evals: switch the default LLM judge to claude-haiku-4-5

Replaces gpt-5-mini-2025-08-07 as the default judge everywhere: the 147
eval configs plus the hardcoded defaults in the evals CLI, recipe composer,
comparison config generator, taubench user-simulator, trial runner and
OptimizeConfig.

Haiku is roughly 3x faster on the judging path and we are out of OpenAI
quota, so gpt-5-mini was failing closed on long eval runs.

src/openjarvis/core/config.py

@@ -1408,7 +1408,7 @@ class OptimizeConfig:
     optimizer_provider: str = "anthropic"
     benchmark: str = ""
     max_samples: int = 50
-    judge_model: str = "gpt-5-mini-2025-08-07"
+    judge_model: str = "claude-haiku-4-5-20251001"
     db_path: str = str(DEFAULT_CONFIG_DIR / "optimize.db")
 
 

src/openjarvis/evals/cli.py

@@ -1141,7 +1141,7 @@ def main():
     "-n", "--max-samples", type=int, default=None, help="Maximum samples to evaluate"
 )
 @click.option("-w", "--max-workers", type=int, default=4, help="Parallel workers")
-@click.option("--judge-model", default="gpt-5-mini-2025-08-07", help="LLM judge model")
+@click.option("--judge-model", default="claude-haiku-4-5-20251001", help="LLM judge model")
 @click.option("-o", "--output", "output_path", default=None, help="Output JSONL path")
 @click.option("--seed", type=int, default=42, help="Random seed")
 @click.option("--split", "dataset_split", default=None, help="Dataset split override")
@@ -1387,7 +1387,7 @@ def run(
     "-n", "--max-samples", type=int, default=None, help="Max samples per benchmark"
 )
 @click.option("-w", "--max-workers", type=int, default=4, help="Parallel workers")
-@click.option("--judge-model", default="gpt-5-mini-2025-08-07", help="LLM judge model")
+@click.option("--judge-model", default="claude-haiku-4-5-20251001", help="LLM judge model")
 @click.option("--output-dir", default="results/", help="Output directory for results")
 @click.option("--seed", type=int, default=42, help="Random seed")
 @click.option("-v", "--verbose", is_flag=True, help="Verbose logging")

src/openjarvis/evals/comparison/make_configs.py

@@ -98,7 +98,7 @@ BENCHMARKS: Dict[str, Dict[str, object]] = {
     },
 }
 
-DEFAULT_JUDGE_MODEL = "gpt-5-mini-2025-08-07"
+DEFAULT_JUDGE_MODEL = "claude-haiku-4-5-20251001"
 
 
 def _template_path() -> Path:

src/openjarvis/evals/configs/deepplanning-monitor-operative.toml

@@ -11,7 +11,7 @@ temperature = 0.6
 max_tokens = 32768
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 max_tokens = 1024
 

src/openjarvis/evals/configs/gaia-gemini-31-pro.toml

@@ -8,7 +8,7 @@ temperature = 0.6
 max_tokens = 8192
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 max_tokens = 4096
 

src/openjarvis/evals/configs/gaia-gemma4-26b.toml

@@ -8,7 +8,7 @@ temperature = 0.6
 max_tokens = 8192
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 max_tokens = 4096
 engine = "cloud"

src/openjarvis/evals/configs/gaia-gpt54.toml

@@ -8,7 +8,7 @@ temperature = 0.6
 max_tokens = 8192
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 max_tokens = 4096
 

src/openjarvis/evals/configs/gaia-nemotron-nano.toml

@@ -8,7 +8,7 @@ temperature = 0.6
 max_tokens = 16384
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 max_tokens = 4096
 

src/openjarvis/evals/configs/gaia-nemotron-super.toml

@@ -8,7 +8,7 @@ temperature = 0.6
 max_tokens = 8192
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 engine = "cloud"
 

src/openjarvis/evals/configs/gaia-nemotron3-super.toml

@@ -8,7 +8,7 @@ temperature = 1.0
 max_tokens = 8192
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 max_tokens = 4096
 engine = "cloud"

src/openjarvis/evals/configs/gaia-qwen-122b.toml

@@ -8,7 +8,7 @@ temperature = 0.6
 max_tokens = 8192
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 engine = "cloud"
 

src/openjarvis/evals/configs/gaia-qwen-27b.toml

@@ -8,7 +8,7 @@ temperature = 0.6
 max_tokens = 16384
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 max_tokens = 4096
 

src/openjarvis/evals/configs/gaia-qwen-2b.toml

@@ -8,7 +8,7 @@ temperature = 0.6
 max_tokens = 8192
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 max_tokens = 4096
 engine = "cloud"

