Novita AI

1 graded endpoint serving gemma-4-31b · back to the matrix

Findings affecting Novita AI

• red harness Images in tool results: rejected, 500'd, or silently blinded on 7 of 11 providers
  The model reads images returned by tools — the reference provider delivers them, perception-judged, and three OpenAI-compat providers prove the standard shape works. Seven providers fail on schema choice, not model...
• red harness Novita silently ignores tool_choice
  "required" and forced-specific constraints are accepted and have no effect on decoding; only "none" is honored. The application consequence matches DeepInfra's: workflows that depend on forced tool calls silently receive...
• yellow harness Tool results replayed out of order are mis-paired on most providers
  The chat dialect's contract is that a tool message is matched to its call by tool_call_id, in any order. After two parallel calls, replaying the results reversed swaps the data between them on these providers — the...