April 17, 2026
Taskade developer changelog — April 17, 2026: New LLM Models Added to the Picker, EVE Custom Bash Commands.
Last updated
Was this helpful?
Taskade developer changelog — April 17, 2026: New LLM Models Added to the Picker, EVE Custom Bash Commands.
Developer highlights. This section tracks API, MCP, SDK, automation, and Taskade Genesis platform changes for developers. For the full product changelog see taskade.com/changelog; for release newsletters see taskade.com/blog/updates.
The following models launched in preview (non-production) on April 17 and are now available in the Taskade model picker for agent chats and automation AI actions:
These models launched in preview (non-production) on April 17; Qwen 3.6 Plus, MiniMax M2.7, and GLM 5.1 reached general availability on May 5, 2026.
Gemma 4 31B
Google (via Vercel AI Gateway)
Small
Instruction-tuned
Gemma 4 26B A4B
Google (via Vercel AI Gateway)
Small
Mixture-of-experts variant
Qwen 3.6 Plus
Alibaba (via Vercel AI Gateway)
Advanced
Replaces Qwen 3 in the picker
MiniMax M2.7
MiniMax (via Vercel AI Gateway)
Advanced
Coding, security, and debugging
GLM 5.1
Z.AI (via Vercel AI Gateway)
Advanced
Auto-mode routing updates:
XL and 2XL plans (Frontier tier) now auto-select Claude Opus 4.6 by default; Claude Opus 4.7 remains available as a manual selection.
Large-context (Business/lg) plans auto-route to Claude Sonnet 4.6 in auto mode.
GPT-5.4 mini and GPT-5.4 nano are now available for standard and lightweight tiers respectively.
Gemini 3.1 Flash Lite replaces Gemini 2.5 Flash Lite in the Small tier.
Credit pricing for all models follows the blended $/1M-token formula. Use the credit display in the model picker to estimate cost before running automations.
EVE's bash sandbox now ships with a set of built-in custom commands that can be invoked directly from the bash tool:
agentChat
Send a message to an agent chat thread from within a sandbox session
teamChat
Post a message to a workspace team chat channel
discoverApis
Enumerate available Taskade API endpoints for a given resource area
fetchUrl
Perform an outbound HTTP GET and return the response body
httpRequest
Full HTTP request (any method, headers, body) from within the sandbox
webSearch
Run a web search and return ranked results
These commands are injected at sandbox creation time and are callable via the bash tool in any EVE or space agent session.
The Genesis app build pipeline now writes a build-status.json file to the integration output directory after every build. This file records the build outcome and can be read by downstream CI/CD integrations to determine pass/fail without parsing log output.
See the Genesis builder docs.
Last updated
Was this helpful?
Was this helpful?