AutoGPT is the vision of accessible AI for everyone, to use and to build on. Our mission is to provide the tools, so that you can focus on what matters.
chore(backend): resync drifted OpenRouter prices for six LLMs (#15621)
Nicholas Tindle119303cd9b76
diff --git a/autogpt_platform/backend/backend/data/block_cost_config_test.py b/autogpt_platform/backend/backend/data/block_cost_config_test.pyindex 821415fe30..7570a29ae9 100644--- a/autogpt_platform/backend/backend/data/block_cost_config_test.py+++ b/autogpt_platform/backend/backend/data/block_cost_config_test.py@@ -240,7 +240,7 @@ def test_open_router_bills_cost_usd_not_the_displayed_token_rate():def test_open_router_display_rate_matches_the_catalog_entry():"""The builder's "$X in / $Y out per 1M" label is the catalog credit ratedivided by the 150 cr/$ margin — SECRT-2701's corrected deepseek figures,- moved again on 2026-09-28 to track OpenRouter's live reprice.+ moved again on 2026-10-17 to track OpenRouter's live reprice."""entry = next(c@@ -248,5 +248,5 @@ def test_open_router_display_rate_matches_the_catalog_entry():if c.cost_filter.get("model") == LLMModel("deepseek/deepseek-chat"))assert entry.token_rate is not None- assert entry.token_rate.input_usd_per_1m == pytest.approx(0.2574)- assert entry.token_rate.output_usd_per_1m == pytest.approx(1.0287)+ assert entry.token_rate.input_usd_per_1m == pytest.approx(0.32)+ assert entry.token_rate.output_usd_per_1m == pytest.approx(0.89)diff --git a/autogpt_platform/backend/backend/data/llm_registry/catalog.py b/autogpt_platform/backend/backend/data/llm_registry/catalog.pyindex 1c0fe3de63..46c66c810f 100644--- a/autogpt_platform/backend/backend/data/llm_registry/catalog.py+++ b/autogpt_platform/backend/backend/data/llm_registry/catalog.py@@ -585,8 +585,8 @@ def _build_catalog() -> CatalogPayload:price_tier=1,cost=CatalogModelCost(run_credits=2,- input_credits_per_1m=38.61,- output_credits_per_1m=154.305,+ input_credits_per_1m=48,+ output_credits_per_1m=133.5,),),CatalogModel(@@ -740,8 +740,8 @@ def _build_catalog() -> CatalogPayload:price_tier=1,cost=CatalogModelCost(run_credits=3,- input_credits_per_1m=112.5,- output_credits_per_1m=562.5,+ input_credits_per_1m=123.75,+ output_credits_per_1m=618.75,),),CatalogModel(@@ -773,12 +773,12 @@ def _build_catalog() -> CatalogPayload:context_window=262144,max_output_tokens=16384,price_tier=1,- # Live OpenRouter rate as of 2026-09-23: $0.09/$0.34 per 1M,- # cache read $0.05/1M (drifted from the $0.14/$0.40 launch price).+ # Live OpenRouter rate as of 2026-10-17: $0.20/$0.40 per 1M,+ # cache read $0.05/1M (flagged by check_openrouter_prices.py).cost=CatalogModelCost(run_credits=1,- input_credits_per_1m=13.5,- output_credits_per_1m=51.0,+ input_credits_per_1m=30,+ output_credits_per_1m=60,cache_read_credits_per_1m=7.5,),),@@ -804,13 +804,13 @@ def _build_catalog() -> CatalogPayload:supports_json_output=True,supports_reasoning=True,supports_parallel_tool_calls=True,- # Live OpenRouter rate as of 2026-09-25: $0.021/$0.0616 per- # 1M, cache read $0.0042/1M (dropped from $0.06/$0.18,- # verified via check_openrouter_prices.py).