src/openjarvis/evals/configs/gaia-qwen-35b.toml

@@ -8,7 +8,7 @@ temperature = 0.6
 max_tokens = 8192
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 engine = "cloud"
 

src/openjarvis/evals/configs/gaia-qwen-9b.toml

@@ -8,7 +8,7 @@ temperature = 0.6
 max_tokens = 8192
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 max_tokens = 4096
 engine = "cloud"

src/openjarvis/evals/configs/gaia-qwen122b.toml

@@ -8,7 +8,7 @@ temperature = 0.6
 max_tokens = 8192
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 max_tokens = 4096
 

src/openjarvis/evals/configs/gaia-qwen35b.toml

@@ -8,7 +8,7 @@ temperature = 0.6
 max_tokens = 8192
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 max_tokens = 4096
 

src/openjarvis/evals/configs/gaia-rerun-claude-opus.toml

@@ -8,7 +8,7 @@ temperature = 0.6
 max_tokens = 8192
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 max_tokens = 4096
 

src/openjarvis/evals/configs/gaia-rerun-claude-sonnet.toml

@@ -8,7 +8,7 @@ temperature = 0.6
 max_tokens = 8192
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 max_tokens = 4096
 

src/openjarvis/evals/configs/gaia-rerun-glm47flash.toml

@@ -8,7 +8,7 @@ temperature = 0.6
 max_tokens = 8192
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 max_tokens = 4096
 

src/openjarvis/evals/configs/gaia-rerun-gpt5mini.toml

@@ -8,7 +8,7 @@ temperature = 0.6
 max_tokens = 8192
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 max_tokens = 4096
 

src/openjarvis/evals/configs/gaia-rerun-minimax-m25.toml

@@ -8,7 +8,7 @@ temperature = 0.6
 max_tokens = 8192
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 max_tokens = 4096
 

src/openjarvis/evals/configs/gaia-rerun-qwen122b.toml

@@ -8,7 +8,7 @@ temperature = 0.6
 max_tokens = 8192
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 max_tokens = 4096
 

src/openjarvis/evals/configs/gaia-rerun-qwen35b.toml

@@ -8,7 +8,7 @@ temperature = 0.6
 max_tokens = 8192
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 max_tokens = 4096
 

src/openjarvis/evals/configs/gaia-rerun-qwen397b.toml

@@ -8,7 +8,7 @@ temperature = 0.6
 max_tokens = 8192
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 max_tokens = 4096
 

src/openjarvis/evals/configs/gaia-trinity-large.toml

@@ -8,7 +8,7 @@ temperature = 0.6
 max_tokens = 16384
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 max_tokens = 4096
 

src/openjarvis/evals/configs/glm-4.7-flash-openhands-remaining.toml

@@ -10,7 +10,7 @@ temperature = 0.0
 max_tokens = 2048
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 max_tokens = 2048
 

src/openjarvis/evals/configs/glm-4.7-flash-openhands.toml

@@ -10,7 +10,7 @@ temperature = 0.0
 max_tokens = 2048
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 max_tokens = 1024
 

src/openjarvis/evals/configs/glm-4.7-fp8-openhands-gaia.toml

@@ -10,7 +10,7 @@ temperature = 0.0
 max_tokens = 2048
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 max_tokens = 1024
 

src/openjarvis/evals/configs/glm5-fp8-monitor-operative.toml

@@ -12,7 +12,7 @@ temperature = 0.6
 max_tokens = 32768
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 max_tokens = 4096
 

src/openjarvis/evals/configs/kimi-k2.5-monitor-operative.toml

@@ -12,7 +12,7 @@ temperature = 1.0              # Kimi-K2.5 thinking mode recommended
 max_tokens = 32768
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 max_tokens = 4096
 

src/openjarvis/evals/configs/livecodebench-claude-opus.toml

@@ -11,7 +11,7 @@ temperature = 0.0
 max_tokens = 4096
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 max_tokens = 1024
 

src/openjarvis/evals/configs/livecodebench-gemma4-26b.toml

@@ -10,7 +10,7 @@ temperature = 0.0
 max_tokens = 4096
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 engine = "cloud"
 

src/openjarvis/evals/configs/livecodebench-gemma4-e4b.toml

@@ -10,7 +10,7 @@ temperature = 0.0
 max_tokens = 4096
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 engine = "cloud"
 

src/openjarvis/evals/configs/livecodebench-kimi-k25.toml

@@ -17,7 +17,7 @@ temperature = 0.0
 max_tokens = 4096
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 engine = "cloud"
 

