From 8ec45c5c1be3cfb152effa157881e247fbec7e40 Mon Sep 17 00:00:00 2001 From: AdrianoDev Date: Mon, 11 May 2026 23:45:52 +0200 Subject: [PATCH] revert(config): rollback tier C a qwen-2.5-72b-instruct (qwen3-235b inferiore) MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Run controllo phase2-qwen25-control-001 (seed 42, stessa pipeline Phase 2, solo tier C switched) ha dimostrato che qwen-2.5-72b รจ qualitativamente SUPERIORE a qwen3-235b sul nostro workload: | metrica | qwen3-235b | qwen-2.5-72b | delta | | ----------------- | ---------- | ------------ | ----- | | max fitness | 0.0238 | 0.0311 | +30% | | median > 0 in gen | mai | 4 gen su 10 | -- | | entropy media | 0.199 | 0.85 | 4.3x | | genomi fit > 0 | 5 | 10 | 2x | | parse success | 97.7% | 100% | + | | durata | 50 min | 28 min | 0.56x | | LLM calls | 148 | 90 | 0.61x | | cost USD | 0.0223 | 0.0122 | 0.55x | Controintuitivo: 235B con context 262k era atteso superiore al 72B legacy. In pratica qwen3-235b in tier C produce strategie meno diverse, meno parsabili e meno ottimizzabili dal GA. Ripristinati prezzi cost_tracker tier C a 0.40/0.40 (qwen-2.5-72b). Co-Authored-By: Claude Opus 4.7 (1M context) --- .env.example | 2 +- README.md | 2 +- src/multi_swarm/config.py | 2 +- src/multi_swarm/llm/client.py | 2 +- src/multi_swarm/llm/cost_tracker.py | 2 +- tests/unit/test_config.py | 2 +- tests/unit/test_cost_tracker.py | 2 +- 7 files changed, 7 insertions(+), 7 deletions(-) diff --git a/.env.example b/.env.example index d88e1ee..cfe2c1b 100644 --- a/.env.example +++ b/.env.example @@ -14,7 +14,7 @@ OPENROUTER_BASE_URL=https://openrouter.ai/api/v1 LLM_MODEL_TIER_S=google/gemini-3-flash-preview LLM_MODEL_TIER_A=deepseek/deepseek-v4-flash LLM_MODEL_TIER_B=deepseek/deepseek-v4-flash -LLM_MODEL_TIER_C=qwen/qwen3-235b-a22b-2507 +LLM_MODEL_TIER_C=qwen/qwen-2.5-72b-instruct LLM_MODEL_TIER_D=openai/gpt-oss-20b # Run config diff --git a/README.md b/README.md index 5f76158..bf7b037 100644 --- a/README.md +++ b/README.md @@ -112,7 +112,7 @@ OPENROUTER_BASE_URL=https://openrouter.ai/api/v1 LLM_MODEL_TIER_S=anthropic/claude-opus-4-7 LLM_MODEL_TIER_A=anthropic/claude-sonnet-4-6 LLM_MODEL_TIER_B=anthropic/claude-sonnet-4-6 -LLM_MODEL_TIER_C=qwen/qwen3-235b-a22b-2507 +LLM_MODEL_TIER_C=qwen/qwen-2.5-72b-instruct LLM_MODEL_TIER_D=meta-llama/llama-3.3-70b-instruct ``` diff --git a/src/multi_swarm/config.py b/src/multi_swarm/config.py index c89c1e2..0ea0ca2 100644 --- a/src/multi_swarm/config.py +++ b/src/multi_swarm/config.py @@ -28,7 +28,7 @@ class Settings(BaseSettings): llm_model_tier_s: str = "google/gemini-3-flash-preview" llm_model_tier_a: str = "deepseek/deepseek-v4-flash" llm_model_tier_b: str = "deepseek/deepseek-v4-flash" - llm_model_tier_c: str = "qwen/qwen3-235b-a22b-2507" + llm_model_tier_c: str = "qwen/qwen-2.5-72b-instruct" llm_model_tier_d: str = "openai/gpt-oss-20b" openrouter_base_url: str = "https://openrouter.ai/api/v1" diff --git a/src/multi_swarm/llm/client.py b/src/multi_swarm/llm/client.py index cc1c61e..0af6913 100644 --- a/src/multi_swarm/llm/client.py +++ b/src/multi_swarm/llm/client.py @@ -17,7 +17,7 @@ from ..genome.hypothesis import HypothesisAgentGenome, ModelTier MODEL_TIER_S = "google/gemini-3-flash-preview" MODEL_TIER_A = "deepseek/deepseek-v4-flash" MODEL_TIER_B = "deepseek/deepseek-v4-flash" -MODEL_TIER_C = "qwen/qwen3-235b-a22b-2507" +MODEL_TIER_C = "qwen/qwen-2.5-72b-instruct" MODEL_TIER_D = "openai/gpt-oss-20b" OPENROUTER_BASE_URL = "https://openrouter.ai/api/v1" diff --git a/src/multi_swarm/llm/cost_tracker.py b/src/multi_swarm/llm/cost_tracker.py index 2fabe89..9db7e54 100644 --- a/src/multi_swarm/llm/cost_tracker.py +++ b/src/multi_swarm/llm/cost_tracker.py @@ -11,7 +11,7 @@ PRICE_PER_M_TOKENS: dict[ModelTier, dict[str, float]] = { ModelTier.S: {"input": 0.50, "output": 3.00}, ModelTier.A: {"input": 0.14, "output": 0.28}, ModelTier.B: {"input": 0.14, "output": 0.28}, - ModelTier.C: {"input": 0.071, "output": 0.10}, + ModelTier.C: {"input": 0.40, "output": 0.40}, ModelTier.D: {"input": 0.03, "output": 0.14}, } diff --git a/tests/unit/test_config.py b/tests/unit/test_config.py index 0f59012..f51ffb9 100644 --- a/tests/unit/test_config.py +++ b/tests/unit/test_config.py @@ -76,6 +76,6 @@ def test_settings_llm_model_defaults(monkeypatch: pytest.MonkeyPatch) -> None: assert s.llm_model_tier_s == "google/gemini-3-flash-preview" assert s.llm_model_tier_a == "deepseek/deepseek-v4-flash" assert s.llm_model_tier_b == "deepseek/deepseek-v4-flash" - assert s.llm_model_tier_c == "qwen/qwen3-235b-a22b-2507" + assert s.llm_model_tier_c == "qwen/qwen-2.5-72b-instruct" assert s.llm_model_tier_d == "openai/gpt-oss-20b" assert s.openrouter_base_url == "https://openrouter.ai/api/v1" diff --git a/tests/unit/test_cost_tracker.py b/tests/unit/test_cost_tracker.py index 0491b25..08733d3 100644 --- a/tests/unit/test_cost_tracker.py +++ b/tests/unit/test_cost_tracker.py @@ -4,7 +4,7 @@ from multi_swarm.llm.cost_tracker import CostTracker, estimate_cost def test_estimate_cost_tier_c(): cost = estimate_cost(input_tokens=1_000_000, output_tokens=1_000_000, tier=ModelTier.C) - assert cost == 0.071 + 0.10 + assert cost == 0.40 + 0.40 def test_estimate_cost_tier_b():