Together built its own agent evals, lifting agent task success from 67% to 96%
togethercompute · x · 2026-11-17
Together Compute shared how they made the Together platform easier for AI agents to use. In a world where agent experience (AX) matters more than developer experience (DX), they used evals to verify agents can actually use the platform—improving agent task success rates from 67% to 96%. The article walks through how the evals were built and what changed.
More from coding & agent
- agent-browser: open-source browser automation so agents can verify fixes and file PRs — evilrabbit_ · 2026-11-17
- OpenRouter compares the context-token cost of MCP, CLI and Skills — rseroter · 2026-11-17
- PSN: storing agent skills as executable programs in an evolvable network — burkov · 2026-11-17
- OunceMCP 1.0 RC: Token Budgets and Safe Writes for MCP Servers, 58-85% Fewer Tokens — gentleGiant96 · 2026-11-17
- Solo Dev Shares 6 Hard-Won Lessons on Prompting Video With a Real Person's Face — pvdyck · 2026-11-17
- Who Can Even Read AI-Generated Code? Garbled Screenshot Sparks Debate — gethackteam · 2026-11-17