FlashLabs Enhances OrcaRouter with Alibaba Qwen 3.7 Max API Support

OrcaRouter's Enhancement with Qwen 3.7 Max API

FlashLabs, a Tokyo-based company, has recently made significant improvements to its AI routing platform, OrcaRouter, by integrating support for Alibaba's Qwen 3.7 Max API. This upgrade provides a multifunctional gateway specifically designed for enterprise AI agent workflows, enhancing capabilities in managing complex reasoning tasks and long document processing.

Background and Objectives

The necessity for AI usage has surged significantly as business products evolve, introducing new cost factors. Traditionally, companies faced a dilemma: either utilize high-performance models for all tasks or resort to manual routing on the application side. Notably, about 65% of prompts in production environments involve standard operations like extraction, classification, formatting, and summarization, which do not require frontier model high performance.

OrcaRouter tackles this issue by assessing the difficulty of each prompt, directing complex reasoning tasks to frontier models while automatically routing standard operations to cost-effective open models. This efficient setup ensures quality while reducing LLM expenditure by about 40%. With the introduction of Qwen 3.7 Max, OrcaRouter can now handle enterprise AI agent workflows that necessitate long-context processing with one million tokens and sustained reasoning activities.

Overview of OrcaRouter and Qwen 3.7 Max

Key Features of Qwen 3.7 Max:

- One Million Tokens Context Support: Capable of processing lengthy documents, multiple files, and conversation histories simultaneously.
- Sustained Long-term Reasoning: Maintains optimization strategies over 1,000 tool calls without losing context.
- Robust Code Generation and Agent Performance: Achieved scores of 69.7 on the Terminal-Bench 2.0 and 78.3 on SWE-Multilingual.

Applications of OrcaRouter:

- Automating complex business workflows (customer support, sales assistance, back office tasks).
- Analyzing, summarizing, and classifying lengthy documents.
- Executing multi-step agent tasks (status maintenance, retries, and failover handling).
- Engaging in advanced code generation and debugging tasks.

Advantages of Integration

- Cost Savings While Maintaining Quality: By processing standard operations (about 65%) with Qwen 3.7 Max and allocating only complex reasoning tasks to frontier models (like Claude Opus 4.7 and GPT 5.5), teams can realize approximately $47,700 in annual savings, based on a $10,000 monthly team size.
- Zero Token Surcharge Transparency: Visibility into each request's difficulty, selected model, provider, and public pricing is maintained, with assessment rationale available through response headers.
- Seamless Integration: The API is compatible with OpenAI, allowing for straightforward integration into existing workflows, including Cursor, Cline, and LangChain.
- Midstream Switching Capability: In case of provider failures, the system automatically switches to a secondary model during streaming, preserving agent loop states and ensuring zero user-facing errors.
- Comprehensive Integration Safeguards: Implements PII Shield, secrets detection, prompt injection prevention, Brand Safety, and Compliance Logging all through one gateway.

Key Statistics of OrcaRouter

Feature	Value
-------	-----
Routing Fee	0%
Cost Reduction While Maintaining Quality	~40%
Number of Supported Models	200+
Routing Delay	<1ms
SLA Uptime (Enterprise)	99.99%

Statement from the CEO

Yoichi Hosoi, CEO of FlashLabs, emphasized: “With the support for one million tokens in Qwen 3.7 Max, OrcaRouter sets a new standard for enterprise AI agent workflows. It is the only platform that can optimize costs without compromising quality, especially for automation involving complex reasoning and extensive document processing. Under the Human-AI Hybrid philosophy, we will further accelerate sales and CX automation for enterprises.”

Company Overview

FlashLabs, based in Chiyoda, Tokyo, is focused on automating sales and customer experience operations, ultimately driving towards autonomy through AI applications. By integrating mechanical speed and precision with human strategic insights in a Human-AI Hybrid framework, it provides results that surpass conventional methods.