Skip to content

A self-improving RLM agent for coding workflows and long-running autonomous tasks.

Rust

fix(ai): preserve nemotron ultra output limit (#420)

ktfefbf34e2f0a
diff --git a/packages/ai/scripts/generate-models.ts b/packages/ai/scripts/generate-models.ts
index cf976738d9..9ec7676163 100644
--- a/packages/ai/scripts/generate-models.ts
+++ b/packages/ai/scripts/generate-models.ts
@@ -144,7 +144,8 @@ const PRIME_INFERENCE_MODEL_METADATA: Record<string, PrimeInferenceModelMetadata
"minimax/minimax-m3": { contextWindow: 524288 },
"moonshotai/kimi-k2-0905": { contextWindow: 98304 },
"nvidia/nemotron-3-super-120b-a12b": { contextWindow: 262144, maxTokens: 4096 },
- "nvidia/nvidia-nemotron-3-ultra-550b-a55b": { contextWindow: 131072 },
+ // Preserve the existing output cap when OpenRouter leaves it unspecified.
+ "nvidia/nvidia-nemotron-3-ultra-550b-a55b": { contextWindow: 131072, maxTokens: 16384 },
// Enforced window is LARGER than OpenRouter's listing.
"qwen/qwen3-30b-a3b-instruct-2507": { contextWindow: 262144 },
// OpenRouter has no max_completion_tokens for the rest of these.