paper §7.3: per-role model leaderboard + GLM-5.2 flagship tier

Single-role drafter on a fixed concision-weighted brief, one turn per backend,
judge-ranked: GLM-4.7 > Claude > Kimi > GLM-5.2. Notable inversion — the
flagship GLM-5.2 ranked last by over-writing the 25-word brief, underscoring
that the best model is role/task/rubric-dependent. Adds the worker_glm5
(glm-5.2 Opus-class) tier to the example config. Gemini excluded this run
(Google June-19 API-key-restriction 403; infra, not model).

Co-Authored-By: Claude Opus 4.8 <[email protected]>

deploy/clawmates-runtime/agent.config.example.toml

@@ -76,6 +76,14 @@ mcp_config = "/zeroclaw-data/clawmates-mcp.json"
 # [agents.worker_glm]
 # model_provider = "claude_cli.glm"
 # risk_profile = "toolfree"
+# Flagship tier (Opus-class): a second alias on glm-5.2 (same z.ai env). Costs
+# 2-3x quota vs glm-4.7 — reserve for hard reasoning roles, not routine ones.
+# [providers.models.claude_cli.glm5]
+# model = "glm-5.2"
+# env = { HOME = "/zeroclaw-data/glm-home", CLAUDE_CODE_OAUTH_TOKEN = "", ANTHROPIC_BASE_URL = "https://api.z.ai/api/anthropic", ANTHROPIC_AUTH_TOKEN = "$ZAI_API_KEY", API_TIMEOUT_MS = "3000000" }
+# [agents.worker_glm5]
+# model_provider = "claude_cli.glm5"
+# risk_profile = "toolfree"
 #
 # Gemini (Google) — built-in `gemini` API family (no image rebuild). Key via env
 # ZEROCLAW_providers__models__gemini__default__api_key=<GEMINI_API_KEY>; model in

papers/dynamic-agentic-topologies.md

@@ -206,6 +206,30 @@ do, the Kimi CLI's stream path does not.) A larger model×role×topology sweep i
 future work (§8), bounded by the subscription plans' 5‑hour/weekly quotas and
 concurrency caps, which favor sequential pipelines over wide swarm/mesh fan‑out.
 
+### 7.3 Per‑role model leaderboard (single‑role, fixed task)
+
+To isolate the model axis from topology, we ran a single‑role drafter on a fixed
+brief — *"a single‑sentence nonprofit mission statement, under 25 words"* — once per
+backend (one turn each; per‑node model via `attrs["agent"]`), then ranked the
+outputs with an independent Claude judge on clarity, impact, and concision:
+
+| Rank | Model (backend) | Note |
+|:----:|-----------------|------|
+| 1 | **GLM‑4.7** (Zhipu, Sonnet‑class) | tightest, most on‑brief |
+| 2 | Claude Sonnet‑4.6 (Anthropic) | strong but slightly longer |
+| 3 | Kimi K2 (Moonshot) | clear, a touch generic |
+| 4 | GLM‑5.2 (Zhipu, Opus‑class) | most capable but *over‑wrote* the 25‑word brief |
+
+A notable inversion: the **flagship GLM‑5.2 ranked last** on this *concision‑weighted*
+task — its longer, richer output is an asset for complex reasoning but a liability
+when the rubric rewards brevity. This is the core lesson the platform is built to
+surface: there is no globally "best" model — the winner is role‑, task‑, and
+rubric‑dependent, so model choice belongs to the same tunable layer as topology,
+measured rather than assumed. (Gemini‑2.5‑flash was excluded from this run: the
+Google key hit a 403 from Google's June‑19 API‑key‑restriction enforcement preview,
+an infrastructure constraint unrelated to the model — it succeeds when the key is
+restricted.)
+
 ## 8. Limitations and future work
 
 - Results are offline; the production path (real models + tool‑using turns with