AiCostCompare

Qwen3.5 397B A17B

by Alibaba (Qwen) · qwen/qwen3.5-397b-a17b · #258 cheapest of 422 paid models

Prices updated Sep 19, 2026, 6:21 PM UTC · refreshed hourly

The Qwen3.5 series 397B-A17B 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. It delivers...

Compare vs:Claude Sonnet 5Claude Opus 4.6Claude Haiku 4.5GPT-5.6 Sol

Pricing

Input

$0.55

per 1M tokens

Output

$3.50

per 1M tokens

Blended (3:1)

$1.29

3 input : 1 output

Cache read

$0.225

per 1M cached tokens

Cache write

per 1M tokens

Cache & batch economics

Effective prices for Qwen3.5 397B A17B with prompt caching.

Cache hit rateEffective blended $/1MExample request*vs no cache
0%$1.29$0.0596
50%$1.17$0.0434−27%
90%$1.07$0.0303−49%

*Example: 100K sticky context + 2K new input + 1K output. Batch discount is an approximate 50% off list rates (available for OpenAI / Anthropic / Google — not flagged for this creator). Confirm on the vendor’s pricing page.

What a request costs

WorkloadInput tokensOutput tokensCost / requestCost / 1K requests
Short chat message500300$0.001325$1.32
Document summary8,0001,000$0.0079$7.90
Codebase question (RAG)30,0002,000$0.0235$23.50
Long-context analysis150,0005,000$0.1$100.00

Performance

Intelligence Index

19.1

AA composite quality

Coding Index

48.2

Math Index

Agentic Index

10.6

Output speed

86 tok/s

median

Time to first token

1.62s

median

Time to first answer

38.85s

after reasoning tokens

Artificial Analysis benchmarks

BenchmarkScore
GPQA Diamond89.3%
Humanity's Last Exam29.0%
SciCode44.8%
IFBench78.8%
AA-LCR (Long Context Reasoning)77.3%
Terminal-Bench Hard40.9%
Terminal-Bench 2.151.3%
τ²-Bench (Telecom)95.6%
τ³-Bench Banking13.4%

Design Arena Elo

CategoryElo
Website1,203
UI components1,178
Data viz1,190
SVG1,163
3D1,190
Game dev1,163
Code categories1,195

Specs

Context window

262K

tokens

Max output

236K

tokens

Input modalities

text, image, video

Tokenizer

Qwen3