions with published benchmark results, Qwen3.8 2.4T A95B ranks highest in Academic Reasoning (IQ 131, #37 of 223) and Computer Use (IQ 126, #33 of 186). Its lowest-ranked measured dimension is Reliability (IQ 118, #58 of 188).
What is the EQ of Qwen3.8 2.4T A95B?
Qwen3.8 2.4T A95B doesn't have an EQ score yet. EQ (emotional reasoning) is calculated from EQ-Bench 4 and AttuneBench, and neither has published a result for Qwen3.8 2.4T A95B. AI IQ doesn't estimate EQ from other benchmarks, so the score appears once one of those results is published.
Which benchmarks has Qwen3.8 2.4T A95B been tested on?
Qwen3.8 2.4T A95B has published results on 14 of the 39 benchmarks behind AI IQ's Composite IQ, including GPQA Diamond, CritPt, GDPval-AA v2.1 normalized score, MCP Atlas and Humanity's Last Exam.
How much does Qwen3.8 2.4T A95B cost?
Qwen3.8 2.4T A95B costs $2.00 per million input tokens and $6.00 per million output tokens. Adjusted for how many tokens it actually consumes on real tasks, its effective cost works out to $13.22 per million input-plus-output tokens.
Is Qwen3.8 2.4T A95B open source?
Yes — Qwen3.8 2.4T A95B is an open-weights model: its trained weights are published, so it can be downloaded and self-hosted.
Who developed Qwen3.8 2.4T A95B?
Qwen3.8 2.4T A95B was developed by Alibaba, based in China.
When was Qwen3.8 2.4T A95B released?
Qwen3.8 2.4T A95B was publicly released on August 8, 2026.
How large is the context window of Qwen3.8 2.4T A95B?
Qwen3.8 2.4T A95B supports a context window of 1,010,000 tokens — the amount of text it can consider in a single request.
How fast is Qwen3.8 2.4T A95B?
In independent measurements, Qwen3.8 2.4T A95B generates about 41 tokens per second, with a typical response time of about 65 seconds.