From 649a3c5e4e3433dc3b10eebcb8b966c3b1e0caef Mon Sep 17 00:00:00 2001 From: Benjamin Admin Date: Mon, 23 Mar 2026 09:12:01 +0100 Subject: [PATCH] perf: switch Pass 0b default model to Haiku 4.5 Benchmark shows Haiku is 2.5x faster than Sonnet at 5x lower cost for this JSON structuring task. Quality is equivalent. $142 vs $705 for 75K obligations, ~2.8 days vs ~7 days. Co-Authored-By: Claude Opus 4.6 --- backend-compliance/compliance/services/decomposition_pass.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/backend-compliance/compliance/services/decomposition_pass.py b/backend-compliance/compliance/services/decomposition_pass.py index d6752ba..cb329dd 100644 --- a/backend-compliance/compliance/services/decomposition_pass.py +++ b/backend-compliance/compliance/services/decomposition_pass.py @@ -39,7 +39,7 @@ logger = logging.getLogger(__name__) # --------------------------------------------------------------------------- ANTHROPIC_API_KEY = os.getenv("ANTHROPIC_API_KEY", "") -ANTHROPIC_MODEL = os.getenv("DECOMPOSITION_LLM_MODEL", "claude-sonnet-4-6") +ANTHROPIC_MODEL = os.getenv("DECOMPOSITION_LLM_MODEL", "claude-haiku-4-5-20251001") DECOMPOSITION_BATCH_SIZE = int(os.getenv("DECOMPOSITION_BATCH_SIZE", "5")) LLM_TIMEOUT = float(os.getenv("DECOMPOSITION_LLM_TIMEOUT", "120")) ANTHROPIC_API_URL = "https://api.anthropic.com/v1"