src/openjarvis/evals/configs/livecodebench-minimax-m25.toml

@@ -14,7 +14,7 @@ temperature = 0.0
 max_tokens = 4096
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 engine = "cloud"
 

src/openjarvis/evals/configs/livecodebench-nemotron-nano.toml

@@ -10,7 +10,7 @@ temperature = 0.0
 max_tokens = 4096
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 engine = "cloud"
 

src/openjarvis/evals/configs/livecodebench-nemotron-super.toml

@@ -8,7 +8,7 @@ temperature = 0.0
 max_tokens = 4096
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 engine = "cloud"
 

src/openjarvis/evals/configs/livecodebench-qwen-122b.toml

@@ -8,7 +8,7 @@ temperature = 0.0
 max_tokens = 4096
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 engine = "cloud"
 

src/openjarvis/evals/configs/livecodebench-qwen-27b.toml

@@ -10,7 +10,7 @@ temperature = 0.0
 max_tokens = 4096
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 engine = "cloud"
 

src/openjarvis/evals/configs/livecodebench-qwen-2b.toml

@@ -10,7 +10,7 @@ temperature = 0.0
 max_tokens = 4096
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 engine = "cloud"
 

src/openjarvis/evals/configs/livecodebench-qwen-35b.toml

@@ -8,7 +8,7 @@ temperature = 0.0
 max_tokens = 4096
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 engine = "cloud"
 

src/openjarvis/evals/configs/livecodebench-qwen-9b.toml

@@ -10,7 +10,7 @@ temperature = 0.0
 max_tokens = 4096
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 engine = "cloud"
 

src/openjarvis/evals/configs/liveresearch-claude-opus.toml

@@ -8,7 +8,7 @@ temperature = 0.6
 max_tokens = 16384
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 max_tokens = 4096
 

src/openjarvis/evals/configs/liveresearch-gemini-31-pro.toml

@@ -8,7 +8,7 @@ temperature = 0.6
 max_tokens = 16384
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 engine = "cloud"
 

src/openjarvis/evals/configs/liveresearch-gemma4-26b.toml

@@ -8,7 +8,7 @@ temperature = 0.6
 max_tokens = 16384
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 max_tokens = 4096
 engine = "cloud"

src/openjarvis/evals/configs/liveresearch-gemma4-31b.toml

@@ -8,7 +8,7 @@ temperature = 0.6
 max_tokens = 16384
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 engine = "cloud"
 

src/openjarvis/evals/configs/liveresearch-gpt54.toml

@@ -8,7 +8,7 @@ temperature = 0.6
 max_tokens = 16384
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 engine = "cloud"
 

src/openjarvis/evals/configs/liveresearch-granite-3.3-8b-lora.toml

@@ -8,7 +8,7 @@ temperature = 0.6
 max_tokens = 16384
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 engine = "cloud"
 

src/openjarvis/evals/configs/liveresearch-granite-3.3-8b.toml

@@ -8,7 +8,7 @@ temperature = 0.6
 max_tokens = 16384
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 engine = "cloud"
 

src/openjarvis/evals/configs/liveresearch-granite-4.0-h-small.toml

@@ -8,7 +8,7 @@ temperature = 0.6
 max_tokens = 16384
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 engine = "cloud"
 

src/openjarvis/evals/configs/liveresearch-granite-4.0-micro.toml

@@ -8,7 +8,7 @@ temperature = 0.6
 max_tokens = 16384
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 engine = "cloud"
 

src/openjarvis/evals/configs/liveresearch-nemotron-nano.toml

@@ -8,7 +8,7 @@ temperature = 0.6
 max_tokens = 16384
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 max_tokens = 4096
 

src/openjarvis/evals/configs/liveresearch-nemotron-super.toml

@@ -8,7 +8,7 @@ temperature = 0.6
 max_tokens = 16384
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 engine = "cloud"
 

src/openjarvis/evals/configs/liveresearch-qwen-122b.toml

@@ -8,7 +8,7 @@ temperature = 0.6
 max_tokens = 16384
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 engine = "cloud"
 

src/openjarvis/evals/configs/liveresearch-qwen-27b-lora-r32.toml

@@ -8,7 +8,7 @@ temperature = 0.6
 max_tokens = 16384
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 max_tokens = 4096
 

src/openjarvis/evals/configs/liveresearch-qwen-27b.toml

@@ -8,7 +8,7 @@ temperature = 0.6
 max_tokens = 16384
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 max_tokens = 4096
 

src/openjarvis/evals/configs/liveresearch-qwen-2b-lora-r16.toml

@@ -8,7 +8,7 @@ temperature = 0.6
 max_tokens = 16384
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 max_tokens = 4096
 engine = "cloud"

