Last updated: October 19, 2026
Release history. Earlier limits may have been replaced by later updates.
October 19, 2026
Windows x64 preview updates the Harness runtime to 0.2.1-alpha.3, adds localized skills and downloads Python dependencies when needed. IteraWork uses its own application data and browser sign-in return link. Object expression settings remain editable. Download and install this unsigned preview manually.
October 16, 2026
Qwen3.8-Flash, GLM-5.2, GLM-5.1, GLM-5, Kimi-K2.6, Kimi-K2.5, MiniMax-M2.7 and MiniMax-M2.5 are now available through the OpenAI-compatible API. The website and Platform include model details, capabilities and current input, cached-input and output pricing in USD per 1M tokens.
October 12, 2026
Windows x64 preview (build 20261012.1) supports the free Qwen3.8-27B 128K context window and compacts long conversations before the window fills. Free now has 160K TPM, Starter 300K TPM, and Pro keeps 300M monthly tokens and 600K TPM. Context and quota failures show clearer messages. All downloads use our Hong Kong server. Download and install this unsigned preview manually.
October 12, 2026
Qwen3.8-27B Free now includes 128K total context and 160K TPM. Starter also provides 128K context with 300K TPM. Monthly allowances remain 30M / 100M / 300M tokens for Free / Starter / Pro; Pro keeps 256K context and 600K TPM. The gateway returns clear context and quota errors, and the new Iterawork preview compacts long conversations before the window fills.
October 10, 2026
Windows x64 preview (build 20261010.1) opens an in-app login window when the system browser cannot open. Your selected Chinese or English language follows you between the website, Platform and app login. The model picker refreshes the online catalog, including the free Qwen3.8 27B model. Download and install this unsigned preview manually.
October 9, 2026
Qwen3.8-27B is now available through the OpenAI-compatible API with 32K free context. Input, cached input and output are all $0 per 1M tokens. The model catalog, pricing and integration guides now reflect its availability. Fair-use rate limits apply; no SLA on free capacity.
Historical limit: the 32K Free context was superseded by 128K on October 12, 2026. Read the update →
October 8, 2026
The Windows x64 preview v0.2.0-rc.2 (build 20261008.1) is now available to download. Choose coding or work mode inside one app to write code, handle office files and schedule recurring tasks.
October 3, 2026
Clarified fair-use rate limits, abuse suspension and multi-account rules for free capacity. See the Free Tier Terms page for details.
October 1, 2026
Revised the Terms of Service covering self-service and enterprise use, including references to the Acceptable Use Policy and Free Tier Terms.
September 28, 2026
Qwen3.8-27B is available again with 32K free context through the OpenAI-compatible API. Fair-use rate limits apply; no SLA on free capacity.
Historical limit: the 32K Free context was superseded by 128K on October 12, 2026. Read the update →