+ # Live OpenRouter rate as of 2026-10-17: $0.06/$0.18 per+ # 1M, cache read $0.0042/1M (flagged by+ # check_openrouter_prices.py).cost=CatalogModelCost(run_credits=1,- input_credits_per_1m=3.15,- output_credits_per_1m=9.24,+ input_credits_per_1m=9,+ output_credits_per_1m=27,cache_read_credits_per_1m=0.63,),),@@ -1123,8 +1123,8 @@ def _build_catalog() -> CatalogPayload:price_tier=1,cost=CatalogModelCost(run_credits=1,- input_credits_per_1m=67.5,- output_credits_per_1m=337.5,+ input_credits_per_1m=75,+ output_credits_per_1m=375,),),CatalogModel(@@ -1204,7 +1204,7 @@ def _build_catalog() -> CatalogPayload:max_output_tokens=131072,price_tier=1,cost=CatalogModelCost(- run_credits=1, input_credits_per_1m=5.55, output_credits_per_1m=25.5+ run_credits=1, input_credits_per_1m=22.5, output_credits_per_1m=90),),CatalogModel(diff --git a/autogpt_platform/backend/backend/data/llm_registry/catalog_test.py b/autogpt_platform/backend/backend/data/llm_registry/catalog_test.pyindex c031c94138..aed9693c21 100644--- a/autogpt_platform/backend/backend/data/llm_registry/catalog_test.py+++ b/autogpt_platform/backend/backend/data/llm_registry/catalog_test.py@@ -166,13 +166,13 @@ def test_deepseek_chat_display_rate_tracks_openrouter():``poetry run python scripts/check_openrouter_prices.py`` and move bothsides together when OpenRouter reprices.- OpenRouter repriced again to $0.2574/$1.0287 live; these pins were moved- to match on 2026-09-28.+ OpenRouter moved back to $0.32/$0.89 live; these pins were moved to+ match on 2026-10-17."""chat = LLMModel("deepseek/deepseek-chat")assert TOKEN_COST[chat].model_dump() == {- "input": 38.61, # $0.2574/1M x 150 cr/$- "output": 154.305, # $1.0287/1M x 150 cr/$+ "input": 48.0, # $0.32/1M x 150 cr/$+ "output": 133.5, # $0.89/1M x 150 cr/$"cache_read": 0.0,"cache_creation": 0.0,}@@ -374,14 +374,14 @@ def test_gemini_4_argon_bills_at_authored_rates():def test_gemini_3_8_flash_bills_at_authored_rates():- """Gemini 3.8 Flash (OpenRouter, Google intro list price $0.75/$3.75- per 1M through 2026-12-31) — flat tier and per-1M projections must+ """Gemini 3.8 Flash (OpenRouter, live rate $0.825/$4.125+ per 1M as of 2026-10-17) — flat tier and per-1M projections mustmatch the authored catalog entry."""flash = LLMModel("google/gemini-3.8-flash")assert MODEL_COST[flash] == 3assert TOKEN_COST[flash].model_dump() == {- "input": 112.5,- "output": 562.5,+ "input": 123.75,+ "output": 618.75,"cache_read": 0.0,"cache_creation": 0.0,}@@ -627,15 +627,14 @@ def test_pareto_bills_at_authored_rates():def test_ling_3_0_flash_vl_bills_at_authored_rates():- """InclusionAI Ling 3.0 Flash VL (OpenRouter live rate $0.021/$0.0616- per 1M, $0.0042/1M cached input as of 2026-09-25 — dropped from- $0.06/$0.18) — flat tier and per-1M projections must match the+ """InclusionAI Ling 3.0 Flash VL (OpenRouter live rate $0.06/$0.18+ per 1M, $0.0042/1M cached input as of 2026-10-17) — flat tier and per-1M projections must match theauthored catalog entry."""ling = LLMModel("inclusionai/ling-3.0-flash-vl")assert MODEL_COST[ling] == 1assert TOKEN_COST[ling].model_dump() == {- "input": 3.15,- "output": 9.24,+ "input": 9.0,+ "output": 27.0,"cache_read": 0.63,"cache_creation": 0.0,}@@ -651,14 +650,14 @@ def test_ling_3_0_flash_vl_bills_at_authored_rates():def test_gemma_4_31b_it_bills_at_authored_rates():- """Google Gemma 4 31B (OpenRouter live rate $0.09/$0.34 per 1M,- $0.05/1M cached input as of 2026-09-23) — flat tier and per-1M+ """Google Gemma 4 31B (OpenRouter live rate $0.20/$0.40 per 1M,+ $0.05/1M cached input as of 2026-10-17) — flat tier and per-1Mprojections must match the authored catalog entry."""gemma = LLMModel("google/gemma-4-31b-it")assert MODEL_COST[gemma] == 1assert TOKEN_COST[gemma].model_dump() == {- "input": 13.5,- "output": 51.0,+ "input": 30.0,+ "output": 60.0,"cache_read": 7.5,"cache_creation": 0.0,}