Qwen3.5-122B-A10Bon Krater.ai

The Qwen3.5 122B-A10B native vision-language model is built on a hybrid architecture that integrates a linear attention mechanism with a sparse mixture-of-experts model, achieving higher inference efficiency. In terms of

Released February 25, 2026

Specifications

Qwen3.5-122B-A10B at a glance

Context

262.1K tokens

Max output

81.9K tokens

Input

Text, Image, Video

Output

Text

Input price

$0.29 / 1M

Output price

$2.40 / 1M

Image inputTool callingStructured outputsReasoning support

Capabilities

What it is good at

Catalog capabilities

  • Accepts image input
  • Generates text responses
  • Supports tool calling
  • Supports structured outputs

Good for

Image understandingLong documentsUse with toolsStructured responsesDeep analysis

Benchmarks

Catalog data, not Krater benchmarks.

Coding index score

45.7

Agentic index score

21.3

Intelligence index score

32.8

FAQ

Questions about Qwen3.5-122B-A10B

Clear answers based on the model catalog and Krater access.

Qwen3.5-122B-A10B supports a context window of 262,144 tokens.

Use Qwen3.5-122B-A10B in your API

Connect with Krater's unified API.

API access

Ready to use Qwen3.5-122B-A10B?

Access leading models in one focused workspace.