Skip to content

Qwen3.5 397B A17B

ModelOpen sourceActive
by Alibaba (Qwen) · family “qwen3b-a17b”

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... (description from the OpenRouter listing)

in: textin: imagein: videoout: textReasoningTool useStructured output
Epoch Capabilities Index
146.6 #67 of 268
90% CI 144.8 – 148.2
Listing ↗
Released
13 Feb 2026
Context window
262K tokens
Max output
236K tokens
Input $ / 1M tokens
$0.55
Output $ / 1M tokens
$3.50
Cached input $ / 1M
$0.23
Each benchmark shown separately with its own source

Benchmark results (8)

All benchmarks
BenchmarkDomainScorevs best recordedSettingRunSource
LMCAagents37.9%
56%
——External ↗
DTBenchreasoning87.5%
88%
——External ↗
Mystery Game Puzzlesgames18.0% ±3.9
21%
none27 Aug 2026Epoch ↗
FrontierMath-Tiers-1-3-v2-Privatemath31.2% ±2.7
33%
none30 Aug 2026Eval log ↗
APEX-Agentsagents24.9%
33%
——External ↗
Chess Puzzlesgames13.0% ±3.4
18%
none7 Aug 2026Eval log ↗
OTIS Mock AIME 2024-2025math88.9% ±4.7
89%
—7 Aug 2026Eval log ↗
GPQA diamondscience86.4% ±2.4
90%
none7 Aug 2026Eval log ↗

Source: Epoch AI Benchmarking Hub (CC BY 4.0). “External” rows are leaderboard results Epoch collects from third parties. Best reported setting per benchmark is shown.

API list price over time

Price history

$ per 1M tokens

8 recorded prices on 3 Mar 2026 (OpenRouter listing + Internet Archive snapshots); re-read on every data refresh, most recently 10h ago. Steps show when the price changed.

Events