feat(providers): add Prism provider

litellm/constants.py

@@ -928,6 +928,7 @@ openai_compatible_endpoints: Final[list] = [
     "https://api.meta.ai/v1",
     "https://api.cognition.ai/v1",
     "https://api.scx.ai/v1",
+    "https://api.prisminference.com/v1",
     "https://gigachat.devices.sberbank.ru/api/v1",
 ]
 
@@ -1000,6 +1001,7 @@ openai_compatible_providers: Final[list] = [
     "meta",  # Meta Model API (Muse Spark) - JSON-configured provider
     "cognition",
     "scx-ai",
+    "prism",
 ]
 
 OPENAI_AUDIO_TRANSCRIPTION_PROVIDERS: Final = frozenset({"openai"} | frozenset(openai_compatible_providers))

litellm/llms/openai_like/providers.json

@@ -200,5 +200,11 @@
       "temperature_max": 1.99
     },
     "supported_endpoints": ["/v1/chat/completions"]
+  },
+  "prism": {
+    "base_url": "https://api.prisminference.com/v1",
+    "api_key_env": "PRISM_API_KEY",
+    "api_base_env": "PRISM_API_BASE",
+    "supported_endpoints": ["/v1/chat/completions"]
   }
 }

litellm/model_prices_and_context_window_backup.json

@@ -74782,5 +74782,27 @@
         "supports_tool_choice": true,
         "supports_vision": false,
         "supports_web_search": false
+    },
+    "prism/deepseek-v4-flash": {
+        "cache_read_input_token_cost": 7e-08,
+        "input_cost_per_token": 1.4e-07,
+        "litellm_provider": "prism",
+        "max_input_tokens": 1000000,
+        "max_output_tokens": 393216,
+        "max_tokens": 393216,
+        "mode": "chat",
+        "output_cost_per_token": 2.8e-07,
+        "source": "https://prisminference.com/pricing",
+        "supported_endpoints": [
+            "/v1/chat/completions"
+        ],
+        "supports_function_calling": true,
+        "supports_native_streaming": true,
+        "supports_parallel_function_calling": true,
+        "supports_prompt_caching": true,
+        "supports_reasoning": true,
+        "supports_response_schema": true,
+        "supports_system_messages": true,
+        "supports_tool_choice": true
     }
 }

litellm/types/utils.py

@@ -4155,6 +4155,7 @@ class LlmProviders(str, Enum):
     PINSTRIPES = "pinstripes"
     COGNITION = "cognition"
     SCX_AI = "scx-ai"
+    PRISM = "prism"
     DARKBLOOM = "darkbloom"
     META = "meta"
     LITELLM_AGENT = "litellm_agent"

model_prices_and_context_window.json

@@ -74782,5 +74782,27 @@
         "supports_tool_choice": true,
         "supports_vision": false,
         "supports_web_search": false
+    },
+    "prism/deepseek-v4-flash": {
+        "cache_read_input_token_cost": 7e-08,
+        "input_cost_per_token": 1.4e-07,
+        "litellm_provider": "prism",
+        "max_input_tokens": 1000000,
+        "max_output_tokens": 393216,
+        "max_tokens": 393216,
+        "mode": "chat",
+        "output_cost_per_token": 2.8e-07,
+        "source": "https://prisminference.com/pricing",
+        "supported_endpoints": [
+            "/v1/chat/completions"
+        ],
+        "supports_function_calling": true,
+        "supports_native_streaming": true,
+        "supports_parallel_function_calling": true,
+        "supports_prompt_caching": true,
+        "supports_reasoning": true,
+        "supports_response_schema": true,
+        "supports_system_messages": true,
+        "supports_tool_choice": true
     }
 }

tests/test_litellm/llms/openai_like/test_prism_provider.py

@@ -0,0 +1,60 @@
+import pytest
+
+import litellm
+
+
+def test_prism_provider_resolution(monkeypatch: pytest.MonkeyPatch):
+    from litellm.litellm_core_utils.get_llm_provider_logic import get_llm_provider
+
+    monkeypatch.setenv("PRISM_API_KEY", "prism-test-key")
+
+    model, provider, api_key, api_base = get_llm_provider(
+        model="prism/deepseek-v4-flash",
+        custom_llm_provider=None,
+        api_base=None,
+        api_key=None,
+    )
+
+    assert model == "deepseek-v4-flash"
+    assert provider == "prism"
+    assert api_key == "prism-test-key"
+    assert api_base == "https://api.prisminference.com/v1"
+
+
+def test_prism_provider_keeps_explicit_credentials(monkeypatch: pytest.MonkeyPatch):
+    from litellm.litellm_core_utils.get_llm_provider_logic import get_llm_provider
+
+    monkeypatch.setenv("PRISM_API_KEY", "prism-env-key")
+
+    _, provider, api_key, api_base = get_llm_provider(
+        model="prism/deepseek-v4-flash",
+        custom_llm_provider=None,
+        api_base="https://prism.internal.example/v1",
+        api_key="prism-explicit-key",
+    )
+
+    assert provider == "prism"
+    assert api_key == "prism-explicit-key"
+    assert api_base == "https://prism.internal.example/v1"
+
+
+def test_prism_model_cost_and_capabilities():
+    from litellm.cost_calculator import cost_per_token
+
+    prompt_cost, completion_cost = cost_per_token(
+        model="prism/deepseek-v4-flash",
+        prompt_tokens=1_000_000,
+        completion_tokens=1_000_000,
+        custom_llm_provider="prism",
+    )
+    model_info = litellm.get_model_info("prism/deepseek-v4-flash")
+
+    assert prompt_cost == pytest.approx(0.14)
+    assert completion_cost == pytest.approx(0.28)
+    assert model_info["cache_read_input_token_cost"] == pytest.approx(7e-08)
+    assert model_info["max_input_tokens"] == 1_000_000
+    assert model_info["max_output_tokens"] == 393_216
+    assert model_info["supports_function_calling"] is True
+    assert model_info["supports_native_streaming"] is True
+    assert model_info["supports_reasoning"] is True
+    assert model_info["supports_response_schema"] is True