Qwen3.7 Plus

Qwen3.7 Plus is a multimodal agent model from Alibaba built on top of the Qwen3.7 text backbone, sitting alongside the heavier Qwen3.7-Max in the same family. It is described as a perception model rather than a generation model: it takes in text, images, and video, but only produces text in response. That design choice shapes its practical role, letting it look at a screen, watch a short clip, or read a diagram, then reason about the scene and act on what it has seen, whether that means writing code from a mockup, clicking through a user interface, or answering a question about visual content.

The model's fit is in interactive, perception-driven workflows where understanding visual context matters as much as language skill. Because it keeps the coding, reasoning, and tool-use strengths associated with its Max sibling while adding visual perception, it works well as a hybrid agent for tasks like screen reading, GUI operation, and diagram-aware problem solving. Independent catalog trackers position it as a well-rounded option with broad multilingual coverage, though practitioners should validate behavior on their specific workloads since third-party benchmark coverage is still partial and no first-party token-rate numbers have been published.

Kilo Gatewayqwen/qwen3.7-plusqwen

Open provider page Provider docs

Quick Info

Provider: Kilo Gateway
Model key: qwen/qwen3.7-plus
Release date: Jun 2, 2026
Last updated: Jun 2, 2026
Knowledge cutoff: 2025-04
Input modalities
Output modalities
Capabilities

Cost

Input token cost: $0.32
Output token cost: $1.28

Limits

Output tokens: 131,072 tokens
Context window: 1,000,000 tokens

Transparent token rates

Compare Qwen3.7 Plus pricing

Rates are shown per one million tokens. Combined means one million input plus one million output tokens.

Browse this family

Latest news about Qwen3.7 Plus

OrcaRouter

Coverage

Alibaba's Qwen team released Qwen3.7-Plus on June 2, 2026, as a multimodal agent model that combines visual perception, GUI control, and code generation within a single autonomous agent loop. The model accepts text, images, and video as input but outputs only text, enabling it to read screens, navigate apps, write code Qwen3.7-Plus leads Alibaba's benchmarks on screen-grounded agent tasks, scoring 79.0 on ScreenSpot Pro versus GPT-5.4's 67.4 and 81.0 on AndroidWorld versus Gemini 3.1 Pro's 70.7, though it trails on pure reasoning tasks. Pricing is set at $0.40 per million input tokens and $1.60 per million output tokens, roughly 6× c

OrcaRouter

CoverageBenchmark

BenchLM.ai rates Qwen3.7-Plus with a Capability score of 55.8/100, ranking it 53rd of 196 tracked models as of September 23, 2026. Speed is measured at 64 tok/s with a first-token latency of 33.7 seconds, and context window is reported at 1M tokens. The model has 52 published benchmark rows across its profile. Category-level rankings show strong multilingual performance (rank 3 of 12, 82nd percentile, score 78.9) and solid multimodal results (rank 19 of 50, 63rd percentile, score 72.4). It ranks 57th of 105 in Agentic (46th percentile, score 35.1), 51st of 135 in Coding (63rd percentile, score 44.3), and 18th of 124 in Instr

OrcaRouter

CoverageBenchmark

Qwen 3.7 Plus launched on June 1, 2026, eleven days after Qwen 3.7 Max, sharing the same 1M context window and 35-hour autonomous run ceiling. The headline difference is pricing: Plus costs $0.40/M input versus Max's $2.50/M, roughly 6× cheaper across input, output, and cached tokens, while also adding vision (text + i For most coding, doc, and agent workloads, Plus is positioned as the default value-tier choice, with Max only earning its 6× premium when a measurable quality win on a specific task mix can be demonstrated. Both models ship through Alibaba's Bailian platform and support 1000+ sequential tool calls. The recommendation t

OrcaRouter

CoverageBenchmark

According to the LLM Stats composite scoring, Qwen3.7-Plus ranks 49th across tracked benchmarks, with standout performance in Legal (rank 2 of 211) and Healthcare (rank 2 of 245), placing it in the top 2% in those categories. It also ranks in the top 10% for Chat (7 of 133) and Long Context (9 of 119), while landing at Specific benchmark scores include IFEval at rank 2 (0.95), MMLU-Redux at rank 3 (0.94), HMMT Feb 26 at rank 4 (0.93), and MRCR v2 at rank 1 (0.92) with a 128k methodology. Quality tracking shows a stable +0.98σ deviation with 62 votes over the past week. The scores are self-reported by the model provider (qwen.ai) and