Qwen3.5 family of multimodal models (Alibaba/Qwen Team) with MoE + Gated DeltaNet architecture, from 0.8B to 397B-A17B, under Apache 2.0.
Context window
262K (do ~1M)
tokens
Parameters
397B / 17B aktywnych (flagowiec MoE); rodzina 0.8Bโ397B
parameters
Release date
16 February 2026
Access:DownloadAPIHostedDeployment:๐ป Localโ Cloud
Overview
Access & deployment
DownloadAPIHosted
LocalCloud
Weights: Open source
Key parameters
๐ Context: 262K (do ~1M)
๐งฉ Parameters: 397B / 17B aktywnych (flagowiec MoE); rodzina 0.8Bโ397B
โ Toolsย ยทย โ Fine-tuning
๐ฅ Input: text, image, video
Technical specification
Context window
262K (do ~1M)
tokens
Parameters
397B / 17B aktywnych (flagowiec MoE); rodzina 0.8Bโ397B
parameters
License
Apache 2.0
Features:โ Tool useโ Fine-tuning
Modalities
โฌ Input
textimagevideo
โฌ Output
textcode
Capabilities and applications
Native model capabilities
Reasoning
The model's ability to reason logically and solve complex problems.
Category: reasoning
Extended thinking mode
A reasoning-model variant with a larger inference budget: more thinking cycles, higher answer precision at the cost of response time. Choice between 'standard' and 'extended' thinking is left to the user (e.g. the selector in GPT-5.2 Pro).
Category: reasoning
Multimodal understanding
Category: multimodal
Image understanding
Analysing and interpreting the content of images.
Category: vision
Video understanding
The model's ability to analyse and interpret video content โ recognising actions, motion, events and relationships between objects over time.
Category: video
Long context
Support for large context windows โ tens to hundreds of thousands (or millions) of input tokens. Enables analysis of entire codebases, long documents, and many parallel conversations without losing earlier information. GPT-5.1 supports 400,000 tokens.
Category: language
Multilingual
Competence in many natural languages (from a few to over a hundred): understanding, generation, translation, and code-switching within a single conversation. Frontier models support a wide range of languages with comparable quality.
Category: language
Coding
Generating, analysing and modifying code in many programming languages. Covers writing functions, debugging, refactoring, code review, and creating tests. Measured by benchmarks such as HumanEval and SWE-bench.
Category: coding
Tool use
The model's ability to call external functions, APIs and tools during a conversation: calculator, search engine, code editor, database. The model decides when and how to use a tool and interprets its result.
Category: planning
Instruction following
Precisely following instructions contained in the prompt: response format, length, style, constraints (e.g. 'reply in six words'). GPT-5.1 significantly improved this capability compared to GPT-5.
Category: language
Mathematical reasoning
The model's ability to solve mathematical tasks requiring multi-step reasoning โ equations, proofs, combinatorics, geometry, calculus and competition-level problems.
Category: reasoning
Agentic capability
The model's ability to autonomously plan and execute multi-step tasks by sequentially using tools, maintaining context, and adapting to intermediate results.
Category: planning
Benchmark results
5 benchmarks
MMLU-Pro
87.8
๐ technical_report
Qwen3.5-397B-A17B (flagship)
GPQA
88.4
๐ technical_report
Qwen3.5-397B-A17B (flagship)
MathVision
88.6
๐ technical_report
Qwen3.5-397B-A17B (flagship)
SWE-bench Verified
76.4
๐ technical_report
Qwen3.5-397B-A17B (flagship)
MMMU
85.0
๐ technical_report
Qwen3.5-397B-A17B (flagship)
Technical architecture
Core Architecture
Model Form
Training Techniques