src/openjarvis/evals/configs/liveresearch-qwen-2b-lora-r32.toml

@@ -8,7 +8,7 @@ temperature = 0.6
 max_tokens = 16384
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 max_tokens = 4096
 engine = "cloud"

src/openjarvis/evals/configs/liveresearch-qwen-2b-mixed.toml

@@ -8,7 +8,7 @@ temperature = 0.6
 max_tokens = 16384
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 max_tokens = 4096
 engine = "cloud"

src/openjarvis/evals/configs/liveresearch-qwen-2b-trackb.toml

@@ -8,7 +8,7 @@ temperature = 0.6
 max_tokens = 16384
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 max_tokens = 4096
 engine = "cloud"

src/openjarvis/evals/configs/liveresearch-qwen-2b.toml

@@ -8,7 +8,7 @@ temperature = 0.6
 max_tokens = 16384
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 max_tokens = 4096
 engine = "cloud"

src/openjarvis/evals/configs/liveresearch-qwen-35b.toml

@@ -8,7 +8,7 @@ temperature = 0.6
 max_tokens = 16384
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 engine = "cloud"
 

src/openjarvis/evals/configs/liveresearch-qwen-9b-lora-r32.toml

@@ -8,7 +8,7 @@ temperature = 0.6
 max_tokens = 16384
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 max_tokens = 4096
 engine = "cloud"

src/openjarvis/evals/configs/liveresearch-qwen-9b-mixed.toml

@@ -8,7 +8,7 @@ temperature = 0.6
 max_tokens = 16384
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 max_tokens = 4096
 engine = "cloud"

src/openjarvis/evals/configs/liveresearch-qwen-9b-trackb.toml

@@ -8,7 +8,7 @@ temperature = 0.6
 max_tokens = 16384
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 max_tokens = 4096
 engine = "cloud"

src/openjarvis/evals/configs/liveresearch-qwen-9b.toml

@@ -8,7 +8,7 @@ temperature = 0.6
 max_tokens = 16384
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 max_tokens = 4096
 engine = "cloud"

src/openjarvis/evals/configs/liveresearch-trinity-large.toml

@@ -8,7 +8,7 @@ temperature = 0.6
 max_tokens = 16384
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 max_tokens = 4096
 

src/openjarvis/evals/configs/liveresearchbench-qwen-27b.toml

@@ -10,7 +10,7 @@ temperature = 0.6
 max_tokens = 16384
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 max_tokens = 4096
 engine = "cloud"

src/openjarvis/evals/configs/liveresearchbench-qwen-2b.toml

@@ -10,7 +10,7 @@ temperature = 0.6
 max_tokens = 16384
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 max_tokens = 4096
 engine = "cloud"

src/openjarvis/evals/configs/liveresearchbench-qwen-9b.toml

@@ -10,7 +10,7 @@ temperature = 0.6
 max_tokens = 16384
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 max_tokens = 4096
 engine = "cloud"

src/openjarvis/evals/configs/long-horizon-claude-opus.toml

@@ -10,7 +10,7 @@ temperature = 0.6
 max_tokens = 8192
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 max_tokens = 4096
 

src/openjarvis/evals/configs/long-horizon-claude-sonnet.toml

@@ -10,7 +10,7 @@ temperature = 0.6
 max_tokens = 8192
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 max_tokens = 4096
 

src/openjarvis/evals/configs/long-horizon-glm47flash.toml

@@ -10,7 +10,7 @@ temperature = 0.6
 max_tokens = 8192
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 max_tokens = 4096
 

src/openjarvis/evals/configs/long-horizon-gpt5mini.toml

@@ -10,7 +10,7 @@ temperature = 0.6
 max_tokens = 8192
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 max_tokens = 4096
 

src/openjarvis/evals/configs/long-horizon-kimik25.toml

@@ -12,7 +12,7 @@ temperature = 0.6
 max_tokens = 8192
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 max_tokens = 4096
 

src/openjarvis/evals/configs/long-horizon-minimax-m25.toml

@@ -10,7 +10,7 @@ temperature = 0.6
 max_tokens = 8192
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 max_tokens = 4096
 

src/openjarvis/evals/configs/long-horizon-qwen122b.toml

@@ -10,7 +10,7 @@ temperature = 0.6
 max_tokens = 8192
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 max_tokens = 4096
 

src/openjarvis/evals/configs/long-horizon-qwen35b.toml

@@ -10,7 +10,7 @@ temperature = 0.6
 max_tokens = 8192
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 max_tokens = 4096
 

