Language modelActive

Qwen 3.5 397B A17B

Alibaba

Released: February 15, 2026
Data date: September 8, 2026
Successor: Qwen 3.8 2.4T A95B

Speed: 86.9 tokens/s
Output throughput
Price: $0.55 · $3.5
Input · Output / 1M tokens
Input: Text, Image, Video
Output: Text

Qwen 3.5 397B A17B is the largest open Qwen3.5 checkpoint, combining early image-text fusion with a sparse Mixture of Experts. Seventeen billion of its 397 billion parameters are active. The model card presents it as a multimodal foundation for coding, agents, and visual understanding, rather than simply a larger text generator.

The repository supports 262,144 native context tokens and extension to 1,010,000. Qwen names Qwen 3.8 2.4T A95B as a later open-weight expansion built on this foundation. For managed inference, Qwen3.5-Plus is a separate configuration with 1M context by default and built-in tools. Keep the weights and hosted API distinct.

Profile

Specifications and access

Published information about this model. Existing estimates are explicitly labeled.

Dated documentation and model-card observations. Provider limits, native context and extended context can differ. Configuration notes retain the source wording.

Specification	Value and source
Weights	Open weightsSource
Access	API, LocalSource
Parameters	397BSource
Active parameters	17BSource
Context window	262,144 tokensSource
Weights repository	Qwen/Qwen3.5-397B-A17BSource
License	Apache-2.0Source
Architecture	hybrid Gated DeltaNet + Gated Attention; dense or sparse MoE with vision encoderSource
Input	Text, Image, VideoSource
Output	TextSource
Reasoning mode	thinking by default; switchable non-thinking/instruct modeSource
Tool calling	YesSource
Structured output	YesSource
Native context window	262,144 tokensSource
Maximum context window	1,010,000 tokensSource
Parameters	397 BSource
Active parameters	17 BSource
Languages	201 languages and dialectsSource
Repository created	2026-02-16 (Repository creation date is a precise publication proxy, but may differ from the public announcement date or an individual checkpoint's first availability.)Source

Measurements without matching peer values

These measurements have no matching peer values under the same test conditions. Their original values and sources remain available here.

MMLU-Pro (Show measurement, test conditions, and source)

Source value: 87.8
Score: 87.8
Metric: percent
Unit: %
Direction: Higher is better
Comparison cohort: research:qwen-3.5-397b-a17b:0
Source type: vendor-reported
Evaluator: Qwen official Hugging Face model card
Status: active
Retrieved at: 2026-09-08
Methodology: Vendor-reported result. Official model-card comparison table; exact target column, setup as published

Source Qwen official Hugging Face model card

MMLU-Redux (Show measurement, test conditions, and source)

Source value: 94.9
Score: 94.9
Metric: percent
Unit: %
Direction: Higher is better
Comparison cohort: research:qwen-3.5-397b-a17b:1
Source type: vendor-reported
Evaluator: Qwen official Hugging Face model card
Status: active
Retrieved at: 2026-09-08
Methodology: Vendor-reported result. Official model-card comparison table; exact target column, setup as published

Source Qwen official Hugging Face model card

SuperGPQA (Show measurement, test conditions, and source)

Source value: 70.4
Score: 70.4
Metric: percent
Unit: %
Direction: Higher is better
Comparison cohort: research:qwen-3.5-397b-a17b:2
Source type: vendor-reported
Evaluator: Qwen official Hugging Face model card
Status: active
Retrieved at: 2026-09-08
Methodology: Vendor-reported result. Official model-card comparison table; exact target column, setup as published

Source Qwen official Hugging Face model card

C-Eval (Show measurement, test conditions, and source)

Source value: 93
Score: 93
Metric: percent
Unit: %
Direction: Higher is better
Comparison cohort: research:qwen-3.5-397b-a17b:3
Source type: vendor-reported
Evaluator: Qwen official Hugging Face model card
Status: active
Retrieved at: 2026-09-08
Methodology: Vendor-reported result. Official model-card comparison table; exact target column, setup as published

Source Qwen official Hugging Face model card

IFEval (Show measurement, test conditions, and source)

Source value: 92.6
Score: 92.6
Metric: percent
Unit: %
Direction: Higher is better
Comparison cohort: research:qwen-3.5-397b-a17b:4
Source type: vendor-reported
Evaluator: Qwen official Hugging Face model card
Status: active
Retrieved at: 2026-09-08
Methodology: Vendor-reported result. Official model-card comparison table; exact target column, setup as published

Source Qwen official Hugging Face model card

IFBench (Show measurement, test conditions, and source)

Source value: 76.5
Score: 76.5
Metric: percent
Unit: %
Direction: Higher is better
Comparison cohort: research:qwen-3.5-397b-a17b:5
Source type: vendor-reported
Evaluator: Qwen official Hugging Face model card
Status: active
Retrieved at: 2026-09-08
Methodology: Vendor-reported result. Official model-card comparison table; exact target column, setup as published

Source Qwen official Hugging Face model card

MultiChallenge (Show measurement, test conditions, and source)

