Points roles.executor in nisse-dev.example.yaml at qwen3.8-27b, and asserts it in the example-file test.
qwen3.8-27b led the executor benchmark (#55): 37/40 over eight tasks, 17/20 on the harder four, ahead of claude-haiku-4-5 (15/20). The runner-up MoE models were faster but weaker (32/40, 30/40). local-fast, the previous example, scored 0/8 on the harder four. Its cost is time on hard tasks (about 5 to 8.5 minutes a run), which suits a milestone-driven executor.
Selection is on the dev executor alone for now, to get one agent working end to end first. The spec's open item on sharing the local host (Part H) is unchanged.
Points `roles.executor` in `nisse-dev.example.yaml` at `qwen3.8-27b`, and asserts it in the example-file test.
qwen3.8-27b led the executor benchmark (#55): 37/40 over eight tasks, 17/20 on the harder four, ahead of claude-haiku-4-5 (15/20). The runner-up MoE models were faster but weaker (32/40, 30/40). `local-fast`, the previous example, scored 0/8 on the harder four. Its cost is time on hard tasks (about 5 to 8.5 minutes a run), which suits a milestone-driven executor.
Selection is on the dev executor alone for now, to get one agent working end to end first. The spec's open item on sharing the local host (Part H) is unchanged.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
It led the executor benchmark (PR #55): 37/40 over eight tasks and 17/20
on the harder four, ahead of claude-haiku-4-5's 15/20. local-fast, the
previous example, scored 0/8 on the harder four. Model selection is on
the dev executor alone for now; the shared-host question (spec Part H)
stays open.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
Points
roles.executorinnisse-dev.example.yamlatqwen3.8-27b, and asserts it in the example-file test.qwen3.8-27b led the executor benchmark (#55): 37/40 over eight tasks, 17/20 on the harder four, ahead of claude-haiku-4-5 (15/20). The runner-up MoE models were faster but weaker (32/40, 30/40).
local-fast, the previous example, scored 0/8 on the harder four. Its cost is time on hard tasks (about 5 to 8.5 minutes a run), which suits a milestone-driven executor.Selection is on the dev executor alone for now, to get one agent working end to end first. The spec's open item on sharing the local host (Part H) is unchanged.
🤖 Generated with Claude Code