src/openjarvis/evals/configs/long-horizon-qwen397b.toml

@@ -10,7 +10,7 @@ temperature = 0.6
 max_tokens = 8192
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 max_tokens = 4096
 

src/openjarvis/evals/configs/paperarena-monitor-operative.toml

@@ -11,7 +11,7 @@ temperature = 0.6
 max_tokens = 32768
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 max_tokens = 1024
 

src/openjarvis/evals/configs/pinchbench-nemotron-super.toml

@@ -8,7 +8,7 @@ temperature = 0.6
 max_tokens = 8192
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 engine = "cloud"
 

src/openjarvis/evals/configs/pinchbench-qwen-122b.toml

@@ -8,7 +8,7 @@ temperature = 0.6
 max_tokens = 8192
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 engine = "cloud"
 

src/openjarvis/evals/configs/pinchbench-qwen-35b.toml

@@ -8,7 +8,7 @@ temperature = 0.6
 max_tokens = 8192
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 engine = "cloud"
 

src/openjarvis/evals/configs/qwen3.5-122b-kg-agentic.toml

@@ -11,7 +11,7 @@ temperature = 0.6
 max_tokens = 8192
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 max_tokens = 4096
 

src/openjarvis/evals/configs/qwen3.5-122b-kg-hightoken.toml

@@ -11,7 +11,7 @@ temperature = 0.6
 max_tokens = 8192          # 8x increase from baseline 1024
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 max_tokens = 4096
 

src/openjarvis/evals/configs/qwen3.5-122b-reasoning-suite.toml

@@ -10,7 +10,7 @@ temperature = 0.6          # Qwen3.5 recommended
 max_tokens = 4096
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 max_tokens = 4096
 

src/openjarvis/evals/configs/qwen3.5-35b-kg.toml

@@ -11,7 +11,7 @@ temperature = 0.6
 max_tokens = 8192
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 max_tokens = 4096
 

src/openjarvis/evals/configs/qwen3.5-35b-monitor-operative.toml

@@ -14,7 +14,7 @@ temperature = 0.6          # Qwen3.5 thinking mode (precise analysis)
 max_tokens = 32768         # Qwen3.5 recommended for general queries
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 max_tokens = 1024
 

src/openjarvis/evals/configs/qwen3.5-397b-monitor-operative.toml

@@ -14,7 +14,7 @@ temperature = 0.6          # Qwen3.5 thinking mode recommended
 max_tokens = 1024          # Keep short to prevent hangs; answers are brief
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 max_tokens = 4096
 

src/openjarvis/evals/configs/smoke_amd.toml

@@ -14,7 +14,7 @@ temperature = 0.0
 max_tokens = 2048
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 max_tokens = 512
 

src/openjarvis/evals/configs/smoke_apple.toml

@@ -13,7 +13,7 @@ temperature = 0.0
 max_tokens = 2048
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 max_tokens = 512
 

src/openjarvis/evals/configs/smoke_nvidia.toml

@@ -13,7 +13,7 @@ temperature = 0.0
 max_tokens = 2048
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 max_tokens = 512
 

src/openjarvis/evals/configs/taubench-claude-opus.toml

@@ -10,7 +10,7 @@ temperature = 0.7
 max_tokens = 4096
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 engine = "cloud"
 

src/openjarvis/evals/configs/taubench-gemini-31-pro.toml

@@ -10,7 +10,7 @@ temperature = 0.7
 max_tokens = 4096
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 engine = "cloud"
 

src/openjarvis/evals/configs/taubench-gemma4-26b.toml

@@ -10,7 +10,7 @@ temperature = 0.7
 max_tokens = 4096
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 engine = "cloud"
 

src/openjarvis/evals/configs/taubench-gemma4-e4b.toml

@@ -10,7 +10,7 @@ temperature = 0.7
 max_tokens = 4096
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 engine = "cloud"
 

src/openjarvis/evals/configs/taubench-gpt54.toml

@@ -10,7 +10,7 @@ temperature = 0.7
 max_tokens = 4096
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 engine = "cloud"
 

src/openjarvis/evals/configs/taubench-kimi-k25.toml

@@ -17,7 +17,7 @@ temperature = 0.7
 max_tokens = 4096
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 engine = "cloud"
 

src/openjarvis/evals/configs/taubench-minimax-m25.toml

@@ -14,7 +14,7 @@ temperature = 0.7
 max_tokens = 4096
 
 [judge]
-model = "gpt-5-mini-2025-08-07"
+model = "claude-haiku-4-5-20251001"
 temperature = 0.0
 engine = "cloud"