From efdd46161bcfdba6270148d9a584ab7f1b6b8707 Mon Sep 17 00:00:00 2001 From: Blake Anderson Date: Fri, 3 Jul 2026 22:00:38 -0500 Subject: [PATCH] feat(pricing): add openrouter/z-ai/glm-5.2 and sakana/fugu-ultra to cost map These two OpenRouter models were missing from the cost map, so LiteLLM reported $0 for streamed calls (which can't use OpenRouter's per-response cost). Add input/output and cache-read pricing plus context windows, sourced from the OpenRouter model API. fugu-ultra is multimodal (supports_vision); both support prompt caching and reasoning. Co-Authored-By: Claude Opus 4.8 (1M context) --- model_prices_and_context_window.json | 33 ++++++++++++++++++++++++++++ 1 file changed, 33 insertions(+) diff --git a/model_prices_and_context_window.json b/model_prices_and_context_window.json index a3dcdaed3f7..17355a83400 100644 --- a/model_prices_and_context_window.json +++ b/model_prices_and_context_window.json @@ -29989,6 +29989,39 @@ "supports_reasoning": true, "supports_tool_choice": true }, + "openrouter/z-ai/glm-5.2": { + "input_cost_per_token": 9.1e-07, + "output_cost_per_token": 2.86e-06, + "cache_read_input_token_cost": 1.69e-07, + "cache_creation_input_token_cost": 0.0, + "litellm_provider": "openrouter", + "max_input_tokens": 1048576, + "max_output_tokens": 128000, + "max_tokens": 128000, + "mode": "chat", + "source": "https://openrouter.ai/z-ai/glm-5.2", + "supports_function_calling": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_tool_choice": true + }, + "openrouter/sakana/fugu-ultra": { + "input_cost_per_token": 5e-06, + "output_cost_per_token": 3e-05, + "cache_read_input_token_cost": 5e-07, + "cache_creation_input_token_cost": 0.0, + "litellm_provider": "openrouter", + "max_input_tokens": 1000000, + "max_output_tokens": 128000, + "max_tokens": 128000, + "mode": "chat", + "source": "https://openrouter.ai/sakana/fugu-ultra", + "supports_function_calling": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_tool_choice": true, + "supports_vision": true + }, "openrouter/minimax/minimax-m2.1": { "input_cost_per_token": 2.7e-07, "output_cost_per_token": 1.2e-06,