Source value: 67.6
Score: 67.6
Metric: percent
Unit: %
Direction: Higher is better
Comparison cohort: research:qwen-3.5-397b-a17b:6
Source type: vendor-reported
Evaluator: Qwen official Hugging Face model card
Status: active
Retrieved at: 2026-09-08
Methodology: Vendor-reported result. Official model-card comparison table; exact target column, setup as published

Source Qwen official Hugging Face model card

AA-LCR (Show measurement, test conditions, and source)

Source value: 68.7
Score: 68.7
Metric: percent
Unit: %
Direction: Higher is better
Comparison cohort: research:qwen-3.5-397b-a17b:7
Source type: vendor-reported
Evaluator: Qwen official Hugging Face model card
Status: active
Retrieved at: 2026-09-08
Methodology: Vendor-reported result. Official model-card comparison table; exact target column, setup as published

Source Qwen official Hugging Face model card

LongBench v2 (Show measurement, test conditions, and source)

Source value: 63.2
Score: 63.2
Metric: percent
Unit: %
Direction: Higher is better
Comparison cohort: research:qwen-3.5-397b-a17b:8
Source type: vendor-reported
Evaluator: Qwen official Hugging Face model card
Status: active
Retrieved at: 2026-09-08
Methodology: Vendor-reported result. Official model-card comparison table; exact target column, setup as published

Source Qwen official Hugging Face model card

GPQA (Show measurement, test conditions, and source)

Source value: 88.4
Score: 88.4
Metric: percent
Unit: %
Direction: Higher is better
Comparison cohort: research:qwen-3.5-397b-a17b:9
Source type: vendor-reported
Evaluator: Qwen official Hugging Face model card
Status: active
Retrieved at: 2026-09-08
Methodology: Vendor-reported result. Official model-card comparison table; exact target column, setup as published

Source Qwen official Hugging Face model card

HLE (Show measurement, test conditions, and source)

Source value: 28.7
Score: 28.7
Metric: percent
Unit: %
Direction: Higher is better
Comparison cohort: research:qwen-3.5-397b-a17b:10
Source type: vendor-reported
Evaluator: Qwen official Hugging Face model card
Status: active
Retrieved at: 2026-09-08
Methodology: Vendor-reported result. Official model-card comparison table; exact target column, setup as published

Source Qwen official Hugging Face model card

HLE-Verified¹ (Show measurement, test conditions, and source)

Source value: 37.6
Score: 37.6
Metric: percent
Unit: %
Direction: Higher is better
Comparison cohort: research:qwen-3.5-397b-a17b:11
Source type: vendor-reported
Evaluator: Qwen official Hugging Face model card
Status: active
Retrieved at: 2026-09-08
Methodology: Vendor-reported result. Official model-card comparison table; exact target column, setup as published

Source Qwen official Hugging Face model card

Page 1 of 8

Model generations

Predecessor and successor

Every relationship names its kind, its qualification and the dated source it rests on.

Later model generation: Qwen 3.8 2.4T A95B
Next generation in the directory
Model lineage evidence
Alibaba's own launch post states Qwen3.8-Max (open-weighted as Qwen3.8-2.4T-A95B the following week) is 'built upon the robust foundation of Qwen 3.5', not Qwen3.6 or Qwen3.7; this is the only predecessor the source names, so the relation targets the flagship Qwen3.5 open-weight profile instead of the nearer-numbered Qwen3.7-Max.
Alibaba Cloud Community: "Alibaba Unveils Qwen3.8-Max: Its Largest and Most Capable Flagship Model to Date" (2026-09-19)

Evidence

Sources and data date

Every statement links to its underlying documentation or leaderboard.

Type	Evidence and data date
Speed	Alibaba. Output speed after the first streaming chunk, not total response time. Retrieved September 8, 2026. Source: Artificial Analysis · Artificial Analysis (retrieved September 8, 2026)Speed measurement methodology
Price	Base API prices without caching or batch discounts. Higher context tiers and other rates are listed under Costs. Source: Pricing · API pricing
Input	Documented modalities. Retrieved September 8, 2026. Source: Model documentation · Model metadata (retrieved September 8, 2026) · Qwen3.5-397B-A17B model card · Editorial description reviewed October 2, 2026
Output	Documented modalities. Retrieved September 8, 2026. Source: Model documentation · Model metadata (retrieved September 8, 2026) · Qwen3.5-397B-A17B model card · Editorial description reviewed October 2, 2026
Research status	Research date September 8, 2026. 3 source URLs checked. This documents the inspected sources, not an exhaustive inventory of every publication.
Not found in the inspected sources	knowledge-cutoff Original wording No exact training-data cutoff was stated in the checked official card/report.
Additional source	Model metadata (retrieved September 8, 2026)
Additional source	Alibaba Cloud Model Studio visual understanding · Editorial description reviewed October 2, 2026
Additional source	OpenRouter: Popularity index (derived), CC BY 4.0
Additional source	Vercel AI Gateway: Popularity index (derived), CC BY 4.0
Additional source	Creative Commons Attribution 4.0
Additional source	Alibaba Cloud Community: "Alibaba Unveils Qwen3.8-Max: Its Largest and Most Capable Flagship Model to Date" (2026-09-19)

Qwen 3.5 397B A17B

Specifications and access

Measurements without matching peer values

Our tests

Pelican on a bicycle

Popularity

Published prices

Predecessor and successor

Sources and data date