ÆAAI
← All models
QWN35 · Qwen 3.5SynthesizedOpen Weights

Qwen3.5 Plus

Alibaba's open-weights frontier — 1M context at $0.26/$1.56 per MTok, strong on coding and math.

Accounting overall

66.6%

Input / Output

$0.26 / $1.56 per MTok

Context

1M

Speed

~100 tok/s

Released

2026-02

Cutoff

2025-12

01Accounting Task Breakdown

Eight accounting-task categories borrowed from DualEntry's 101-task benchmark. Measured where published, synthesized from adjacent benchmarks otherwise.

Transaction Class.
80.0
Journal Entry
78.0
Accounts Payable
70.0
Accounts Receivable
69.0
Bank Reconciliation
65.0
Financial Reporting
53.0
Month-End Close
41.0
Accounting Knowledge
77.0
02Research

Qwen 3.5-Plus released February 15, 2026. The headline variant Qwen3.5-397B-A17B (397B total, 17B active MoE) is Apache-2.0 open-weights. Reported benchmarks: 83.6 LiveCodeBench v6, 91.3 AIME26, 85.33% LiveCodeBench (edging Opus 4.6 at 84.68%). Alibaba also ships Qwen3.5-Flash at $0.10/MTok — roughly 1/13th Sonnet 4.6's rate on comparable tasks.

Accounting sub-scores are **synthesized** — no DualEntry coverage. We estimate ~70% overall based on adjacent benchmark positioning. GDPVal-AA currently lists the sibling Qwen3.6 Max at 1,510 Elo (#10) which implies Qwen 3.5-Plus is likely a touch below that.

For price-sensitive, long-context agentic accounting workloads — especially where a 1M-token context matters for multi-year ledger ingestion — Qwen3.5-Plus is the cheapest serious option after Grok 4.1 Fast. The 1M-context-at-flat-pricing combination is rare at this price tier.

Citations