ZenMux's blog index (zenmux.ai/blog) surfaces several developer-relevant updates from June–July 2026: the Ling-3.0-flash API became available on the gateway (Jul 23, 2026) with 256K context positioning it for agentic coding and tool workflows; a guide details pointing the Claude Agent SDK at ZenMux to route phases of a The same blog also documents ZenMux's first-party research: a "Who are you?" study in which the team spent roughly $1,000 querying 27 frontier LLMs from 16 vendors 29,700 times and found ~7.1% misidentification across brands, with full data and tooling open-sourced (Jun 4, 2026), plus a follow-up piece arguing that sel

ZenMux announced the availability of GLM 5.3 through its enterprise LLM platform on August 21, 2026, accompanied by a free trial program aimed at letting developers and technical teams evaluate the model before making longer-term routing decisions. The model is offered via ZenMux's unified API environment, which suppor The trial is positioned around a common evaluation challenge: comparing multiple models across providers, pricing, and API formats requires separate accounts and varying integration work. ZenMux's platform provides a single access layer for model discovery, routing, and API operations, and the free trial is intended to

ZenMux announced limited-time free access to DeepSeek V4 Flash on August 13, 2026, through a dedicated model access page that consolidates provider routing, pricing, latency, throughput, uptime, supported APIs, and free-tier usage limits in one location. The page also provides a direct model slug for implementation, al DeepSeek V4 Flash is positioned as an efficiency-focused member of the DeepSeek V4 family, targeting workloads where speed and cost predictability matter alongside reasoning depth. ZenMux lists support for a one-million-token context window, reasoning capabilities, and OpenAI-style chat completion compatibility, making

The DeepSeek V4 Flash (free) model page on ZenMux (published 3 days before current date) documents a new free routing tier for the 0731 version of DeepSeek V4 Flash, with 1M context, an explicit per-provider table (DeepSeek at 96.0% cache hit rate, BaiduSunset as alternate), and live operational metrics: $0.14/M input, Technically, the page describes V4 Flash as the efficiency-oriented variant of the V4 series, open-sourced alongside V4-Pro, retaining reasoning capability close to V4-Pro on simple agent tasks with a measurable gap only on the most demanding workflows, and inheriting DeepSeek Sparse Attention (DSA) plus token-dimensio

Agnes AI, a Singapore-based company that has built and trained its own multimodal AI models in-house, has made its proprietary model suite available through ZenMux via API token access. The integration positions ZenMux as a distribution channel for Agnes AI's reasoning, generation, and execution capabilities, aiming to Alongside the ZenMux API availability, Agnes AI separately launched Agnes Token Plans through its official website, offering low-cost token packages aimed at making the models more accessible for real-world deployment. The combined moves reflect a broader strategy of bundling multimodal capabilities into a single integ

ZenMux published a head-to-head technical comparison of Qwen3.8-Max and Kimi K3 on August 13, 2026, aimed at developers choosing between the two models. Kimi K3 has 2.8 trillion total parameters with 104 billion active, while Qwen3.8-Max has 2.4 trillion total and 95 billion active; both support approximately 1 million On ZenMux pricing, Qwen3.8-Max charges $2 per 1M input tokens versus $3.00 for Kimi K3, with output rates of $6 versus $15.00 respectively, though task-level cost depends on usage patterns and Kimi K3's cache-hit pricing may benefit workloads that reuse context. Both target coding and long-horizon agentic workflows, an

The ZenMux platform has added the Claude Opus 4.6 model to its free plan and offers a two-week free trial, attracting attention in the AI community. The model f