# DeepInfra and Parasail validate tool_choice but do not enforce it high · inference · probed live 2026-07-26 (DeepInfra) · 2026-08-02 (Parasail) · affects DeepInfra and Parasail · HTML version: [https://inferencecanary.com/tool-choice-unenforced](https://inferencecanary.com/tool-choice-unenforced) Both providers parse and validate the `tool_choice` field — send an invalid value or force a tool that isn't in `tools` and you get a 4xx error. Every surface signal says the capability exists. But a *valid* constraint has zero effect on what the model generates. Validation without enforcement is worse than ignoring the field outright: nothing warns you the guarantee is absent. ## What was measured on DeepInfra `tool_choice: "required"` on a prose-inviting prompt ([https://inferencecanary.com/assets/repro/tool-choice-required.deepinfra.json](https://inferencecanary.com/assets/repro/tool-choice-required.deepinfra.json)): ```json { "model": "google/gemma-4-31B-it", "messages": [ { "role": "user", "content": "Just say hello." } ], "tools": [ { "type": "function", "function": { "name": "log_event", "description": "Logs an event message.", "parameters": { "type": "object", "properties": { "message": { "type": "string" } } } } } ], "tool_choice": "required", "chat_template_kwargs": { "enable_thinking": false }, "temperature": 0, "max_tokens": 200 } ``` Received — HTTP **200**, prose, zero tool calls — and a self-contradicting envelope: ```json { "index": 0, "message": { "role": "assistant", "content": "Hello!", "reasoning_content": null, "name": null, "tool_calls": null }, "finish_reason": "tool_calls", "logprobs": null } ``` `finish_reason: "tool_calls"` with `tool_calls: null`: the constraint reached the layer that stamps the finish reason, and never reached decoding. Forcing `get_time` by name on a weather prompt, with `get_weather`, `get_news` and `get_time` offered ([https://inferencecanary.com/assets/repro/tool-choice-specific.deepinfra.json](https://inferencecanary.com/assets/repro/tool-choice-specific.deepinfra.json)): ```json { "model": "google/gemma-4-31B-it", "messages": [ { "role": "user", "content": "What's the weather in Paris right now?" } ], "tools": [ { "type": "function", "function": { "name": "get_weather", "description": "Current weather for a city.", "parameters": { "type": "object", "properties": { "city": { "type": "string" } } } } }, { "type": "function", "function": { "name": "get_news", "description": "Latest news on a topic.", "parameters": { "type": "object", "properties": { "topic": { "type": "string" } } } } }, { "type": "function", "function": { "name": "get_time", "description": "Current time in a timezone.", "parameters": { "type": "object", "properties": { "timezone": { "type": "string" } } } } } ], "tool_choice": { "type": "function", "function": { "name": "get_time" } }, "chat_template_kwargs": { "enable_thinking": false }, "temperature": 0, "max_tokens": 200 } ``` Received — HTTP **200**, one call, to the *wrong* tool: ```json { "index": 0, "message": { "role": "assistant", "content": "", "reasoning_content": null, "name": null, "tool_calls": [ { "id": "chatcmpl-tool-a52fe694592aca60", "type": "function", "function": { "name": "get_weather", "arguments": "{\"city\": \"Paris\"}" } } ] }, "finish_reason": "tool_calls", "logprobs": null } ``` `get_time` was forced; `get_weather` — the obvious tool for the prompt — came back. That is the signature of the constraint being dropped, not misread. ## What was measured on Parasail The same signature. Parasail enforced all three `tool_choice` conditions when probed on 2026-07-27; since 2026-08-01 it ignores `"required"` and forced-specific constraints, stable across three runs through 2026-08-02. `tool_choice: "required"` on the same prose-inviting prompt ([https://inferencecanary.com/assets/repro/tool-choice-required.parasail.json](https://inferencecanary.com/assets/repro/tool-choice-required.parasail.json)): ```json { "model": "google/gemma-4-31B-it", "messages": [ { "role": "user", "content": "Just say hello." } ], "tools": [ { "type": "function", "function": { "name": "log_event", "description": "Logs an event message.", "parameters": { "type": "object", "properties": { "message": { "type": "string" } } } } } ], "tool_choice": "required", "chat_template_kwargs": { "enable_thinking": false }, "temperature": 0, "max_tokens": 200 } ``` Received — HTTP **200**, prose, zero tool calls, and the same self-contradicting envelope: `finish_reason: "tool_calls"` over a message that contains no tool call at all: ```json { "index": 0, "message": { "role": "assistant", "content": "Hello!", "reasoning": null }, "finish_reason": "tool_calls", "logprobs": null } ``` Forcing `get_time` by name on the weather prompt ([https://inferencecanary.com/assets/repro/tool-choice-specific.parasail.json](https://inferencecanary.com/assets/repro/tool-choice-specific.parasail.json)): ```json { "model": "google/gemma-4-31B-it", "messages": [ { "role": "user", "content": "What's the weather in Paris right now?" } ], "tools": [ { "type": "function", "function": { "name": "get_weather", "description": "Current weather for a city.", "parameters": { "type": "object", "properties": { "city": { "type": "string" } } } } }, { "type": "function", "function": { "name": "get_news", "description": "Latest news on a topic.", "parameters": { "type": "object", "properties": { "topic": { "type": "string" } } } } }, { "type": "function", "function": { "name": "get_time", "description": "Current time in a timezone.", "parameters": { "type": "object", "properties": { "timezone": { "type": "string" } } } } } ], "tool_choice": { "type": "function", "function": { "name": "get_time" } }, "chat_template_kwargs": { "enable_thinking": false }, "temperature": 0, "max_tokens": 200 } ``` Received — HTTP **200**, one call, to the *wrong* tool — here even the finish reason gives up the pretense and reads `"stop"`: ```json { "index": 0, "message": { "role": "assistant", "content": "", "tool_calls": [ { "id": "chatcmpl-tool-90e0ba0e0470cacf", "type": "function", "function": { "name": "get_weather", "arguments": "{\"city\": \"Paris\"}" } } ], "reasoning": null }, "finish_reason": "stop", "logprobs": null } ``` And the validation half: `tool_choice: "banana"` → 400 *"Invalid value for `tool_choice`… Only named tools, "none", "auto" or "required" are supported"*; forcing a name not in `tools` → 400 *"The tool specified in `tool_choice` does not match any of the specified `tools`"*. The field is read and checked; the constraint still never reaches decoding. ## How this breaks the contract `"required"` obliges at least one tool call; a named `tool_choice` obliges a call to that exact function. Both providers return 200 on both and satisfy neither — and because the field is validated at the surface, every cheap client-side check (no 4xx, plausible finish_reason, well-formed JSON) reports success. ## Why it matters Structured extraction and guaranteed tool loops built on forced calls silently get prose, or a call to a different tool with different semantics. The failure only shows up when downstream code tries to use a result that never arrived. The Parasail history sharpens the risk: an integration validated against enforcing behavior can lose the guarantee later with no signal at the API surface. ## Run it yourself ``` curl "https://api.deepinfra.com/v1/openai/chat/completions" -H "Authorization: Bearer $KEY" \ -H "Content-Type: application/json" -d @tool-choice-required.deepinfra.json ``` ``` curl "https://api.parasail.io/v1/chat/completions" -H "Authorization: Bearer $KEY" \ -H "Content-Type: application/json" -d @tool-choice-required.parasail.json ``` Three checks on either provider: (1) `"required"` on "Just say hello." — watch prose come back under `finish_reason: "tool_calls"`; (2) forced `get_time` — watch `get_weather` return; (3) the validation half: change `tool_choice` to `"banana"`, or force a name not in `tools` — a 4xx either way (DeepInfra answers 400 and 422, Parasail 400 for both). That last pair proves the field is read, which makes the non-enforcement a choice, not a parsing gap.